International Journal of Recent Technology and Engineering (IJRTE)
ISSN: 2277-3878, Volume-7 Issue-5S2, January 2019
302
Published By:
Blue Eyes Intelligence Engineering
& Sciences Publication
Retrieval Number: ES2053017519/19©BEIESP
Abstract: Tumor suppressor genes are always linked with
stress, directly or indirectly which results in mutation. Therefore
the probability of turning these mutations into cancer increases.
Identification of major tumor suppressor genes and its presence
among Indian population is analyzed. Due to great advancement
in the field of deep learning, and wide variety of scopes in future,
deep learning is incorporated in this project to perform the
classification task .The requirement of large amount of data to
perform classification task is one of the major drawback of deep
learning. In order to solve this problem, one-shot learning
algorithm is introduced which gave the accuracy of 70.2%. A
secure data sharing platform has been developed using
blockchain technique.
Index Terms: Block Chain Technique, Deep Learning,
Tumor suppressor genes.
I. INTRODUCTION
Stress is a typical natural response to dangerous situations
of any kind of demand or threat, which is accompanied with
a number of chemical and biological aberrations and resulted
by the release of chemicals and hormones such as adrenaline
and cortisol. In a biomolecular perspective, it is found to be
extended even up to the cellular level. Cellular stresses is
generally resulted out of the cells being exposed to external
and internal stressors which damages the integrity of the cell
and its genome such as cigarette smoke, hypoxia, ionizing
radiations, oxidative stress, carcinogens, oncogene activation
etc. [1]. Apart from mental and physical stresses, another
main stress is the cellular stress. Cellular stresses can happen
when human cell is exposed to external and internal stressors
which damages the integrity of the cell and its genome such
as cigarette smoke, hypoxia, ionizing radiations, oxidative
stress, carcinogens, oncogene activation etc. [1]. This can
lead to damage of DNA and also malignant transformation of
cells. To ensure the survival of the organisms, cells have built
up various methodologies to adjust to stressors i.e. the
“Tumor Suppressor Genes” [2]. Due to this evolving nature
of biological systems, it becomes difficult to carry out the big
data analysis using the basic techniques.
Revised Manuscript Received on January 25, 2019
Kaajal Nishandh, Department of Electronics and Communication
Engineerining, Amrita School Of Engineerining, Amrita Vidhyapeetham,
Coimbatore, India.
Sanjay Kumar P, Amrita Molecular modeling and synthesis (AMMAS)
Research Lab, Computational Engineerining and Networking, Amrita School
Of Engineerining, Amrita Vidhyapeetham, Coimbatore, India.
P.K Krishnan Namboori, Molecular modeling and synthesis (AMMAS)
Research Lab, Computational Engineerining and Networking, Amrita School
Of Engineerining, Amrita Vidhyapeetham, Coimbatore, India.
This gap could be bridged using emerging machine learning
techniques such as deep learning which works with huge data
sets. With the implementation of Siamese net or One- shot
learning, one could carry out the classification problem with
minimal data with better prediction accuracy [3].
Gene expression analysis is important to understand the
association of tumor suppressor genes in the regulation of the
disease. The ‘gene expression profiling’ can be considered as
the ‘mutation signature’ of the disease. Although gene
expression profiling is presently used as a primary research
tool, many other potential clinical applications of this
method are being instigated. The expression profile renders
prognostic data about the aggressiveness nature of tumor,
response to therapies, sensitivity and resistance to different
chemotherapies [4]. The expression level varies from person
to person, demanding a pharmacogenomic investigation in
the subject [4] [5]. The knowledge of pharmacogenomic with
genomic, epigenomic, metagenomic and environmental
genomic components, helps in developing safe and effective
medications according to the person’s genetic composition.
The proneness of the disease as well as the susceptibility of
drugs depends up on the individual variations. Epigenomics
deals with the study of the entire set of epigenetic alterations
on the genetic material of a cell (epigenome), Metagenomics
is the study of genetic material taken from micro-organisms
such as bacteria, yeast etc and environmental genomics deals
with the prediction of how the organisms respond to external
environmental factors [6]. Pharmacogenomics promises
personalized treatment for patients suffering from common
diseases, especially with multiple treatment modalities [7].
Pharmacogenomics provides the population specific marker
frequency profiles and also drug efficiency. Various machine
learning applications in healthcare helps in early diagnosis
and prediction of suitable treatment strategy to the patient.
Recently, deep learning algorithm has been identified as
most suitable for identifying cancerous tumors from
mammogram. At present, training the machine with
minimum number of samples could be possible with modern
learning experiences such as ‘one shot learning or Siamese
net’. One-shot learning ranks the similarity of inputs in
which the classification is done using K-nearest neighbor
calculating the Euclidean distance between the test and train
data, classifying the nearest ones as in the same class. Since
health care deals with confidential patient
Expression Profiling & Classification using
Convolutional Neural Networks of Tumor
Suppressor Genes Linked with Stress
Kaajal Nishandh, Sanjay Kumar P, P.K Krishnan Namboori