Computational Engineering

Cancer Genomics 

Genomics is the study of the genes in the genome and their interactions with the environment. Genomic analysis concerns withbiomolecules such as DNA sequences and RNA sequences in (i) identifying them,(ii) comparing their features, and (iii) measuring their structural variation, gene expression, or regulatory and functional element annotation. Next Generation Sequencing (NGS) technologies help in the characterization of point mutations of a wide range of cancers and their structural alterations. Complete genome sequences of numerous cancer types are becoming increasingly available, which provide a comprehensive view of cancer development and growth.Cancer genome studiesenable understanding the gene abnormalities that are the root cause of many types of cancer. This improved understanding of cancer biology helps in developing new waysof diagnosis and treatment.

Cancer accounts for approximately 13% of all deaths worldwide and contributes significantly to global socio-economic burden to the society. There exists evidenceto support the theory that cancer is a disease of genome; aberrant genomic alterations are the hallmark of cancer cells. At the computational engineering lab, current research in cancer genomics is focused mainly on understanding intratumor heterogeneity whichplays a major role in tumor metastasis and drug resistance to therapeutics. To better understand the tumor genome heterogeneity in tumor niche, a global approach of sequencing the whole cancer genome was undertaken. Since then, several sequencing studies have identified a strong correlation between genomic alterations with tumor progression and resistance to conventional therapies. However, effective treatment of cancer rely on early detection, precise diagnosis, and implementation of specific therapeutic strategies that give rise to the concept of personalized medicine.

Although, whole genomic sequences provide important insights into clonal evolution of mutations, lack of proper tools and pipelines to analyze and interpret the framework of data has limited the use of sequencing technology for diagnosis and therapeutics. Independent studies have identified several driver mutations with specific cancer types, but analyzing how each of these mutations work individually or in a cohort towards tumor progression is limited. Recent advancement in computational biology and mathematical modeling has shown great potential in analyzing huge sequencing data and in understanding the origin and progression of cancer without much of the wet lab experiments and ethical issues.

Figure 1: Genome Analysis Pipeline

Genome Analysis Pipeline

A computational model for identifying multiple cancer specific biomarkers from the whole genome sequence data of cancer patient and analyzing cancer prognosis using predictive machine learning models is proposed. The source of the whole genome sequencing data of cancer cells will be obtained from the European Genome-Phenome Archive (EGA), The Cancer Genomics Hub (CGHub) and other relevant databases. The source for the potential biomarkers known for each cancer type is will be obtained from 1000 genomes browser and COSMIC databases and the reference sequences of these biomarkers will be obtained from NationalCenter for Biotechnology Information (NCBI). The pipeline developed will be a cluster of four main groups based on their application such as assembly, alignment, variation discovery and annotation. For predictive modeling, machine learning approaches will be used, that have unique characteristics for prediction and classification such ascapacity control of the decision function, the use of the kernel functions, and the scarcity of the solution. The three-way data split using this model will be applied for training, validation, and testing. The proposed pipeline and a schematic representation of the frontend of the pipeline are shown below.

Figure 2(a): Frontend of the MBICGD Pipeline

Figure 2(b): Frontend of the MBICGD Pipeline (contd.)

Research Objectives

• To recognize and use new opportunities offered by NGS technologies in the field of cancer genomics
• To overcome challenges raised by the massive amount of sequence data that has been and will be generated
• To create a dedicated bioinformatics facility for genome analysis related to cancer studies
• To empower biomedical teams working in the field of cancer to best exploit the huge data generated by the ongoing genomic projects

 Research Outcome

• High-throughput tools for DNA and RNA characterization that facilitate comprehensive analyses of cancer genomes including all somatic alterations.
• Cancer genome methodology to understand the relationship between the genetic mutations and the clinical response to assist in cancer therapy.
• New methodologies for cancer diagnosis and prognosis.
• Comprehensive catalogs of the genomic changes in respect of specific cancer types.
• Technologies for complete molecular profiling of tumors enabling genome-informed personalized cancer medicine.
• New techniques for validation of genomic data to (i) distinguish mutations responsible for disease pathogenesis resulting from genomic instability, (ii) define genes responsible for cancer initiation, progression, and maintenance, and (iii) identify the effective ways to therapeutically exploit this information.

