Highlights:

Event:

Projects:

Designing High-Performing, Graph Neural Network Based Novel Computational Tools for Multi-layered Protein Interaction Networks (PINs) Alignment and Identification of Network Modules/Motifs for Prioritizing Common Anti-Viral Drug Targets for Multiple Hepatotropic Viral Diseases

Funding Agency:  Dept of Biotechnology (Ministry of Science and Technology-GoI) 

Duration: 3 Years                  Starting Date: May 2025                  Cost: Rs. 16.72 Lakhs                  Status: Ongoing

Principal Investigator: Swarup Roy

Description:

Construction and analysis of PINs have been shown to deliver many fundamental aspects of protein functions at the systems level. Especially, analysis of host-pathogen PINs has been helping in understanding the molecular basis of how a pathogenmight subvert a robust host system. For example, recent studies have shown that viral proteins (SARS-CoV-2) primarily interact with the human proteins occupying central positions in the network. It is desirable to compare two or more PINs of related species to identify evolutionally conserved sub-networks. Aligning PPI networks (PIN) of evolutionally related species enable to discover conserved PINs and sub-network modules/motifs. Very few methods are available for the simultaneous alignment of more than two PINs. Available methods are computationally expensive and do not guarantee optimal alignments in a singleprocessor system. We propose to leverage Graph Neural Network (GNN) based framework to align multiple PINs. We propose to use the newly developed tool on many host-hepatotropic PINs to detect common sub-PINs representative of core host functions commonly targeted by pathogens. In the case of host-hepatotropic PINs, such conserved core components may relate to basic liver functions commonly targeted by multiple viruses. Hence, their discovery and analyses may enable us to design more effective therapeutic multi-viral drug targets. An effort will be made to prioritize suitable hepatotropic anti-viral drug targets.

Prioritizing Alzheimer’s Disease (AD) Responsible Genes Through Multi-Modal Network Integration and Advanced AI: Leveraging Graph Deep Learning and Large Language Models (LLMs)

Funding Agency:  IDEAS - TIH at Indian Statistical Institute,  Kolkata (Under DST-GoI) 

Duration: 1 Year                   Starting Date: March 2025                  Cost: Rs. 13.63 Lakhs                  Status: Ongoing

Principal Investigator: Swarup Roy

Description:

Alzheimer’s Disease (AD) is a complex neurodegenerative disorder that affects millions worldwide, yet its underlying mechanisms remain poorly understood. A growing body of evidence suggests that genes associated with AD may play crucial roles in biological networks, often acting as hub or connector genes. However, these assumptions require further investigation to uncover the regulatory mechanisms underlying AD. This project aims to develop a comprehensive gene prioritization framework by integrating multi-omics data using Graph Deep Learning and Large Language Models (LLMs).

Our approach focuses on constructing a high-confidence, integrated biological network by combining data from multiple omics sources, including microarray, single-cell RNA sequencing, and single-nuclei RNA sequencing. The network is further enriched using protein-protein interaction (PPI) and gene ontology (GO) data to capture the complexity of AD-related molecular interactions. By employing graph deep learning models, we aim to identify key genes involved in AD pathogenesis, while leveraging LLMs to enhance interpretability and contextualize findings with existing scientific knowledge.

The project is expected to deliver a prioritized set of genes highly relevant to Alzheimer’s Disease (AD), identified through the integration of multi-omics data and advanced network analysis. This ranked gene set will provide valuable insights into the regulatory roles of AD-associated genes, including both central hub genes and often-overlooked peripheral genes, thereby challenging conventional paradigms in disease research. By leveraging Large Language Models (LLMs), the findings will be contextualized with detailed, interpretative summaries for each gene, offering actionable insights for therapeutic exploration. Additionally, the project will establish a reproducible and scalable pipeline for gene prioritization, which can be readily adapted to investigate other neurodegenerative and systemic disorders, making a significant contribution to precision medicine and translational research.

Large Scale Biological Network Reconstruction and Analysis for   Identifying Therapeutic Targets for Alzheimer’s Disease: A Big Data Initiative

Funding Agency:  Department of Science & Technology (DST), Govt. of India (Under DST-ICPS Data Science Cluster Research Program) 

Duration: 3 Years                   Starting Date: April 2019                  Cost: Rs. 54 Lacks                  Status: Completed

Principal Investigator: Swarup Roy

Description:

The volume of data is growing fast in bioinformatics research. The importance of ‘big data’ in biology is enormous as huge quantities of high-throughput experimental data are available in public domain and growing almost every day. DNA microarray provides a platform to analyze in-silico the expression levels of thousands of genes over different conditions, such as different environmental conditions or separate developmental stages of diseases, in turn helps in effective drug design. Gene regulatory networks (GRN) alterations underlie many anomalous conditions, such as cancer. Inferring GRN and their alterations from high-throughput microarray data is a fundamental but challenging task. Thus GRN inference offers a platform for investigations into data intensive sciences. GRNs can be merged with protein-protein interaction (PPI) networks to provide a more complete picture of cellular processes. It has been observed that central or hub genes are the influential genes and the cause of most of the diseases. Investigating genetic causes of neurodegenerative diseases like Alzheimer or Parkinson are challenging. This is because, instead of any single gene, a group of genes interact with each other causing perturbation in the molecular pathways which lead to such complex diseases. Identifying key genes responsible for the diseases may help in designing patient-specific drugs. Reconstruction of a disease regulatory network from expression data allows researchers to understand the molecular basis of complex traits and diseases, as well as the discovery of direct drug targets. By understanding the dynamics of these networks, one can shed light on the mechanisms of diseases and helps pharmacologically to identify potential drug targets.

To date, data obtained from numerous expression microarray experiments have been made publicly available. Several existing inference algorithms implemented by the original authors are also available online in R platform in open access forum such as Bioconductor. However they are not convenient for use by the non-programmers. In this proposal we intend to investigate a set of related issues in creating, comparing and analyzing large gene GRNs and PPI networks, with tens of thousands of nodes, up to genome-wide dimensions. We intend to develop an integrated web tool with a graphical user interface (GUI) and instruction manual for the tasks of inference, visualization of large scale GRNs. An integrated platform for performance evaluation of various inference methods along with inference and visualization is missing in state-of-the-art inference tools. We make the developed tool available to biomedical researchers with all the above facilities including frequent feature updates where user can choose the suitable algorithms for inference of large scale network.

This proposal seeks to extend our prior work by building large-scale gene regulatory networks using publicly available expression datasets generated from thousands of experiments in combination while addressing issues of scaling and stability. This will necessitate developing new and improved efficient algorithms for analyzing large-scale graphs. Large cellular networks will be built with heterogeneous nodes and edges and analysed using the graph algorithms we develop. Our final objective is to develop a user-friendly extensible software framework, along with a graphical user interface and thorough documentation, where existing inference algorithms implemented in R will be used in the background following good software development principles. The proposed research will provide high-valued software for the biomedical researcher and the student. We will develop algorithms that can efficiently build biological graphs from a large number of diverse inputs obtained from disparate experiments and sources. We will implement the proposed software in user friendly framework to be able to provide them in a convenient package for use by the biomedical community.

Team