Cover page

Image Name: NGC 7496

Image Credit:NASA, ESA, CSA, Janice Lee (NSF’s NOIRLab)

The spiral arms of NGC 7496, one of a total of 19 galaxies targeted for study by the Physics at High Angular

resolution in Nearby Galaxies (PHANGS) collaboration, are ﬁlled with cavernous bubbles and shells overlapping

one another in this image from Webb’s Mid-Infrared Instrument (MIRI). These ﬁlaments and hollow cavities

are evidence of young stars releasing energy and, in some cases, blowing out the gas and dust of the interstellar

medium surrounding them.

For more information read

Managing Editor Chief Editor Editorial Board Correspondence

Ninan Sajeeth Philip Abraham Mulamoottil K Babu Joseph The Chief Editor

Ajit K Kembhavi airis4D

Geetha Paul Thelliyoor - 689544

Arun Kumar Aniyan India

Sindhu G

Journal Publisher Details

Publisher : airis4D, Thelliyoor 689544, India

Website : www.airis4d.com

Email : nsp@airis4d.com

Phone : +919497552476

Editorial

by Fr Dr Abraham Mulamoottil

airis4D, Vol.2, No.7, 2024

www.airis4d.com

This edition of the Journal starts with Dr.Balamuralidhar

Purushothaman, the former chief scientist at TCS Re-

search Bangalore. The article discusses the increas-

ing importance of remote sensing in Earth observation

and sustainability eﬀorts, driven by advancements in

satellite technology and private sector investments like

Tata Technologies’ TSAT. It highlights the challenge

of integrating domain knowledge with data-driven ap-

proaches in remote sensing, which is crucial for ac-

curate and scalable analysis. Traditional knowledge-

driven models struggle due to complex natural interac-

tions, while data-driven models beneﬁt from abundant

training data but lack interpretability. The article ex-

plores methods to integrate domain knowledge with

deep learning models, emphasizing enhanced perfor-

mance, interpretability, and generalizability through

feature engineering, model architecture adjustments,

and hybrid approaches. Future directions include over-

coming scalability issues and improving integration

methods to advance impactful remote sensing applica-

tions.

In ”Rotating Black Holes: Singularity, Event Hori-

zons and All That,” Ajit Kembhavi explores the unique

characteristics of Kerr black holes, contrasting them

with Schwarzschild black holes. He highlights that

while Schwarzschild black holes have a singularity at

a point, Kerr black holes exhibit a ring singularity in

their equatorial plane. The structure of the event hori-

zon in Kerr black holes is more complex due to their

spin, featuring two distinct event horizons. Kembhavi

discusses orbits within the Kerr metric, emphasizing

their non-planar nature and the inﬂuence of the black

hole’s spin on particle trajectories. He also touches

upon the extraction of rotational energy from within

the ergosphere of Kerr black holes, a phenomenon ﬁrst

proposed by Roger Penrose. This summary encapsu-

lates Kembhavi’s detailed exploration of the dynamics

and properties of rotating black holes.

The article ”X-ray Astronomy: Through Mis-

sions” by Aromal P discusses the evolution of X-ray

astronomy through various satellite missions. Early

experiments could not take direct X-ray images due

to the high energy of X-ray photons. This changed

with the launch of NASA’s HEAO-2 satellite, renamed

Einstein, in 1978, which used a grazing incidence X-

ray focusing telescope to capture high-resolution im-

ages. The 1980s saw the launch of several impor-

tant satellites: Japan’s Hintori (1981), which focused

on solar ﬂares; Tenma (1983), which studied X-ray

sources and discovered hot plasma and iron absorption

lines; ESA’s EXOSAT (1983), which identiﬁed quasi-

periodic oscillations in low mass X-ray binaries; and

Japan’s Ginga (1987), which analyzed the variability

of X-rays from active galaxies. These missions laid the

groundwork for the advancements in X-ray astronomy

in the subsequent decades.

In ”Exploring Stellar Clusters: Insights from Color-

Magnitude Diagrams, Part-2,” Sindhu G discusses the

Hertzsprung-Russell (HR) diagrams of globular clus-

ters, focusing on Messier 55. The HR diagrams of

globular clusters like Messier 55 provide critical in-

sights into stellar evolution, highlighting their age and

chemical composition. These clusters typically lack

high-mass main-sequence stars but contain many red

giants and blue stragglers. Messier 55’s HR diagram

shows it to be an older cluster with many evolved stars.

The article emphasizes the diﬀerences in stellar popula-

tions and chemical abundances between globular and

open clusters, noting that globular clusters generally

have fewer heavy elements.

In her article ”Precision Medicine: Role in Phar-

macogenomics and Drug Discovery (Part 2),” Geetha

Paul discusses the transformative impact of precision

medicine on healthcare. Pharmacogenomics uses ge-

netic information to tailor drug therapies to individual

patients, improving treatment eﬃcacy and minimiz-

ing adverse eﬀects. This approach, rooted in ancient

practices, is now advanced by modern genomics and

bioinformatics. The U.S. Precision Medicine Initiative

aims to integrate this approach widely, exempliﬁed by

the AllofUs study, which gathers extensive genetic and

health data to reﬁne treatments. Pharmacogenomics

encompasses the study of the entire genome’s eﬀect

on drug response, while pharmacogenetics focuses on

speciﬁc genes. Both ﬁelds enhance drug development

and patient care by predicting responses and reduc-

ing adverse reactions. Genetic variations inﬂuence

key processes such as drug absorption, distribution,

metabolism, and excretion, emphasizing the need for

personalized treatment plans. Advances in genomic

technologies promise further improvements in preci-

sion medicine, oﬀering more eﬀective and safer health-

care solutions.

The article ”CpG Sites: The Tiny Turn Signals of

Our DNA” by Jinsu Ann Mathew explains the critical

role of CpG sites in DNA. CpG sites, where a cyto-

sine nucleotide is followed by a guanine nucleotide,

play a signiﬁcant role in gene regulation through DNA

methylation. This methylation can turn genes on or

oﬀ, inﬂuencing cellular functions and health. CpG is-

lands, clusters of CpG sites, are often found near gene

promoters and help regulate gene expression. Under-

standing CpG sites and their methylation patterns can

provide insights into disease diagnosis and treatment,

oﬀering potential for personalized medicine.

The article ”Scaling Mechanisms in Software Sys-

tems” by Arun Aniyan discusses how large websites

and applications handle massive user loads eﬃciently

through scaling. It explains two primary scaling meth-

ods: vertical and horizontal scaling. Vertical scaling

involves enhancing a single server’s capacity by up-

grading its hardware (like processors, RAM, and SSDs)

and optimizing software (like operating systems and

databases). Although simpler, it has limitations such

as hardware constraints and potential downtime. Hor-

izontal scaling, preferred by large enterprises, involves

adding more servers to distribute the load, using tech-

niques like distributed architecture, data distribution,

and load balancing. This method oﬀers better fault

tolerance and ﬂexibility but is more complex to imple-

ment and maintain. The article concludes that eﬀective

scaling, often a mix of both methods, is essential for the

stability and performance of modern software systems.

iii

Contents

Editorial ii

I Artiﬁcial Intelligence and Machine Learning 1

1 Domain Knowledge Integration in Automated Analysis for Remote Sensing 2

1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2

1.2 Domain Knowledge for Remote Sensing and Analysis . . . . . . . . . . . . . . . . . . . . . . . . 2

1.3 Knowledge and Features . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3

1.4 Integrating Domain Knowledge with DL . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4

1.5 Future Directions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5

II Astronomy and Astrophysics 7

1 Black Hole Stories-10

Rotating Black Holes-Singularity, Event Horizons and All That 8

1.1 The Singularity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8

1.2 The Event Horizon . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8

1.3 Orbits in the Kerr Metric . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9

2 X-ray Astronomy: Through Missions 12

2.1 Accipere Pictures: Satellites with Optics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12

2.2 Satellites in 1980s . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13

3 Exploring Stellar Clusters: Insights from Color-Magnitude Diagrams, Part-2 15

3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15

3.2 The Hertzsprung-Russell Diagram of Globular Clusters . . . . . . . . . . . . . . . . . . . . . . . 15

3.3 The Hertzsprung-Russell Diagram of Globular Cluster Messier 55 . . . . . . . . . . . . . . . . . 16

III Biosciences 18

1 Precision Medicine

Role in Pharmacogenomics and Drug Discovery (Part 2) 19

1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19

1.2 Pharmacogenomics and Pharmacogenetics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20

1.3 Absorption of Drugs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21

1.4 Distribution of Drugs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21

1.5 Metabolism of Drugs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22

1.6 Metabolism Pathway-Phase 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22

1.7 Metabolism-Pathway Phase 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22

1.8 Metabolism Pathway-Phase 3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23

CONTENTS

2 CpG Sites: The Tiny Turn Signals of Our DNA 25

2.1 What are CpG Sites? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25

2.2 CpG Islands . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26

2.3 Why CpG Sites Matter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26

2.4 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27

IV General 29

1 Scaling Mechanisms in Software Systems 30

1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30

1.2 Vertical Scaling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30

1.3 Horizontal Scaling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31

1.4 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33

Part I

Artiﬁcial Intelligence and Machine Learning

Domain Knowledge Integration in Automated

Analysis for Remote Sensing

by Balamuralidhar P

airis4D, Vol.2, No.7, 2024

www.airis4d.com

1.1 Introduction

Earth observation through remote sensing is gain-

ing more attention with the emphasis on sustainability

goals committed by many countries. More and more

satellites are being put in orbit for advanced remote

sensing capabilities and applications. Private indus-

tries are also started investing in satellite based remote

sensing. Recently launched TSAT by Tata Technolo-

gies is one example. There are several other startups

including Pixxel, Galaxye are active in this area. This

brings abundance of earth observation data with re-

quire powerful and scalable automated processing to

derive essential information from remote sensing data

for a wide range of Earth system and socio-economic

researches and applications.

In an article published in previous issue, we have

reviewed the impact of AI in remote sensing and ob-

served the potential of deep neural network based archi-

tectures giving substantial performance enhancements.

However there was a challenge observed in integrat-

ing domain knowledge along with the data driven ap-

proaches which is an important requirement for remote

sensing problems. In this short article we review this

challenge and gather some insights / directions.

Generally we see knowledge-driven and data-driven

models as the dominant paradigms for spatio-temporal

modeling and spatio-temporal decision-making in the

remote sensing ﬁeld. Generally knowledge driven

models have the capability of encapsulating domain

knowledge better compared to data driven models.

However with suﬃcient training data available, data

driven approaches are relatively easier to handle.

In traditional approach, geoscience knowledge and

geoscience features (GK/GFs) from relevant expert ex-

perience and geoscience data as a series of logical infer-

ences and mathematical models. Physical models used

to represent interactions and relations between remote

observations and geoscientiﬁc parameters related to en-

vironmental domains, such as the atmosphere, ocean,

vegetation and hydrology. However, with the limited

prior knowledge about the interactions between sens-

ing and natural phenomenon made the development of

such physical models very diﬃcult. Also it aﬀected the

construction of accurate knowledge reasoning models

as well as knowledge discovery processes. Thus the

knowledge driven methods face substantial challenges

in large scale remote sensing applications. Neverthe-

less, knowledge-driven methods are still regarded as

one of the most important research directions because

of their advanced reasoning ability and interpretability.

