Ashwin Gopinath

MetaMuse: A Multi-Agent AI System for Biomedical Metadata Curation and Harmonization

MetaMuse curates and harmonizes biomedical metadata using a multi-agent AI system, achieving high accuracy and scalability in standardizing unstructured data from repositories like GEO.

GeneJEPA: A Predictive World Model of the Transcriptome

This paper introduces a self-supervised foundation model for single-cell transcriptomes that predicts gene sets from context and outperforms baselines in downstream tasks.

REVIVE-Flow: A Foundation Model for Blood DNAm Aging

Revive-Flow predicts and reverses blood DNA methylation aging using a flow-matching model trained on epigenetic data, achieving high accuracy in rejuvenation.

Guided multi-agent AI invents highly accurate, uncertainty-aware transcriptomic aging clocks

The paper introduces K-Dense, an AI system that creates accurate transcriptomic aging clocks with uncertainty estimates, showing continuous evolution of aging signatures.

The Limitations of TabPFN for High-Dimensional RNA-seq Analysis

This study evaluates TabPFN's performance on high-dimensional RNA-seq data and finds that adaptations underperform compared to classical methods when using full feature sets.

Transfer Learning Of Gene Expression Using Reactome

This paper introduces a gene expression transfer learning model that enables cross-species gene mapping using reactome profiles, improving functional gene understanding and drug response prediction.

Gene Expression Reactomes Across Species Do Not Correlate with Gene Structural Similarity

The paper measures gene expression responses to drugs in mouse and rat genomes across organs and finds that reactomes do not correlate with structural similarity, proposing a new cross-species mapping approach based on expression patterns.

Method for Identifying Proteomic Biomarkers of Health

The paper presents an algorithm to identify proteomic biomarkers for health status, applying it to find 9-protein indicators for atopic dermatitis and 6-protein indicators for heart failure.

BIOFORMERS: A SCALABLE FRAMEWORK FOR EXPLORING BIOSTATES USING TRANSFORMERS

The paper introduces BIOFORMERS, a transformer-based framework for analyzing biological systems, demonstrating its ability to capture biological insights from single-cell transcriptomics data.

High Frequency Longitudinal RNAseq Reveals Temporally Varying Genes and Recovery Trajectories in Rats

This study uses longitudinal RNAseq in rats to identify temporally variable genes and recovery trajectories after chemical exposure.