About
Computational biologist developing deep learning methods — primarily graph neural networks — for biomarker identification and protein dynamics analysis.
I am actively doing my research in the field of Computational & Systems Biology , at Department of Bioinformatics, University of North Bengal, where I am pursuing my PhD Coursework. I earned my M.Sc. in Bioinformatics from the same University in 2025, and my B.Sc. in Zoology (Hons.) from Kalimpong College in 2023.
My research sits at the intersection of deep learning, machine learning, graph theory, structural bioinformatics and systems biology. Biological systems have inherent structure — interaction networks, ontological hierarchies, spatial contact graphs — and I try to encode that structure directly into learning architectures rather than treating biology as unstructured tabular data.
I currently focus on two problems: identifying biomarkers from gene ontology combining knowledge graphs and natural language processing with graph neural networks, and predicting protein fluctuations using physics-constrained neural operators. Both efforts have produced novel algorithms and web tools for researchers, with manuscripts in progress and under review.
I am also interested in the broader methodological questions this work raises: how do we make hierarchical priors like the Gene Ontology differentiable? How do we handle oversmoothing when message-passing on biological networks? What are the failure modes of physics-informed losses in noisy regimes?
Additional information can be found in the publications, software, and research sections. For inquiries or potential collaborations, feel free to get in touch.
Currently developing LGNM and CABIgoWEB — both web-based tools serving the computational biology community.
Selected Contributions
-
Web Server
LGNM
Learned Graph Network Model for protein flexibility, normal modes, and conformational dynamics prediction.
-
Web Server
CABIgoWEB
Biomarker identification using gene-ontology features and deep learning. Combines Graph Neural Networks with Gene Ontology feature on PPI networks input.
-
Open Source
ScriptaDocX
Privacy-first document tools running entirely in the browser — PDF, OCR, merge/split, no server upload.
-
Research
Graph Learning on Biological Networks
Ongoing methodological work on ontology-aware embeddings, physics-informed graph operators, and interpretability in biological GNNs.