1.2 Domain Knowledge for Remote

Sensing and Analysis

In automated analysis, domain knowledge is re-

quired to interpret the remote sensing data correctly

and often required human intervention. For example

in a land cover classiﬁcation analysis hill shade on one

side of a mountain is may get confused with water

bodies This situation may be improved if topography-

1.3 Knowledge and Features

related knowledge or features such as from digital ele-

vation models (DEMs) are introduced into the analysis.

Domain knowledge include experience and theories of

remote sensing interpretation and existing related sci-

entiﬁc laws.

Knowledge-driven methods and data driven DL

models are complementary, and an integrated approach

combining both could lead to better processing frame-

work for remote sensing.

An overview of some of the key knowledge fea-

tures that will be useful for remote sensing are outlined

below:

1.3 Knowledge and Features

Geographic features refer to intuitive features that

can be extracted directly from remote sensing data or

auxiliary data, such as texture features and geometric

features. Geographic knowledge include geoscience

processes such as hydrological, ecological and atmo-

spheric processes, and their internal driving mecha-

nisms, and geoscience attributes such as spatial distri-

bution and time variation, of ground objects.

They can be grouped into three types:

1.3.1 Spatial knowledge/features:

This deals with the spatiality of the target and can

be subdivided into four forms:

Spatial vision features

Spatial geometry features

Spatial distribution knowledge

Spatial relationship knowledge

The topological relation

1.3.2 Physical knowledge/features

Physical knowledge/features are related mainly to

the imaging principle and process of the sensor, as well

as the mechanism of electromagnetic waves reﬂected

and radiated by ground objects.

Physics based models are one form of knowledge.

Quantitative remote sensing physical models transform

electromagnetic wave information into useful knowl-

edge by establishing physically meaningful equations

and model. In plant growth, the terrestrial water cy-

cle, carbon and nitrogen cycles, land surface radiation,

and other geoscientiﬁc processes, such physical models

express the mechanisms of action between geoscience

variables and remote sensing observations and trans-

mission media through clear formulae.

Spectral reﬂectance characteristics of ground ob-

jects is another form of knowledge. Spectral index is

one such feature. For example, required biophysical

or environmental parameters, such as biomass, chloro-

phyll content and vegetation coverage can usually be es-

timated by establishing the relationship between them

and vegetation indices.

1.3.3 Regional knowledge/features

Regional knowledge/features are related to the

temporal evolution of ground objects, the physical geo-

graphical environment, regional diﬀerentiation and the

socio-economic conditions within a region. Here, we

subdivide regional knowledge/features into temporal

knowledge and environmental knowledge/features.

Temporal knowledge reﬂects primarily the changes

in ground objects over time. The most typical temporal

knowledge used in remotely sensed information extrac-

tion is the seasonal variation of plant growth.

Environmental knowledge/features refer to the sur-

rounding environmental conditions of the ground ob-

jects and knowledge of the spatial dependence from a

rather macroscopic perspective. Among the various

factors aﬀecting the formation of the land surface en-

vironment, topography is undoubtedly a dominant fac-

tor. In previous studies, DEMs, digital surface models

(DSMs), slope, aspect and related terrain indices were

used frequently as prior information and combined with

remote sensing data and other georeferenced GIS data

to establish classiﬁcation and analysis models, demon-

strated to signiﬁcantly increase the mapping accuracy

of LULC, vegetation type, soil type, and landslide sus-

ceptibility.

Another type of environmental knowledge/fea-

tures is embedded in historical products related to

physical regionalization, classiﬁcation of ground ob-

jects and other zoning.

1.4 Integrating Domain Knowledge with DL

Now given a high level understanding of domain

knowledge we can see how to integrate them with DL

models for remote sensing and analysis.

1.4 Integrating Domain Knowledge

with DL

It is well known that DL models are black box

models lacking explainability. In many cases getting

labelled remote sensing data is diﬃcult. Training with

incomplete labeled data makes DL models become un-

reliable. Also it makes generalization bit diﬃcult under

diverse geospatial contexts. This makes the processing

by domain experts inevitable in complex interpretation.

Integrating domain knowledge with deep learn-

ing (DL) models can signiﬁcantly enhance their per-

formance, interpretability, and generalizability. Here

are several ways to achieve this integration:

Feature Engineering:

Manual Feature Extraction: Incorporate domain-

speciﬁc features that are known to be relevant.

This can include statistical measures, domain-

speciﬁc transformations, or handcrafted features.

Feature Augmentation: Combine raw data with

domain knowledge-based features to create a

richer input dataset.

Data Preprocessing:

Domain-Speciﬁc Data Cleaning: Apply clean-

ing techniques that leverage domain knowledge

to handle noise, outliers, or missing values ef-

fectively.

Normalization and Scaling: Use domain-speciﬁc

criteria to normalize and scale data, ensuring it

is appropriately prepared for the DL model.

Model Architecture:

Custom Layers and Architectures: Design neu-

ral network architectures that reﬂect domain-

speciﬁc processes or constraints. For example,

in biology, embedding biological pathway infor-

mation into the network.

Embedding Domain Knowledge: Incorporate

domain knowledge directly into model architec-

ture through embeddings, attention mechanisms,

or by constraining certain parts of the network

based on known relationships.

Regularization:

Domain-Inspired Regularization: Use regular-

ization techniques informed by domain knowl-

edge, such as sparsity constraints for models in

areas where most features are irrelevant.

Transfer Learning: Pretrain models on related

domain-speciﬁc tasks and ﬁne-tune them on the

target task, leveraging prior knowledge.

Data Augmentation:

Domain-Speciﬁc Augmentation: Apply augmen-

tation techniques that make sense within the do-

main context, such as rotations in image data for

vision tasks or adding synthetic examples in a

manner consistent with domain rules.

Hybrid Models:

Combining DL with Traditional Methods: Use

traditional models (e.g., statistical models, rule-

based systems) in conjunction with DL models to

leverage both domain knowledge and the learn-

ing power of DL.

Ensemble Methods: Create ensembles that com-

bine DL models with other models that incorpo-

rate domain knowledge.

Knowledge Graphs and Ontologies:

Embedding Knowledge Graphs: Use embed-

dings of domain-speciﬁc knowledge graphs to

inform the model. These graphs can encode re-

lationships and hierarchies known in the domain.

Ontology-Based Constraints: Apply constraints

derived from ontologies to guide the learning

process and ensure outputs adhere to domain

rules.

Incorporating Expert Feedback:

Human-in-the-Loop: Integrate expert feedback

into the training loop, allowing for model adjust-

ments based on domain expertise.

Interactive Learning: Use interactive methods

where domain experts can iteratively reﬁne the

model by correcting errors or providing insights.

Explainability and Interpretability:

Model Interpretation Tools: Use tools and tech-

niques to make model decisions interpretable in

1.5 Future Directions

the context of domain knowledge, helping ex-

perts understand and validate the model’s rea-

soning.

Post-Hoc Analysis: Perform post-hoc analysis to

ensure that model predictions align with domain

knowledge, adjusting the model as necessary.

Rule-Based Integration:

Hard Constraints: Embed domain-speciﬁc rules

directly into the model as hard constraints, en-

suring outputs conform to known rules.

Soft Constraints: Use soft constraints to gen-

tly guide the model towards domain-consistent

behavior without enforcing strict rules.

By thoughtfully integrating domain knowledge

into DL models, one can achieve models that not only

perform better but are also more robust, interpretable,

and aligned with expert understanding.

1.5 Future Directions

Generally rule based knowledge integration is pop-

ular in remote sensing but there are some issues. Static

rule-based representation is rigid and lacks learnabil-

ity whereas DL models lacks in interpretability. The

algorithm of embedding rules into neural network is

relatively more complex; the rule-based approach may

not be suitable for scaling up.

Semantic network in the form of ontologies and

knowledge graphs is good at constructing the organic

relationship between various complex things, espe-

cially suitable for expressing relational knowledge. On-

tologies and knowledge graphs (KGs) are two represen-

tative modern implementations of semantic networks.

However the construction process of ontologies and

KGs is relatively complicated, and it is not very easy

to embed them into the optimization process of neu-

ral networks at present. However there are some good

progress is reported in this direction.

Rather than pixels, object based analysis is be-

ing pursued in Geographic object-based image anal-

ysis. Object characteristics like spectral information,

texture, geometry, the location distribution of ground

objects and contextual information are used as infor-

mation elements in the analysis. Application of Object

based CNN and Graph Convolution Networks (GCN)

have been successfully used here.

Physical models that are explicitly mathematical

formulas designed by human experts to better analyze

and understand processes or phenomena in the real-

world system, also used for knowledge integration.

However, there is a lack of sound physical models since

real physical processes are generally highly complex.

At the same time, high computational cost is often

needed to estimate the parameters of physical models.

Given this, DL models have been exploited to provide

a feasible and alternative scheme for approximating

complex physical processes.

DL itself is a powerful representation-learning

technology from raw data. When rich auxiliary geo-

science data and shallow GFs are available DL-based

knowledge discovery can be eﬀective.

Currently, beneﬁting from the ﬂexible architec-

tural design of DL model, the above approach can be

attempted via the following six main means.

Introducing auxiliary geoscience data/features as

input into DL models through individual neural

network processing stream.

Designing DL models with two neural network

branches to learn deep features separately from

input auxiliary geoscience data/features and re-

mote sensing images and then fuse them.

Leveraging attention modules to achieve feature

enhancement or feature selection when fusing

features.

Transferring the GK/GFs that pretrained net-

works contain to downstream tasks of remotely

sensed information extraction.

Combining diﬀerent DL models to excavate var-

ious GK/GFs for more accurate prediction.

Normalizing the loss function of the DL models

by the constraints of various GK/GFs extracted

from DL models to characterize geoscience fea-

tures.

Considering the importance of domain knowledge

integration in automated remote sensing analysis, the

methodologies for the same with DL models will be

signiﬁcant. There have been good progress in this

direction, however application of this in impactful re-

1.5 Future Directions

mote sensing applications require more attention and

accelerated eﬀort. This is also a rich area for future

research.

References

1. Ossai, Eze & Ao, Oliha. (2024). Integration of

Geoinformatics and Artiﬁcial Intelligence: En-

hancing Surveying Applications through Advanced

Data Analysis and Decision-Making.

2. Yong Ge, Xining Zhang, Peter M. Atkinson, Al-

fred Stein, Lianfa Li, Geoscience-aware deep

learning: A new paradigm for remote sensing,

Science of Remote Sensing, Volume 5,

3. Wang, Shu et al. “Geographic Knowledge Graph

(GeoKG): A Formalized Geographic Knowledge

Representation.” ISPRS Int. J. Geo Inf. 8

(2019): 184.

4. Hindustan Times

About the Author

Dr.Balamuralidhar Purushothaman is a former Chief Scientist at TCS Research Bangalore. Currently

he is continuing in TCS Research as a research advisor. He obtained his PhD from Aalborg university Denmark,

MTech from IIT Kanpur and BTech from TKM Engg Quilon. His research interest and contributions are in IoT,

Sensor Informatics, Remote Sensing and Robotics & AI. He has over 170 publications and 110 patents in these

technology and application areas. He has published a book on ‘IoT Technical Challenges and Solutions’.

