Sajib Acharjee Dip

This paper introduces PerturbGraph, a graph neural network framework that predicts transcriptional responses to unseen gene perturbations by integrating biological priors, outperforming existing methods on benchmark datasets.

NicheAgent identifies spatial niches in transcriptomics data without training, using LLMs to refine assignments based on gene expression and spatial context.

This study evaluates lightweight LLM agents for improving spatial transcriptomics annotation using rule-based heuristics and multirole reasoning across multiple datasets.

UnCOT-AD improves Alzheimer's Disease prediction by integrating unpaired multi-omics data using a novel deep learning framework with cycle consistency and contrastive learning.

DeepAge uses deep learning to estimate biological age from DNA methylation data, outperforming existing methods by capturing complex interactions and improving accuracy.

PathoLM is a pathogen language model that improves pathogenicity identification in bacterial and viral sequences by leveraging pre-trained DNA models and minimal data, outperforming existing methods.