Beneficiaries

• Cancer patients
• Medical practitioners
• Genome scientists
• Research scholars

Wrist Pulse Analyzer

Wrist pulse analysis is a simple non-invasive technique used in ancient Indian medicine Ayurveda and traditional Chinese medicine (TCM)to diagnose health. Most modern pathological diagnosis tools give reports in terms of content and chemical changes in the patient’s body. Very few devices give reports based on body constitution of the patient. Wrist pulse diagnosis is an ancient method of diagnosing human body constitution. Computerized pulse diagnosis uses sensors to acquire wrist pulse signals and uses machine learning techniques to analyze patient’s health based on the acquired pulse signals.Wrist pulse indicates the significantly varied blood flow of an organ in abnormal health conditions. Patient health conditions manifest themselves in reflected waves of the wrist pulse signals in different ways. With careful interpretation of these signals, one can perform an initial diagnosis to great accuracy.

 Research Objectives

• To design an efficient and accurate pulse acquiring device without losing vital radial pulse information of the patient
• To acquire radial arterial acoustic vibrations through a non-invasive pulse detector
• To develop an indigenous application to analyze, characterize and classify the human radial pulse obtained through the detector
• To provide low-cost pulse diagnosis system for alternative and integrated medical practitioners

To provide complete analysis of the overall constitution of a person in accordance with the TCM and Ayurveda principles through charts and bar graphs

Figure 3: Wrist Pulse Analysis

Research Outcome

• A wrist pulse detector which takes acoustic vibrations of the radial arterial pulse and an intelligent machine which analyses these variations and provides vital five elements energy level information which aids in the correct diagnosis of ailments.
• Low-cost pulse diagnosis system for alternative and integrated medical practitioners.
• A system for complete analysis of the overall constitution of a person in accordance with the TCM and Ayurveda principles through charts and bar graphs.

Beneficiaries

• Alternative medicine practitioners (AYUSH) for getting complete body constitution of the patient.
• Patients in terms of getting better and affordable alternative therapies.
• Researchers to pursue their research on pulse diagnosis and alternate therapies.

Whole-Cell Modeling and Simulation

Understanding and engineering of biological systems require comprehensive models of cellular physiology with 100% predictability. These whole-cell models guide experiments in molecular biologyand enable simulation and computer-aided design in synthetic biology. Theyhelp personalized medical treatment. Constructing comprehensive whole cell models with sufficient detail and validating them is a massive work. Modeling larger cells and more complex physiology involve (i) model building and integration, (ii) model validation, (iii) data curation, (iv) experimental interrogation, (v) analysis and visualization, (vi) accelerated computation, (vii) community development and (viii) collaboration. A broader multidisciplinary research community is needed to innovate in all these areas. As a first step, a study on the issues and challenges is initiated at the lab and is in progress.

Figure 4: Whole-Cell Modeling Process

Research Objectives

• To set up and configure tools for modelling.
• To estimate parameters and feed the estimated data to a tool.
• To design submodules depending on the cell process.
• To design a model by combining all the submodules.
• To speed up the whole-cell model simulations on high-performance heterogeneous platform, exploiting parallel computing technology.

Research Outcome

• The precise model by making use of available whole-cell model software and tools.
• To represent cell functions and characteristic using whole-cell modelling.
• To validate the new model with experimental data.

Beneficiaries

• Clinicians
• Bioinformatics scientists
• Research scholars

Cyber Security

Digital technology is growing rapidly, causing a wide spread of distribution of digital documents and images over the Internet. The security of digital documents, images and other multimedia data has thus become extremely important for common people and the government. Secured transmission of digital images is an important issue in information security field. Cryptographic encryption techniques provide an effective security to data by converting it into an un-understandable form to attackers. Conventional cryptographic encryption methods are unsuitable for image encryption. In recent years, chaotic theory has attracted research community for image encryption. A simple and secured scheme for image encryption using one-dimensional logistic maps has been studied. This image encryption scheme first shuffles the position of pixel values and then changes the gray values leading to a complex relationship between the original plain image and the encrypted image. Two operations, namely, image scrambling and diffusing, are performed by logistic maps. Various experiments test the robustness and the security aspects of the algorithm and it is found that such schemes are resistant to different cryptanalytic attacks and provides adequate security.

a) Original Image                            b) Encrypted Image

Figure 5: Image Encryption based on Chaotic Theory

Tongue Diagnosis using Deep Learning

Analysis of the tongue is crucial in the evaluation of the human body in Ayurveda medicine. There is substantial research scope for tongue diagnosis in Ayurvedic therapy. Tongue pictures are classified into vata, pitta and kapha (tri-dosha) with the color, coating, texture and shape features. These characteristics would also be determined largely by the portion of the tongue part (upper region, middle region and sides). Segmentation is the most important procedure in defining the region of interest in the tongue image. These features include various color models and differentiated parts of the tongue body. Coating and background are also considered for feature set to feed them to the training-testing model for classification. The final outcome of the research is to blend in with the traditional Ayurvedic methods to the machine learning and deep learning model as it describes the basic class of inspection in medication.

Figure 6: Tongue Diagnosis using Deep Learning