Part II

Astronomy and Astrophysics

Black Hole Stories-10

Rotating Black Holes-Singularity, Event

Horizons and All That

by Ajit Kembhavi

airis4D, Vol.2, No.7, 2024

www.airis4d.com

In Black Hole Stories 9 (BHS-9), we introduced

the Kerr metric which corresponds to a black hole with

mass and spin. We then considered some of its prop-

erties in the Boyer-Lindquist coordinates t, r, θ, φ. In

this story, we will describe the singularity in the Kerr

black hole, the structure of the event horizon and some

concepts including the ergosphere and frame dragging.

1.1 The Singularity

In the Schwarzschild black hole, the singularity

is at the point r=0, which corresponds to the origin

and where the black hole is located. Since the black

hole has ﬁnite mass M but zero size, the density of

the matter there is inﬁnitely large. Consequently there

is a singularity in the geometry, with the curvature of

space-time too becoming inﬁnitely large.

In the Kerr black hole, by adopting suitable co-

ordinates it can be shown that the Boyer-Lindquist co-

ordinate r=0 corresponds to a disc in the equatorial

plane with radius a, where a is the angular momentum

parameter we introduced in BHS-9. The set of points

with r=0, θ=π/2 correspond to the edge of this disk,

which is a ring with radius a. It can be shown that the

Kerr metric has a space-time singularity at r=0, θ=π/2,

much like the singularity at r=0 in the Schwarzschild

case. But while the Schwarzschild singularity is at a

point, in the Kerr case, it is a ring with radius a in the

equatorial plane θ=π/2.

1.2 The Event Horizon

The event horizon for a Schwarzschild black hole

has a simple structure. It is a sphere with radius R

= 2GM/c

, from the interior of which no light ray can

exit to the outside world. Since a light ray cannot es-

cape, no particles with mass can escape either, because

their velocity is always smaller than the velocity of

light. Therefore, no signal whatsoever can be trans-

mitted from the inside the event horizon to the outside,

giving the enclosed black hole its name. But it is pos-

sible for light rays and particles to fall into the event

horizon from the outside, as we have seen in the anal-

ysis of their trajectories in BHS-6, using the concept

of the eﬀective potential. Once a trajectory enters the

event horizon, it inexorably falls towards the singular-

ity, eventually reaching it. A light ray or particle which

is outside the event horizon can escape to inﬁnity under

the right conditions.

In the three dimensional space of our experience,

the orientation of a surface at any point is determined

by the normal to the surface, which is perpendicular

to the surface at that point. Small displacements in

the surface at the point are deﬁned by tangents to the

surface. By deﬁnition, the tangents and normal are

perpendicular to each other. In contrast, the event

horizon is a null surface, which means that in the space-

time geometry, the direction of the normal lies in the

surface! This is possible because of the nature of the

1.3 Orbits in the Kerr Metric

metric of the four dimensional space-time geometry.

Because of the spin of the Kerr black hole, the

nature of the event horizon is more complex than in

the Schwarzschild case. It can be shown that the Kerr

geometry has two event horizons, which are at the

values of r given by

where µ=GM/c

and a=J/Mc, with J the angular

momentum of the black hole. We introduced the pa-

rameter a in BHS-9. For the zero angular momentum

case a=0, which is the Schwarzschild case, r

reduces

to the familiar r = R

= 2GM/c

, and r

= 0. It is also

apparent that the event horizon radii have real values

only for a < GM/c

, which is consistent with our state-

ment in BHS-6 that there is a maximum value of the

parameter a permitted, with a

max

=GM/c

. At this value

of a, both the radii of the Kerr event horizon have the

common value r

= r

= µ = GM/c

The structure of the Kerr black hole is illustrated

in Figure 1. As described above, there are two event

horizons, the inner one at r

and the outer one at r

The

ring singularity at r=0, in the equatorial plan deﬁned

by θ=π/2 is also shown. There are two other surfaces

shown, which are known as the inner surface of inﬁ-

nite redshift and the outer surface of inﬁnite redshift.

Since the outer surface is outside the outer event hori-

zon, except at the poles on the z-axis where the two

surfaces touch, a photon leaving this surface can travel

to inﬁnite distances from the black hole. The observed

wavelength of such a photon would be inﬁnitely large,

giving the surface its name.

The region between the outer surface of inﬁnite

redshift and the outer event horizon is known as the

ergosphere. This region has the interesting property

that no particle in the region can remain at rest with

respect to observers who are inﬁnitely far away. To

understand this, let us ﬁrst consider a particle which

is at a large distance from the black hole outside the

surfaces we have considered. If such a particle had

no angular momentum, it would fall towards the black

hole because of the gravity. But if it had thrusters

which could act in the direction opposite to the gravity,

then it could remain stationary at a given point under

the opposite actions of the two forces. We know that

Figure 1: The structure of the Kerr black hole. The

ﬁgure is described in the text. The ﬁgure is from

Lingyan Guan+, Journal of Physics: Conference Series

2022. Creative Commons Licence.

such a stationary situation cannot happen inside the

event horizon of a Schwarzschild black hole. There

the gravity is so strong that every particle necessarily

falls into the singularity. A similar situation applies

inside the ergosphere of a Kerr black hole. There no

particle can remain stationary at a given point, however

strong the thrusters available to it. Every particle is

dragged along, because the spin is not just the property

of the black hole, it is also a property of the space-time

around the black hole. This phenomenon is known as

frame dragging. It is possible for a particle to remain at

some constant r and θ by rotating around the black hole

with the axis of rotation being in the same direction as

the spin axis of the black hole.

Another interesting aspect of the Kerr black hole

is that it is possible to extract the rotational energy of

the black hole from within the ergosphere, which was

ﬁrst shown by Roger Penrose. The extracted energy

can power, for example, the active galactic nuclei asso-

ciated with the supermassive black holes in galaxies.

We will consider rotational energy extraction from ro-

tating black holes in a future story. This extraction is

diﬀerent from the energy released due to the infall of a

particle which we will consider below.

1.3 Orbits in the Kerr Metric

In BHS-6 to 8, we have studied in detail the orbits

of particles and photons in the Schwarzschild geom-

etry. One simplifying aspect there was the spherical

1.3 Orbits in the Kerr Metric

symmetry of the metric, which led to the conservation

of angular momentum. This meant that every orbits

was in a plane, with the direction of the angular mo-

mentum perpendicular to the plane. We introduced the

concept of an eﬀective potential, from which the nature

of possible orbits could be easily studied. We found

that for a given angular momentum, depending on the

energy of the particle (1) the orbits could be parabolic

or hyperbolic in which case the particles or photons

come from inﬁnite distance and return to it, (2) orbits

which come from inﬁnity and spiral into the black hole,

(3) for particles with mass bound precessing elliptical

orbits are possible, a special case of which was a stable

circular orbit when the particle was at the minimum

of the eﬀective potential and (4) an unstable circular

orbit when the particle was at the maximum of the ef-

fective potential. For photons, the only closed orbits

were unstable circular orbits at the maximum of the

eﬀective potential for photons, since the potential had

no minimum.

For massive particles in the Schwarzschild geom-

etry, we found in BHS-8 that as the angular momentum

of the particle decreases, the minimum in the eﬀec-

tive potential moves closer to the black hole, so that

the radius of the corresponding circular orbit becomes

smaller and the particle is more tightly bound. The

closest circular orbit occurs when the angular momen-

tum has the value L = 2

√

3GM. The radius of this in-

nermost stable circular orbit is r

ISCO

= 6GM/c

= 3R

Orbits with smaller radii are not possible because for

lesser angular momentum, the eﬀective potential has

no minimum. When a particle is in an elliptical orbit in

the Schwarzschild geometry, if energy and angular mo-

mentum are extracted from it, the orbit becomes more

circular and moves closer to the black hole. The pro-

cess can continue until r

ISCO

is reached, but no further

energy extraction is possible. If the angular momen-

tum decreases further, the particle will plunge into the

black hole. We mentioned in BHS-7 that a maximum

5.7 percent of the rest mass energy of the particle can

be extracted in the process.

The Kerr black hole has no spherical symmetry,

therefore the total angular momentum is not in general

conserved, and so an orbit does not have to lie in a

plane. The orbits can in general have very complex

behaviour and have been studied in detail by using

analytical and numerical techniques. One simplifying

factor is that the rotation of the black hole deﬁnes an

axis around which there is symmetry. The component

of the angular momentum in the direction of the spin

axis is therefore conserved and orbits in the equatorial

plane, which is perpendicular to the spin axis, will

remain conﬁned to it. For such orbits, the component of

the angular momentum in the direction of the black hole

spin is in fact the total angular momentum. The study

of these orbits is relatively simple and yet produces

very useful insights.

For equatorial orbits, taking into account the con-

servation of energy and angular momentum, the equa-

tion for the radial coordinate can be written in much

the same manner as for the Schwarzschild orbits we

considered in BHS-6 and 7. The equation can be

written in terms of the eﬀective potential, which, as

in the Schwarzschild case, has three terms: one pro-

portional to -1/r, which is the Newtonian part which

applies at large r, a second term proportional to 1/r

which is the centrifugal term, and a third term propor-

tional to -1/r

which is the general relativistic term. In

the Schwarzschild case the eﬀective potential depends

only on the square of the angular momentum of the

particle and not on the energy. But in the Kerr case

the potential depends on the angular momentum and

its square, as well as energy. The orbits therefore de-

pend on whether the initially the particle is moving in

the spin direction, i.e., corotating or counter to it, i.e.,

counterrotating. An interesting feature is that a parti-

cle which falls radially towards a Kerr black hole from

inﬁnity, conﬁned to the equatorial plane, sweeps out

an angle ∆ϕ, which is due to the frame dragging we

mentioned above. In the Schwarzschild case, an initial

radial orbit, which has zero angular momentum, would

always remain radial.

For the Kerr metric, the radial coordinate of the

innermost stable circular orbit in the equatorial plane

depends on the ratio of the angular momentum param-

eter a to GM/c

. This is shown in Figure 2.

In the diagram the dependence r

ISCO

/M on a/M is

shown for counterrotating as well as corotating orbits.

1.3 Orbits in the Kerr Metric

Figure 2: For a Kerr black hole, the r coordinate of the

innermost stable circular orbit r

ISCO

in the equatorial

plane, as a function of the angular momentum param-

eter a. In the ﬁgure r

ISCO

and a are indicated by their

ratios to GM/c

, with units chosen so that G=c=1. The

ﬁgure is from the book Gravity by James B. Hartle

For the Schwarzschild case which corresponds to a/M =

0, r

ISCO

= 6GM/c

, as discussed earlier. For the extreme

Kerr black hole with a=GM/c

, we have r

ISCO

= GM/c

while for the counterrotating case we have the much

larger value r

ISCO

= 9GM/c

. The smaller the value

of r

ISCO

, the more tightly bound is the particle in the

circular orbit, and therefore the greater is the energy

extracted from it. If a particle of rest mass m has

energy e per rest mass in its orbit, then the energy of

the particle is emc

. When the particle is at rest at

inﬁnity, the energy of the particle is just the rest mass

energy mc

. Therefore the fractional energy lost by the

particle per unit rest mass when it is in orbit is (mc

– emc

)/mc

= 1 – e. This is known as the binding

energy per unit rest mass of the particle, since it is the

energy per unit mass which would have to be provided

to the particle to take it to inﬁnity. A plot of the binding

energy per unit mass against a/M (in units of c=1, G=1)

is shown in Figure 3.

In Figure 3, for a = 0, we recover the Schwarzschild

value of 0.057 or 5.7 percent. The binding energy

keeps increasing with a for a given mass. For the ex-

treme value of a = GM/c

, the binding value is 0.42 or

42 percent. The eﬃciency of energy release in trans-

porting a particle from large distances to a black hole

to the innermost stable orbit, in the Kerr as well as

the Schwarzschild cases, is far more than the eﬃciency

with which energy is released during nuclear fusion

Figure 3: A plot of binding energy per unit mass, in the

innermost stable circular orbit, against a/M. The orbit

is in the equatorial plane of a Kerr black hole. Plots

for corotating and counterrotating orbits are shown.

in the interior of stars. This is just ˜0.7 percent for

the fusion of hydrogen to helium and ˜0.8 percent for

fusion all the way to iron, beyond which no release of

energy through fusion is possible. We will see in a fu-

ture story how such energy extraction is made possible

through the agency of an accretion disk. This energy

extraction because of the increase in the binding energy

of the particle, is diﬀerent from the rotational energy

extraction which we mentioned above. We will see

in a future story how such energy extraction is made

possible through the agency of an accretion disk.

About the Author

Professor Ajit Kembhavi is an emeritus

Professor at Inter University Centre for Astronomy

and Astrophysics and is also the Principal Investiga-

tor of the Pune Knowledge Cluster. He was the former

director of Inter University Centre for Astronomy and

Astrophysics (IUCAA), Pune, and the International

Astronomical Union vice president. In collaboration

with IUCAA, he pioneered astronomy outreach ac-

tivities from the late 80s to promote astronomy re-

search in Indian universities. The Speak with an

Astronomer monthly interactive program to answer

questions based on his article will allow young enthu-

siasts to gain profound knowledge about the topic.

X-ray Astronomy: Through Missions

by Aromal P

airis4D, Vol.2, No.7, 2024

www.airis4d.com

2.1 Accipere Pictures: Satellites with

Optics

So far, we have discussed the satellites, rocket pay-

loads, and balloon experiments used to study cosmic

X-rays. But none of them could take direct pictures of

an X-ray source. In other words, those experiments did

not have specialized optics to focus X-rays and capture

the image of these sources. One of the great challenges

in focusing X-rays is the high energy of the photons.

Whenever the X-ray photons come into contact with a

surface, they penetrate the surface rather than reﬂect

from it, making it hard to focus in a plane. However,

by reﬂecting at very high angles, called as grazing an-

gles, the low-energy X-rays can be focused. We can

discuss such telescopes in more detail in the upcoming

chapters. Before the success of the Uhuru mission,

proposals for a focusing telescope for X-rays were sub-

mitted. Later, a team led by Giaconni assembled to

develop the Large Orbiting X-ray Telescope (LOXT)

in 1970.

2.1.1 Einstein

Eﬀorts to develop an X-ray satellite with focus-

ing optics succeeded with the launch of the second

satellite of NASA’s High Energy Astrophysical Ob-

servatories (HEAO-2), which was renamed Einstein

after the launch. Einstein was launched in November

1978 and became the ﬁrst high-resolution imaging X-

ray telescope launched into space. The main science

objectives of the mission were imaging and spectro-

scopic studies of X-ray sources. For this, the satellite

was equipped with a Wolter Type-1 grazing incidence

X-ray focusing telescope and several other instruments.

Focusing enabled much better constraints in the posi-

tion of the sources and dramatically reduced the parti-

cle background. As this satellite was 100 times more

sensitive than the Uhuru, it enabled us to study the

diﬀuse emission, image extended objects, and a large

number of fainter sources. In its operational lifetime of

around two and a half years, Einstein provided the data

that revolutionized the ﬁeld of X-ray astronomy and

set the foundation for the great missions yet to come.

The scientiﬁc instruments in the satellite studied X-rays

in the energy range 0.2-20 keV. The instruments were

rotated one at a time into the focal plane of the op-

tics. Main instruments onboard of the satellite were an

Imaging Proportional Counter (IPC), a High Resolu-

tion Imager (HRI), a Solid State Spectrometer (SSSn)

and a Focal Plane Crystal Spectrometer (FPCS). Apart

from these instruments, the satellite also carried a non-

focusing Monitor Proportional Counter array (MPC)

and an Objective Grating Spectrometer (OGS).

With the help of the highly sensitive and focusing

instruments, morphological studies of supernova rem-

nants were conducted at high spatial resolution and

many faint sources were resolved in the M31 and Mag-

ellanic clouds. Einstein studied the X-ray emissions

from the hot intra-cluster medium in the clusters of

galaxies for the ﬁrst time, revealing the cooling inﬂow

and cluster evolution. It discovered that the X-ray jets

from Cen A and M87 aligned with the radio jets. Ein-

stein did the ﬁrst medium and deep X-ray survey, and it

was the ﬁrst NASA mission to have a Guest Observer

program.

2.2 Satellites in 1980s

Figure 1: Einstein view of galactic centre of An-

dromeda Galaxy (M31) Credits:NASA

2.2 Satellites in 1980s

2.2.1 Hintori

Hintori, Japanese for Phoenix, was the ﬁrst satel-

lite in the ASTRO satellite series dedicated to study

the solar phenomena. Hintori was launched in Febru-

ary 1981 from the Kagoshima Space Center in Japan

and was operational till October 1982. The satellite

specially studied the solar ﬂares and for this purpose,

it carried a Solar X-ray Telescope (SXT), which con-

sisted of two sets of bi-grid modulation collimators

for imaging the hard X-ray emission using the rotating

modulation collimator technique. In addition to SXT,

it had a Solar X-ray Aspect Sensor (SXA) to determine

the ﬂare position with an accuracy of 5 arcsec and a Soft

X-ray Crystal Spectrometer (SOX) for the spectroscopy

of X-ray emission lines from highly ionized Iron during

ﬂares. There were other instruments onboard to moni-

tor the solar ﬂares over a large energy band. The main

scientiﬁc output we got from the satellite includes the

time proﬁle and spectrum of the X-ray ﬂares and the

discovery of high-temperature phenomena reaching up

to 50 million Kelvin.

2.2.2 Tenma

Tenma was the second satellite in the ASTRO

series and the second Japanese satellite for X-ray as-

tronomy. It was launched in February 1983 and was

operational till November 1985. There were four in-

struments onboard Tenma: a Scintillation Proportional

Counter (SPC), an X-ray Focusing Collector (XFC), a

Transient Source Monitor (TSM), and a radiation belt

and Gamma-ray monitor. SPC was devoted to the spec-

tral and temporal studies of X-ray sources, while XFC

was used to observe the very soft X-ray sources. TSM

served as an X-ray monitor due to its wide ﬁeld of view.

The main discovery of Tenma was the study of the iron

line region of many classes of source. Apart from that,

it also discovered hot plasma of several millions of de-

grees located along the galactic center, Iron absorption

lines in the energy spectra of X-ray bursts, which was

red-shifted in the strong gravitational ﬁeld of a neutron

star and the identiﬁcation in low mass X-ray binaries

of X-ray emission region on the surface of the neutron

star and the accretion disc.

2.2.3 EXOSAT

The European Space Agency’s X-ray Observa-

tory was launched in May 1983 and was active for

three years in space. It was initially launched into

a highly eccentric orbit with a revolution period of

around 90.6 hr with an apogee and perigee of 191,000

km and 350 km, respectively. This orbit enabled undis-

turbed observation of the source for a long time. EX-

OSAT was equipped with focusing optics; it had Low

Energy Telescopes composed of two identical Wolter

Type-1 telescopes. Each of the telescopes can operate

in imaging mode as well as in spectroscopy mode. Be-

sides the telescope, it had Medium Energy instruments

(ME) and the Gas Scintillation Proportional Counter

(GSPC). During the performance veriﬁcation phase,

the Position Sensitive Detector(PSD) associated with

the two Low Energy telescopes had failed, and after

the six months of operation, one of the channel plates

failed. However, the Low Energy Telescope functioned

up to the end. The remarkable discovery of EXOSAT

was the quasi-periodic oscillations (QPOs) of low mass

X-ray binaries and the soft excess from AGN. It also

discovered many new transient sources during its op-

eration period.

2.2 Satellites in 1980s

Figure 2: EXOSAT observation of Cyg X-1 Cred-

its:HEASARC

2.2.4 Ginga

The third satellite of the ASTRO series, ASTRO-

C, was launched in February 1987 and was renamed

Ginga, which means galaxy in Japanese. It was a

collaboration between the Japanese institutions, the

University of Leicester and the Rutherford-Appleton

Laboratory in the UK, and the Los Alamos National

Laboratory in the USA. This mission was operational

until November 1991. The primary mission objective

was to study the time variability of X-rays from active

galaxies such as Seyfert galaxies, BL Lac objects, and

quasars in the energy range 1.5-30 keV. Accurate timing

analysis of the galactic X-ray sources was also one of

the mission objectives. Payloads consisted of a Large

Area Proportional Counter (LAC), an All Sky Monitor

(ASM), and a Gamma-ray Burst Detector (GBD).

So far, we have discussed the satellites launched

in the 1980s. After this period, the X-ray astronomy

took a giant leap. From 1990 to the present age is

referred to as the golden period of X-ray astronomy,

where groundbreaking results were produced from the

data collected from even more sensitive and capable

satellites. We will discuss the missions that made these

breakthroughs in X-ray astronomy in the upcoming ar-

ticle.

References

Santangelo, Andrea and Madonia, Rosalia and

Piraino, Santina A Chronological History of X-

Ray Astronomy Missions. Handbook of X-ray

and Gamma-ray Astrophysics.ISBN 9789811645440

HEAO 2

X-ray Impact, Riccardo Giacconi Einstein Ob-

servatory (HEAO-2)

About the Author

Aromal P is a research scholar in Depart-

ment of Astronomy, Astrophysics and Space Engi-

neering (DAASE) in Indian Institute of Technology

Indore. His research mainly focuses on studies of

Thermonuclear X-ray Bursts on Neutron star surface

and its interaction with the Accretion disk and Corona.

Exploring Stellar Clusters: Insights from

Color-Magnitude Diagrams, Part-2

by Sindhu G

airis4D, Vol.2, No.7, 2024

www.airis4d.com

3.1 Introduction

In the earlier article titled ”Exploring Stellar Clus-

ters: Insights from Color-Magnitude Diagrams, Part-

1,” we explored distinctive characteristics of HR dia-

grams in globular clusters compared to younger open

clusters. We also examined challenges encountered

when constructing HR diagrams for globular clusters.

In this article we explore the HR diagram of globular

star cluster Messier 55.

3.2 The Hertzsprung-Russell

Diagram of Globular Clusters

Star clusters are dynamically bound groups of

hundreds to thousands of stars that share the same for-

mation history, age, and composition. These clusters,

which contain stars with a wide range of masses, pro-

vide invaluable insights into stellar evolution through

their Hertzsprung-Russell diagram(HRD). To extract

this information, precise distances and high-accuracy

photometric data are necessary. A globular cluster is

a densely packed group of thousands to millions of

stars bound together by gravity. Figure: 1 depicts a

Hertzsprung-Russell diagram. Figure: 2 shows major

branches on the HR diagram.

Figure: 3 presents a combined Hertzsprung-Russell

diagram for 14 globular clusters, created using data

from the second Gaia data release. The primary diﬀer-

ences depicted are in chemical composition, indicated

by color: stars from metal-poor clusters are shown in

Figure 1: The HR Diagram. (Image Credit: Chandra

X-ray Observatory )

Figure 2: Major Branches on the H-R Diagram. (Im-

age Credit: Chandra X-ray Observatory)

3.3 The Hertzsprung-Russell Diagram of Globular Cluster Messier 55

Figure 3: Combined HRD for 14 globular clusters

made with the second Gaia data release data. (Im-

age Credit: ESA/Gaia/DPAC, Carine Babusiaux and

co-authors of the paper ”Gaia Data Release 2: Obser-

vational Hertzsprung-Russell diagram” )

Figure 4: Globular star cluster Messier 55 is located

at a distance of about 17000 light-years from Earth.

(Image Credit: ESO/J. Emerson/VISTA )

blue, while stars from higher-metallicity clusters are

shown in red. The image illustrates how these diﬀer-

ences inﬂuence stellar evolution.

3.3 The Hertzsprung-Russell

Diagram of Globular Cluster

Messier 55

Figure: 4 shows the globular star cluster Messier

55. The Hertzsprung-Russell diagram for the globular

star cluster Messier 55 is presented in Figure: 5. The

HR diagram of M55 demonstrates that globular clus-

ters are typically older than open clusters, as indicated

Figure 5: HR diagram of M55. (Image Credit: AT-

NF/: Hubble Heritage Team (STScI / AURA), A. Cool

(SFSU) et al., NASA )

by a higher proportion of evolved stars. Additionally,

globular clusters lack high-mass stars that remain on

the main sequence. Interestingly, upon examining the

diagram above, you can observe a cluster of hot stars

that seem to be positioned on the main sequence be-

yond the turnoﬀ point. These stars are identiﬁed as

blue stragglers. Within the dense stellar environments

of globular clusters, astronomers hypothesize that cer-

tain stars can merge and amalgamate. This merging

process results in a combined mass that causes the new

star to appear hotter (bluer) and brighter compared to

the majority of stars in the cluster.

HRD of globular clusters lacks main sequence

stars of types O, B, A, and F but includes many red

giants. The brightest stars in a globular cluster are at the

tip of the red giant branch in the HRD, which explains

why the bright stars in color images of these clusters

appear red, as seen in the image above. Additionally,

you can observe stars on the horizontal branch , the

asymptotic giant branch, and even some stars with the

colors and magnitudes of F-type stars, though there are

far fewer of these compared to the G-type stars just

below and to the right of them on the main sequence.

Figure: 6 illustrates a comparison between two

HR diagrams of a typical globular cluster. It shows

the absence of main sequence stars of types O, B, or

A, a densely populated lower main sequence and red

3.3 The Hertzsprung-Russell Diagram of Globular Cluster Messier 55

Figure 6: LEFT: Schematic HR diagram of a globu-

lar cluster. RIGHT: Color Magnitude (HR) Diagram

of M55 from real data. (Image Credit: LEFT: Penn

State Astronomy & Astrophysics, RIGHT: Astronomy

Picture of the Day)

giant branch, as well as a relatively signiﬁcant pop-

ulation of white dwarf stars. When we observe the

stars in globular clusters using spectroscopy, we can

measure the abundance of chemical elements in their

atmospheres. This reveals another diﬀerence between

globular clusters and open clusters. In globular clus-

ters, the abundance of elements heavier than helium is

typically only 1-10% of their abundance in the Sun and

in stars found in open clusters. In the upcoming article,

we will delve into the Hertzsprung-Russell diagram of

various clusters.

References:

Star Clusters

Globular Clusters

Globular cluster

HR Diagram, Star Clusters,and Stellar Evolution

The Colour-Magnitude-Diagram of Globular Clus-

ters and Stellar Evolution

Gaia DR2 Hertzsprung-Russell diagram of glob-

ular clusters

Messier 55 HR Diagram

Pulsating Variable Stars and the Hertzsprung-

Russell (H-R) Diagram

About the Author

Sindhu G is a research scholar in Physics

doing research in Astronomy & Astrophysics. Her

research mainly focuses on classiﬁcation of variable

stars using diﬀerent machine learning algorithms. She

is also doing the period prediction of diﬀerent types

of variable stars, especially eclipsing binaries and on

the study of optical counterparts of X-ray binaries.

Part III

Biosciences

Precision Medicine

Role in Pharmacogenomics and Drug

Discovery (Part 2)

by Geetha Paul

airis4D, Vol.2, No.7, 2024

www.airis4d.com

1.1 Introduction

The integration of precision medicine into phar-

macogenomics and drug discovery is transforming health-

care. Pharmacogenomics is a ﬁeld that leverages ge-

netic information to tailor drug therapies to individual

patients. By understanding how genetic variations in-

ﬂuence drug response, pharmacogenomics enhances

treatment eﬃcacy and minimises adverse eﬀects, ush-

ering in a new era of customised healthcare. By

leveraging genetic information, healthcare providers

can oﬀer more eﬀective and personalised treatments,

while pharmaceutical companies can develop safer and

more targeted drugs. The ongoing advancements in

genomics, data analytics, and bioinformatics are driv-

ing this revolution, promising a future where medical

treatments are tailored to the unique genetic makeup

of each individual. In modern medicine, the concept

of one-size-ﬁts-all treatments is rapidly being replaced

by a more personalised approach known as precision

medicine. Pharmacogenomics explores how an indi-

vidual’s genetic makeup inﬂuences their response to

medications. Variations in genes can aﬀect the ab-

sorption, metabolism, and eﬃcacy of drugs and the

likelihood of adverse reactions. By identifying these

genetic diﬀerences, healthcare providers can prescribe

medications that are most likely to be eﬀective and

safest for each patient.

While precision medicine may seem like a mod-

ern innovation, its roots can be traced back 2,500 years

ago to ancient Greece, speciﬁcally to Hippocrates, of-

ten hailed as the “Father of Western Medicine”. In

an enlightening article by Sykiotis et al., Hippocrates’s

approach is highlighted, emphasising his belief in the

uniqueness of diseases and the importance of prescrib-

ing “diﬀerent drugs to diﬀerent patients”. When deter-

mining the appropriate treatment regimen, Hippocrates

considered various factors, such as a patient’s constitu-

tion, age, physical attributes, and seasonal inﬂuences.

By embracing this personalised approach, healthcare

providers can optimise treatment outcomes, minimise

adverse eﬀects, and ultimately enhance patient care.

Recognising this potential, the U.S. Congress ap-

proved a $2 billion increase in NIH funding for the

Precision Medicine Initiative (PMI) in 2016. The

long-term goals of the PMI aim to integrate precision

medicine into all facets of health and healthcare on a

large scale. A key component of this initiative is the

AllofUs study, which will involve a cohort of at least

one million volunteers from across the United States.

These participants will provide genetic data, biological

samples, and health-related information. Researchers

will utilise this wealth of data to investigate a broad

spectrum of diseases, aiming to predict disease risk

better, understand disease occurrence, and develop im-

proved diagnostic and treatment strategies.

Variations among individuals stem primarily from

diﬀerences in their genetic makeup. These genetic

1.2 Pharmacogenomics and Pharmacogenetics

variances contribute to not all patients responding uni-

formly to the same medications. In the United States,

adverse drug reactions (ADRs) represent a signiﬁcant

public health concern, ranking as the fourth leading

cause of death. Annually, prescription drugs are linked

to approximately 2.74 million ADRs and contribute to

about 128,000 deaths. The economic impact of ADRs

is substantial, totalling $136 billion annually, exceed-

ing the combined costs of cardiovascular and diabetes

care. Furthermore, ADRs are responsible for one out

of every ﬁve injuries or deaths among hospitalised pa-

tients each year.

1.2 Pharmacogenomics and

Pharmacogenetics

Pharmacogenomics (PGx) broadly encompasses

studying how an individual’s entire genome (all of

their genes) aﬀects their drug response. It considers

how variations across multiple genes inﬂuence drug

metabolism, eﬃcacy, and adverse reactions. Pharma-

cogenomics typically involves large-scale analysis of

genetic data to identify patterns and associations be-

tween genetic variations and drug responses across

populations. Meanwhile, pharmacogenetics (PGt)

focuses more narrowly on the eﬀect of speciﬁc genetic

variations or mutations within single genes on drug re-

sponse. It examines how variations in a particular gene

(e.g., a speciﬁc enzyme involved in drug metabolism)

can impact an individual’s ability to metabolise or re-

spond to a drug.

Integrating pharmacogenomics and pharmacoge-

netics into clinical practice promises to improve pa-

tient outcomes, enhance medication safety, and re-

duce healthcare costs by minimising trial-and-error

prescribing. As genomic technologies continue to ad-

vance and become more accessible, the ﬁeld of phar-

macogenomics is expected to expand, oﬀering new in-

sights into how genetic information can be used to tailor

treatments to individual patients eﬀectively.

Pharmacogenomics plays two major roles in pre-

cision medicine. First, it guides pharmaceutical com-

panies in drug discovery and development. Second, it

(Image

courtesy:https://docs.google.com/document/d/11JZnhGcrKRNCYrZQuv5JeITcBkA B6JJi-

ls I7kozg)

Figure 1: The ﬁgure illustrates the inter-individual

variability with drug response.

guides physicians in selecting the right drug for patients

based on their genetic makeup, avoiding ADR, and in

maximising drug eﬃcacy by prescribing the right dose.

The Human Genome Project (HGP) completion

in April 2003 marked a signiﬁcant milestone in scien-

tiﬁc achievement, providing a comprehensive map of

the human genetic blueprint. This monumental eﬀort

revealed that humans possess approximately 20,500

genes, with an astonishing 99.5 percent of these genes

shared among individuals. The remaining 0.5 percent

comprises genetic variations contributing to diverse

traits such as eye colour, blood group, and susceptibil-

ity to certain diseases.

Among the most prevalent genetic variations iden-

tiﬁed are single nucleotide polymorphisms (SNPs), of-

ten abbreviated as “snips”. SNPs are single base pair

diﬀerences in DNA sequence that occur in at least 1

percent of the population. These variations act as bio-

logical markers throughout the genome and play pivotal

roles in determining an individual’s response to med-

ications, susceptibility to environmental factors like

toxins, and risk of developing various diseases.

The human genome is estimated to contain ap-

proximately 11 million SNPs, occurring on average

once every 1,300 base pairs. This abundance of SNPs

underscores their importance as genetic markers for

understanding human diversity and disease suscepti-

bility. By studying SNPs, researchers can uncover

genetic factors inﬂuencing how individuals metabolise

drugs, respond to therapies, and interact with their en-

vironment.

In addition to SNPs, another signiﬁcant category

of genetic variation is structural variations (SVs). SVs

involve larger changes in DNA sequence, including

1.3 Absorption of Drugs

(Image

courtesy:https://docs.google.com/document/d/11JZnhGcrKRNCYrZQuv5JeITcBkA B6JJi-

ls I7kozg)

Figure 2: The ﬁgure illustrates the pharmacokinetics

(PK) process.

deletions, insertions, tandem repeats, inversions, and

copy number variations (CNVs). These structural changes

can profoundly aﬀect gene expression, protein func-

tion, and disease susceptibility, further enriching our

understanding of genetic diversity and individual health

proﬁles.

The insights gained from the HGP (Human Genome

Project) and subsequent genomic research have paved

the way for personalised medicine, where treatments

can be tailored based on an individual’s unique genetic

proﬁle. Pharmacogenomics, for example, leverages

knowledge of genetic variations such as SNPs to op-

timise drug therapy by predicting how patients will

respond to medications and minimising the risk of ad-

verse reactions. Future advancements in genomic tech-

nologies and data analysis are expected to deepen our

understanding of how genetic variations contribute to

human health and disease. As genomic research unrav-

els the complexities of the human genome, integrating

this knowledge into clinical practice holds tremendous

promise for improving healthcare outcomes and ad-

vancing personalised approaches to medicine.

To successfully develop personalised dosing reg-

imens for patients, understanding pharmacokinetics

(PK), pharmacodynamics (PD), and PGx is important.

Every drug enters the body through absorption, distri-

bution, metabolism, and excretion (ADME). The sum

of all these processes is PK, which determines how

much of the drug is needed to reach the site of action

for an eﬀective therapeutic outcome.

PK describes a drug’s absorption, distribution,

metabolism, and excretion, inﬂuencing its concentra-

tion and duration in the body. PD, conversely, details

how a drug interacts with its target cells or tissues to

produce its eﬀects. The delicate balance between PK

and PD is crucial: it ensures a drug achieves its in-

tended therapeutic eﬀect while minimising potential

adverse reactions. Genetic polymorphisms can disrupt

this balance, altering how individuals metabolise drugs

and how their bodies respond to treatment. Pharma-

cogenetics (PGx) harnesses this genetic information to

personalise medication regimens, considering individ-

ual metabolism and drug response variations. By in-

tegrating PGx data with factors like environment, diet,

and health status, personalised medicine aims to opti-

mise treatment outcomes and enhance patient safety.

1.3 Absorption of Drugs

Absorption is the process by which a drug moves

from its administration site into the bloodstream. This

intricate process involves various membrane-bound drug

transporters, such as P-glycoprotein (P-gp, MDR1) and

multidrug resistance (MDR) transporters encoded by

the ABC genes. For instance, the ABCB1 gene, re-

sponsible for encoding P-gp, exhibits over 50 diﬀerent

single nucleotide polymorphisms (SNPs) that vary in

prevalence across diﬀerent ethnic groups. These ge-

netic variations signiﬁcantly inﬂuence the bioavailabil-

ity of the drug, which is the proportion of the adminis-

tered drug that reaches the bloodstream or its intended

site of action.

1.4 Distribution of Drugs

The volume of distribution (Vd) of a drug is a

theoretical concept that describes the extent to which

a drug disperses into body compartments and tissues,

inﬂuenced by the drug’s physical-chemical properties.

This volume is aﬀected by physiological factors such as

body mass index (BMI) and fat deposits, with higher

BMI and fat content increasing the Vd for lipophilic

drugs. Additionally, Vd depends on pharmacogenomics

(PGx) for distribution to speciﬁc compartments like

the brain and breast milk, which rely on transporter

1.5 Metabolism of Drugs

genes such as ABC. Overexpression of these genes

can lead to drug resistance by increasing drug eﬄux

from cells. Polymorphisms in genes like ABCB1 and

SLCO1B1 also impact drug distribution. ABCB1 af-

fects P-glycoprotein function, inﬂuencing drug pene-

tration across the blood-brain barrier, while SLCO1B1

aﬀects hepatic drug uptake, altering distribution and

clearance. Thus, drug distribution is a complex inter-

play of chemical properties, physiological factors, and

genetic variations.

1.5 Metabolism of Drugs

Drug metabolism is the metabolic breakdown of

drugs, usually through specialised enzymatic systems.

Most drug metabolism occurs in the liver and intestine.

Drug metabolism is divided into three phases.

1.6 Metabolism Pathway-Phase 1

Oxidation (via cytochrome P450), reduction, and

hydrolysis reactions

Conversion of a parent drug to more polar (wa-

ter soluble) active metabolites by unmasking or

inserting a polar functional group (-OH, -SH,

-NH2).

The metabolism of drugs is a critical process that often

involves various biochemical reactions. One primary

drug metabolism pathway is oxidation, primarily me-

diated by cytochrome P450 enzymes. These enzymes

catalyse the addition of oxygen to the drug molecule,

making it more polar and water-soluble. Another sig-

niﬁcant pathway is reduction, where the drug under-

goes a gain of electrons, often leading to the addition

of hydrogen atoms, which can also increase the drug’s

polarity. Additionally, hydrolysis reactions play a cru-

cial role in drug metabolism. These reactions involve

the cleavage of chemical bonds in the drug molecule

by adding water, further enhancing its solubility.

The ultimate goal of these metabolic reactions—

oxidation, reduction, and hydrolysis—is to convert the

parent drug into more polar (water-soluble) active metabo-

lites. This is achieved by unmasking or inserting polar

functional groups such as hydroxyl (-OH), sulfhydryl

(-SH), or amino (-NH2) groups into the drug molecule.

By increasing the polarity of the metabolites, the body

can more easily excrete them via renal or biliary routes,

thus facilitating the drug’s elimination and reducing its

potential toxicity. This metabolic transformation is es-

sential for the drug’s therapeutic eﬃcacy and safety

proﬁle, ensuring it reaches its intended target in the

body and is eventually cleared from the system.

1.7 Metabolism-Pathway Phase 2

Glucuronidation, acetylation, and sulfation reac-

tions.

“Conjugation reactions” that increase the water

solubility of a drug with a polar moiety glu-

curonate, acetate, and sulphate, respectively.

Conversion of a parent drug to more polar (water

soluble) inactive metabolites by conjugation of

subgroups to -OH, -SH, -NH2 functional groups

in the drug.

Drugs are metabolised via phase II reactions that

are excreted via the kidney.

Glucuronidation involves the covalent addition of glu-

curonic acid to a chemical is an elimination and detoxi-

ﬁcation mechanism for a host of lipophilic compounds

that includes drugs, environmental chemicals, and an-

tibiotics.

Patients who are deﬁcient in acetylation capacity,

known as slow acetylators, may experience prolonged

or toxic responses to normal doses of certain drugs due

to their decreased rates of metabolism. Acetylation

is a crucial metabolic pathway where an acetyl group

is transferred to a drug molecule, typically mediated

by N-acetyltransferase enzymes. In slow acetylators,

the activity of these enzymes is signiﬁcantly reduced,

leading to slower metabolism and clearance of drugs

that undergo acetylation.

As a result, the drugs and their active metabolites

may accumulate in the body to toxic levels, prolonging

their pharmacological eﬀects and increasing the risk of

adverse reactions. This can be particularly problematic

for medications with a narrow therapeutic index, where

the diﬀerence between therapeutic and toxic doses is

small. Consequently, slow acetylators require careful

1.8 Metabolism Pathway-Phase 3

dosage adjustments and monitoring to prevent potential

toxicity and ensure therapeutic eﬃcacy. Understand-

ing an individual’s acetylation capacity is crucial for

personalised medicine, allowing healthcare providers

to tailor drug therapies to the patient’s metabolic proﬁle

for optimal safety and eﬀectiveness.

1.8 Metabolism Pathway-Phase 3

Further modiﬁcation of the conjugated drug and

excretion.

A detoxiﬁcation process and transportation of

the conjugates against a concentration gradient

out of the cell into the interstitial space between

cells.

The conjugated drug enters the capillary system

and then the main bloodstream and is ﬁltered by

the kidneys.

Once conjugated, the drug undergoes a detoxiﬁcation

process, transporting the conjugates against a concen-

tration gradient out of the cell. This active transport is

mediated by eﬄux transporters, which move the drug

conjugates from the intracellular environment into the

interstitial space between cells. The conjugated drug

enters the capillary system and the main bloodstream.

As the conjugated drug circulates in the blood-

stream, it reaches the kidneys, where ﬁltration occurs.

The kidneys ﬁlter the blood, selectively reabsorbing

essential substances while allowing waste products,

including drug conjugates, to pass into the urine for

excretion. This sequence of processes ensures that the

drug and its metabolites are eﬀectively detoxiﬁed and

removed from the body, maintaining homeostasis and

preventing potential toxicity.

Genes involved in drug metabolism are the cy-

tochrome P450 (CYP) genes. They exhibit the highest

levels of polymorphism. There are approximately 50

CYP genes, consisting of 49 functional genes and one

pseudogene. These genes produce numerous isoforms,

enzyme variants derived from a single gene. CYP

isoforms are categorised into families and subfamilies

based on sequence homology; families share at least

40% sequence homology, while subfamily members

share at least 55% homology. Notably, about a dozen

(Image

courtesy:https://docs.google.com/document/d/11JZnhGcrKRNCYrZQuv5JeITcBkA B6JJi-

ls I7kozg)

Figure 3: The ﬁgure illustrates the major CYP 450

genes involved in drug metabolism.

(Image courtesy:

https://docs.google.com/document/d/11JZnhGcrKRNCYrZQuv5JeITcBkA B6JJi-ls I7kozg)

Figure 4: Table: 1 ADME (Adsorption, Distribution,

Metabolism, and Excretion) genes—core list.

enzymes from the 1, 2, and 3 CYP families are primar-

ily responsible for the metabolism of most drugs and

other foreign compounds.

A group of 32 ADME (Adsorption, Distribution,

Metabolism, and Excretion) genes for pharmacoge-

nomics study for drug development. ADME genes

are provided in Table 1.

Understanding pharmacogenomics is crucial in

pharmacodynamics for selecting the appropriate molec-

ular targets for a drug’s action. These targets can be lo-

cated on the cell surface, such as receptors ion channels,

or within the cell, like enzymes or regulatory proteins.

Figure 6 highlights two major determinants in drug de-

velopment: pharmacokinetics (PK) and pharmacody-

namics (PD). Both PK, which involves the absorption,

distribution, metabolism, and excretion of drugs, and

PD, which concerns the drug’s eﬀects on the body,

are signiﬁcantly inﬂuenced by genetic polymorphisms.

These genetic variations can aﬀect drug eﬃcacy and

safety, underscoring the importance of personalised

medicine in optimising drug therapy.

In clinical drug development, pharmacogenomics

1.8 Metabolism Pathway-Phase 3

(Image courtesy:

Image courtesy:

https://docs.google.com/document/d/11JZnhGcrKRNCYrZQuv5JeITcBkA B6JJi-ls I7kozg)

Figure 5: Two major determinants of drug develop-

ment: PK and PD.

continues to play a crucial role. Genetic screening of

participants can identify those most likely to beneﬁt

from the drug and those at risk of adverse reactions.

This stratiﬁcation improves clinical trial eﬃciency and

success rates by ensuring patients receive the right

treatments. Integrating pharmacogenomics through-

out drug development leads to more eﬀective, safer,

and personalised therapies.

References

Sykiotis GP, Kalliolians GD, Papavassiliou AG.

Pharmacogenetic principles in the Hippocratic writ-

ings. J Clin Pharmacol. 2005;45(11):1218–1220.

Adams J. Pharmacogenomics and personalised

medicine. Nature Education, 2008;1(1):194.

Higgins G. “The role of pharmacogenomics (PGx)

in drug discovery” in Drug Discovery: Practices, Pro-

cesses, and Perspectives, eds. Li JJ, Corey EJ.

Kaiser J. Senate panel approves $2 billion raise

for NIH in 2016. Science. June 23, 2015.

http://www.sciencemag.org/news/2015/06/senate-panel-approves-2-billion-raise-nih-2016.

Jancova P, Siller M. “Phase 2 drug metabolism”

in Pharmacology, Toxicology and Pharmaceutical Sci-

ence, Paxton J, ed.

https://www.intechopen.com/books/topics-on-drug-metabolism/

phase-ii-drug-metabolism.

https://www.mlo-online.com/continuing-education/

article/13009247/the-role-of-pharmacogenomics-in-precision-medicine

About the Author

Geetha Paul is one of the directors of

airis4D. She leads the Biosciences Division. Her

research interests extends from Cell & Molecular Bi-

ology to Environmental Sciences, Odonatology, and

Aquatic Biology.

CpG Sites: The Tiny Turn Signals of Our

DNA

by Jinsu Ann Mathew

airis4D, Vol.2, No.7, 2024

www.airis4d.com

Imagine driving down a complex highway sys-

tem. Traﬃc signals guide you, ensuring a smooth ﬂow

and preventing collisions. Within the intricate world

of DNA, CpG sites act like tiny turn signals, silently

inﬂuencing the ﬂow of genetic information. While they

appear simple – just a cytosine (C) followed by a gua-

nine (G) – their role in regulating gene expression is

anything but basic.

This section dives into the fundamentals of CpG

sites, laying the groundwork for understanding their

profound impact on cellular function and potential links

to disease. By understanding these tiny turn signals,

we gain a deeper appreciation for the intricate dance

between DNA and gene expression. This knowledge

paves the way for future exploration into how CpG

sites can inﬂuence our health and well-being, poten-

tially oﬀering insights into disease diagnosis and even

personalized medicine approaches.

2.1 What are CpG Sites?

Our DNA is a double helix formed by two strands

linked by pairs of chemical units called nucleotides.

These nucleotides come in four varieties, distinguished

by their nitrogenous bases: Adenine (A), Thymine (T),

Cytosine (C) and Guanine (G). CpG sites speciﬁcally

refer to locations on the DNA strand where a cyto-

sine (C) nucleotide is followed by a guanine (G) nu-

cleotide 1. These CpG sequences are relatively com-

mon throughout the genome, but their distribution and

function vary depending on their location.

(Image courtesy:Helixitta - Own work, CC BY-SA 4.0,

https://commons.wikimedia.org/w/index.php?curid=47075331)

Figure 1: CpG site and CG base pairing in DNA.

Here’s where things get interesting: CpG sites

are key players in epigenetics, the study of how fac-

tors beyond the DNA sequence itself can inﬂuence

gene expression. Epigenetic modiﬁcations act like tiny

switches on our genes, turning them on or oﬀ without

altering the underlying DNA code.

Let’s take a closer look at the individual compo-

nents of a CpG site:

Cytosin(C)

Cytosine is one of the four main nitrogenous bases

found in DNA. It has a unique property that makes it

a key player in gene regulation. Cytosine can undergo

a chemical modiﬁcation known as methylation. In this

process, a methyl group (CH3) is added to the cytosine

molecule. This modiﬁcation can change how genes

are expressed. For example, when a cytosine in a CpG

site is methylated, it can prevent certain proteins from

binding to the DNA, thereby silencing the associated

gene.

2.2 CpG Islands

(Image courtesy:https://www.nature.com/articles/npp201085)

Figure 2: CpG Island

Guanine (G)

Guanine is another of the four main nitrogenous

bases in DNA, and it pairs with cytosine in the DNA

double helix. In the context of CpG sites, guanine has a

speciﬁc but passive role. Although guanine does not di-

rectly undergo modiﬁcations like cytosine, its presence

next to cytosine is crucial. The combination of cytosine

followed by guanine (CpG) creates a speciﬁc sequence

that is recognized by enzymes involved in the methy-

lation process. These enzymes, called DNA methyl-

transferases, target CpG sites to add methyl groups to

the cytosine.

’p’ in CpG

The ”p” in CpG represents the phosphate group

that links the cytosine (C) and guanine (G) nucleotides

together. This phosphate group is a crucial part of the

DNA backbone, which provides structural integrity and

stability to the DNA molecule. The backbone, consist-

ing of alternating sugar and phosphate groups, main-

tains the double helix structure and ensures the DNA’s

robustness, allowing it to function properly within the

cell.

2.2 CpG Islands

Within the vast landscape of our DNA, CpG sites

are scattered like individual beads. However, these

sites aren’t distributed uniformly. There are speciﬁc re-

gions where CpG sites cluster together, forming dense

neighborhoods called CpG islands 2. These islands

stand out from the surrounding DNA landscape due to

their unique characteristics.

Unlike the rest of the genome where CpG sites

are relatively rare, CpG islands boast a high concen-

tration of guanine (G) and cytosine (C) nucleotides.

This creates a higher density of CpG dinucleotides

compared to surrounding regions. Interestingly, CpG

islands are often found near the beginning (promoter

region) of genes. This strategic placement positions

them to directly inﬂuence gene expression. Promoters

act like on/oﬀ switches for genes, and CpG methylation

within these islands plays a crucial role in regulating

that switch.

While methylation can occur at CpG sites through-

out the genome, CpG islands are typically unmethy-

lated. This allows the associated genes to remain ac-

tive and readily express their proteins when needed by

the cell. The unmethylated state of CpG islands in

promoters allows for eﬃcient gene expression. When

methylation does occur within these islands, it can act

as a more precise control mechanism, ﬁne-tuning the

level of gene activity. For example, slight changes

in methylation within a CpG island might result in a

gene being expressed at a lower level compared to its

completely unmethylated state.

2.3 Why CpG Sites Matter

CpG sites may seem like minor components of

our vast genome, but their importance cannot be over-

stated. These small sequences play crucial roles in the

regulation of gene expression, impacting various bi-

ological processes and overall health. Understanding

why CpG sites matter gives us insight into the intricate

mechanisms that control our genetic functions. Here

are key reasons why CpG sites are vital:

2.3.1 Gene Regulation

CpG sites play a pivotal role in the regulation

of gene expression. Through a process called DNA

methylation, a methyl group (CH3) can be added to the

cytosine at CpG sites. This modiﬁcation can either re-

press or activate gene expression 3. When CpG sites in

gene promoter regions are methylated, the associated

gene is often silenced. Conversely, demethylation can

activate gene expression. This precise control mecha-

nism is essential for normal development, cellular dif-

ferentiation, and response to environmental changes.

2.4 Conclusion

(Image courtesy:https://evolutionletters.wordpress.com/evolution-learning-zone/evolution-

explained/dna-and-gene-expression-how-they-make-us-who-we-are/)

Figure 3: Regulation of gene expression.

2.3.2 Epigenetic Memory

CpG methylation patterns can be inherited through

cell divisions, providing what is known as epigenetic

memory. This means that when a cell divides, the

methylation patterns on its CpG sites are copied to the

new cells. These patterns help cells ”remember” their

identity and continue performing their speciﬁc func-

tions.

For example, once a cell becomes a muscle cell,

it keeps its muscle cell identity because of the stable

methylation patterns at its CpG sites. These patterns

control which genes are turned on or oﬀ, ensuring the

cell behaves like a muscle cell. Similarly, liver cells

remain liver cells through generations of cell division

due to these stable methylation patterns that regulate

liver-speciﬁc genes.

This epigenetic memory is crucial for the body’s

proper functioning, as it maintains the specialization

and consistency of diﬀerent cell types. It ensures that

each cell type continues to perform its unique role,

contributing to the overall health and stability of an

organism.

2.3.3 Health and Disease

Abnormal methylation of CpG sites is associated

with various diseases, including cancer. Hypermethy-

lation of CpG islands in tumor suppressor gene pro-

moters can lead to gene silencing, removing critical

controls on cell growth and contributing to tumor de-

velopment. On the other hand, hypomethylation can

activate oncogenes, which can also promote cancer.

Understanding these methylation changes is crucial for

developing diagnostic tools and treatments for cancer

and other diseases.

2.3.4 Response to Environment

CpG sites allow our cells to respond to environ-

mental changes in a ﬂexible and dynamic way. Fac-

tors like diet, lifestyle choices, and even exposure to

stress can inﬂuence the methylation state of CpG sites.

This adaptability is crucial for organisms to cope with

a changing environment and maintain optimal health.

By adjusting gene expression through CpG methyla-

tion, cells can adapt to various stressors and ensure

survival.

2.4 Conclusion

In conclusion, CpG sites, once viewed as mere

sequences within our DNA, have emerged as powerful

regulators of gene expression. Like tiny turn signals,

they guide the ﬂow of genetic information, inﬂuenc-

ing cellular function, development, and even our sus-

ceptibility to disease. Understanding the methylation

patterns at CpG sites opens exciting possibilities for

the future of medicine. By analyzing these patterns,

researchers may be able to develop non-invasive diag-

nostic tools for early disease detection. Additionally,

manipulating CpG methylation holds promise for novel

therapeutic approaches, potentially oﬀering ways to re-

store normal gene expression and combat disease. As

our knowledge of CpG sites continues to evolve, they

have the potential to revolutionize healthcare, paving

the way for personalized medicine and a deeper under-

standing of human health.

References

CpG site

CpG Islands

CpG and Non-CpG Methylation in Epigenetic

Gene Regulation and Brain Function

CpG Islands Track Settings

Collaborations between CpG sites in DNA methy-

lation

2.4 Conclusion

Comprehensive analysis of CpG islands in hu-

man chromosomes

About the Author

Jinsu Ann Mathew is a research scholar

in Natural Language Processing and Chemical Infor-

matics. Her interests include applying basic scientiﬁc

research on computational linguistics, practical appli-

cations of human language technology, and interdis-

ciplinary work in computational physics.

Part IV

General

Scaling Mechanisms in Software Systems

by Arun Aniyan

airis4D, Vol.2, No.7, 2024

www.airis4d.com

1.1 Introduction

Ever wondered how websites and applications like

Google and Facebook are able to serve data and provide

service to millions of users worldwide at the same time?

People around the world may access the same media

or webpages at the same time and we hardly feel the

lag and delay unless caused by network issues. This

is achieved by a method called “scaling” in software

systems. Scaling makes sure that service provided

remains stable and the users feel no delay with the

applications they use.

As software systems grow in complexity and user

base, the need for eﬀective scaling mechanisms be-

comes crucial. Scaling refers to a system’s ability to

handle increased load without compromising perfor-

mance or reliability. This article explores various scal-

ing strategies and mechanisms used in modern software

architecture.

1.2 Vertical Scaling

Vertical scaling is the process of increasing the ca-

pacity of a single node or server in a system by adding

more resources to it. This approach is often the ﬁrst

consideration when a system needs to handle an in-

creased load. This is a very intuitive approach since

to do more work a system needs more muscle power.

So to do the heavy lifting simply adding more com-

pute resources eases the heavy lifting. There are both

hardware and software methods for vertical scaling.

1.2.1 Hardware Upgrade

The ﬁrst and foremost upgrade is adding a more

powerful processor or increasing the number of cores.

Powerful processors are not just about the clock speed

they use but also the design of the compute engine. For

example for computations that include large ﬂoating

point calculations processors having dedicated engines

like Altivec are used. If the application involves more

threading and multiprocessing, upgrading to a proces-

sor that supports more threads with a larger number of

cores are often preferred.

In terms of memory, bumping up the RAM based

on data usage is done. When there is large amount of

read / write operation with huge volume, using faster

Solid State Disks (SSDs) over conventional spinning

hard drives are preferred since they considerably help

with reducing data latency over the device. Data pro-

cesses like Dynamic Memory Access (DMA) which

does not involve CPU cycles will be a lot faster.

Network traﬃc speeds are also considered in hard-

ware scaling and hence cards with faster data rates and

reduced redundancy are used. This makes sure the

latency and throughput are optimal when services are

accessed through a network.

1.2.2 Software Optimisations

The ﬁrst level of software optimization involves

upgrading the Operating System to use a fast and eﬃ-

cient one. There are OS versions and even ﬂavors that

are speciﬁcally ﬁne-tuned and optimized for large-scale

enterprise applications. They basically make sure of

eﬃcient resource management.

1.3 Horizontal Scaling

Many applications make use of diﬀerent databases

like SQL or even NO-SQL. There are diﬀerent varieties

of them which are designed for a speciﬁc set of use

cases. For example, if an application requires to store

large data but be very eﬃcient in terms of read speeds,

then Postgres SQL is a good choice. But if the applica-

tion intends to store humungous amounts of data with

less frequent reads, then solutions like CockroachDB

is an ideal choice.

The next level of software optimizations involves

ﬁne-tuning the application code to better utilize the

available resources. For example, if the application

queries the database very often, use caching systems

to cache the data in memory. Redis caching is a good

example. Similarly, if there are processes that can work

independently, then run them in threads or Async fash-

ion. The code should be reviewed thoroughly to make

sure the applications free unused resources at diﬀerent

stages and do not unnecessarily use up memory. Code

changes in these aspects can improve the application

performance to a large extent.

The following are the main considerations when

deciding to do vertical scaling.

1. Performance Bottlenecks: Identify the primary

bottleneck (CPU, memory, I/O) before scaling to

ensure you’re addressing the right issue.

2. Application Architecture: Some applications may

not eﬀectively utilize additional resources with-

out code changes.

3. Cost-Beneﬁt Analysis: Compare the cost of up-

grading hardware versus other scaling strategies.

4. Future Growth: Consider whether vertical scal-

ing will meet long-term growth projections or if

a diﬀerent approach might be needed eventually.

5. Backup and Recovery: As the system grows

vertically, ensure that backup and recovery pro-

cesses can handle the increased data volume and

processing requirements.

The main advantages of vertical scaling are simplicity,

consistency, and lower latency. Simplicity is in the

sense that often requires minimal changes to applica-

tion architecture or code. Consistency refers to a sim-

pler system architecture with fewer nodes to manage.

Single-node operations typically have lower latency

than distributed systems.

Even though vertical scaling is simple and eﬃ-

cient it is limited by the following disadvantages.

1. Hardware constraints: There’s a physical limit

to how much a single machine can be upgraded.

2. Downtime: Upgrades often require system down-

time, which can be problematic for high-availability

services. This can be minimized by opting for

cloud-based solutions like AWS and GCP which

have very low downtime when switching re-

sources.

3. Cost: High-end hardware can be exponentially

more expensive than commodity hardware.

4. Single Point of Failure: Reliance on a single ma-

chine increases the risk of system-wide outages.

When these limits are reached, organizations of-

ten need to consider horizontal scaling or other ar-

chitectural changes to continue growing their system

capacity.

1.3 Horizontal Scaling

Horizontal scaling is the process of adding more

machines or nodes to a system to distribute the load

and increase overall capacity. This approach is fun-

damental to building large-scale, distributed systems

capable of handling massive workloads. This is a style

of scaling often preferred by large enterprises which

has a huge customer base.

The key aspect of horizontal scaling is distributed

systems. This involves distributing the load across

multiple machines as compared to a single node with

vertical scaling. To achieve horizontal scaling initially

more servers are added to the network. Then the tasks

and data are distributed across multiple nodes. And this

requires architectural changes to support distributed

processing.

1.3.1 Distributed Architecture

The primary step in distributed architecture is

breaking down applications into smaller, distributable

components. This is done when designing the appli-

cation architecture itself. There are often dependen-

1.3 Horizontal Scaling

cies among components and many components do not

need to wait for the other. Such components can be

distributed among nodes and can communicate when

required. This is a fundamental design principle with

high-performance computing systems.

To allow the distribution of workloads, compo-

nents are made statelessness. This means that the com-

ponents are never in waiting states and are made sure

they are churning data or processes. This makes sure

there are minimal dependencies between components.

The other consideration is implementing service

discovery mechanisms for dynamic environments. This

is often when the system has added features with up-

grades and new releases. The components need to

know and also discover speciﬁc parts of the system.

Querying and routing mechanisms are usually put in

such places.

1.3.2 Data Distribution

This is speciﬁcally done to the database system

that is used. For large databases partitioning or shard-

ing data across multiple nodes is done. This allows for

eﬃcient read and write of large datasets and fast ac-

cess. For the distributed components of the application,

some of them may need to access certain tables more

often than others. Partition allows this to be made pos-

sible eﬃciently. Consistent hashing for eﬃcient data

distribution is also often used.

Another aspect of data distribution is managing

data replication for redundancy and improved read per-

formance. One can create replicas of tables or parts of

the databases to improve data access.

1.3.3 Load Balancing

Load balancing is a critical component of hori-

zontal scaling, distributing incoming requests across

multiple servers. The diﬀerent independent compo-

nents will make the distributed system heterogeneous

in nature with respect to data and resource access.

This situation should be eﬀectively managed with load-

balancing techniques. Load balancing mainly involves

distributing incoming requests across multiple nodes

using various load-balancing algorithms and software

and hardware load balancers.

Load balancing algorithms include methods like

round-robin techniques which distribute load sequen-

tially across diﬀerent servers. Another strategy is

called least connections which sends requests to the

server with the fewest active connections.

Software and hardware load balancing mainly deals

with network traﬃc. There are dedicated hardware load

balancers which include a high-speed network switch

with specialized load-balancing ﬁrmware. Software

load balancing is done with proxy software like Nginx

and HAProxy. DNS load balancing is also a strategy

that is often employed.

1.3.4 Consistent and Coordination

Distributed systems require strict coordination and

consistent transfer patterns to keep stable operations.

For example, many processes may be written to the

database at the same time. Some instances require

locking the instance and handling the other connec-

tions at the same time and releasing the lock state after

completion. These kinds of instances need to be eﬀec-

tively handled.

The following are the main considerations taken

into account while doing horizontal scaling.

CAP Theorem: Understand the trade-oﬀs be-

tween Consistency, Availability, and Partition

Tolerance in distributed systems.

Network Latency: Design systems to minimize

unnecessary network communication.

Data Locality: Consider data access patterns

when designing data distribution strategies.

Operational Complexity: Invest in robust moni-

toring, logging, and tracing systems.

Cost Management: Implement proper resource

allocation and scaling policies to optimize costs.

Security: Ensure proper network segmentation

and access controls in distributed environments.

Horizontal scaling is advantageous in terms of scal-

ing by adding nodes, provides better fault tolerance by

adding resilience and redundancy. It provides more

ﬂexibility to scale only speciﬁc components and al-

lows global distribution of services across geographic

1.4 Conclusion

locations.

But one can evidently see that even though it is

advantageous to have horizontal scaling, it is extremely

complex to design and deploy. Data consistency has to

be maintained all across. Debugging speciﬁc issues has

to be done within the distributed environment which is

very challenging.

1.4 Conclusion

Eﬀective scaling is crucial for the success of mod-

ern software systems. The choice of scaling mecha-

nisms depends on various factors, including the nature

of the application, expected growth, budget constraints,

and technical expertise. Often, a combination of these

strategies is employed to create a robust, scalable ar-

chitecture. As systems evolve, it’s important to contin-

uously monitor performance, reassess scaling strate-

gies, and adapt to changing requirements. With the

right approach, software systems can gracefully han-

dle growth and provide consistent performance even

under increasing loads.

Reference

What is Scalability in Software Development

How to Scale a Software Product?

Why Software Scalability is Important for App

Development

What is software scalability, and why should

your company take it seriously?

About the Author

Dr.Arun Aniyan is leading the R&D for Arti-

ﬁcial intelligence at DeepAlert Ltd,UK. He comes from

an academic background and has experience in design-

ing machine learning products for diﬀerent domains.

His major interest is knowledge representation and com-

puter vision.

About airis4D

Artiﬁcial Intelligence Research and Intelligent Systems (airis4D) is an AI and Bio-sciences Research Centre.

The Centre aims to create new knowledge in the ﬁeld of Space Science, Astronomy, Robotics, Agri Science,

Industry, and Biodiversity to bring Progress and Plenitude to the People and the Planet.

Vision

Humanity is in the 4th Industrial Revolution era, which operates on a cyber-physical production system. Cutting-

edge research and development in science and technology to create new knowledge and skills become the key to

the new world economy. Most of the resources for this goal can be harnessed by integrating biological systems

with intelligent computing systems oﬀered by AI. The future survival of humans, animals, and the ecosystem

depends on how eﬃciently the realities and resources are responsibly used for abundance and wellness. Artiﬁcial

intelligence Research and Intelligent Systems pursue this vision and look for the best actions that ensure an

abundant environment and ecosystem for the planet and the people.

Mission Statement

The 4D in airis4D represents the mission to Dream, Design, Develop, and Deploy Knowledge with the ﬁre of

commitment and dedication towards humanity and the ecosystem.

Dream

To promote the unlimited human potential to dream the impossible.

Design

To nurture the human capacity to articulate a dream and logically realise it.

Develop

To assist the talents to materialise a design into a product, a service, a knowledge that beneﬁts the community

and the planet.

Deploy

To realise and educate humanity that a knowledge that is not deployed makes no diﬀerence by its absence.

Campus

Situated in a lush green village campus in Thelliyoor, Kerala, India, airis4D was established under the auspicious

of SEED Foundation (Susthiratha, Environment, Education Development Foundation) a not-for-proﬁt company

for promoting Education, Research. Engineering, Biology, Development, etc.

The whole campus is powered by Solar power and has a rain harvesting facility to provide suﬃcient water supply

for up to three months of drought. The computing facility in the campus is accessible from anywhere through a

dedicated optical ﬁbre internet connectivity 24×7.

There is a freshwater stream that originates from the nearby hills and ﬂows through the middle of the campus.

The campus is a noted habitat for the biodiversity of tropical Fauna and Flora. airis4D carry out periodic and

systematic water quality and species diversity surveys in the region to ensure its richness. It is our pride that

the site has consistently been environment-friendly and rich in biodiversity. airis4D is also growing fruit plants

that can feed birds and provide water bodies to survive the drought.