Devansh Pandey
I am a PhD student in Computational Biology and Bioinformatics at The University of Texas at Austin, advised by Dr. Vagheesh M. Narasimhan . My research focuses on developing multimodal AI frameworks that integrate imaging, genetics, and metabolic data to improve precision preventive cardiology and understand human evolution.
Previously, I completed my dual degree (B.Tech & M.Tech) in Biotechnology and Biochemical Engineering at the Indian Institute of Technology (IIT) Kharagpur.
- Multimodal AI
- Deep Learning
- Genomic Foundation Models
- Precision Preventive Cardiology
- Human Evolutionary Genomics
Research Interests
Genetic Architecture of Atherosclerosis
Investigating the genomic basis of atherosclerosis and its causal links to cardiovascular disease progression and clinical endpoints.
Precision Risk Stratification
Enhancing population-level risk prediction for Coronary Artery Disease (CAD) through the integration of large-scale imaging and genetic data.
Multimodal Phenotyping & Drug Discovery
Leveraging multimodal data to derive precise phenotypes for identifying novel, druggable genetic risk loci for cardiovascular diseases.
News
Research Experience
- Investigating the genetic architecture and heritability of rupture-prone plaque features using imaging and genomic data from 30,000 individuals in the SCAPIS cohort.
- Determining the genetic correlation between high-risk plaque features and major cardiovascular outcomes (CAD, stroke).
- Developed a novel multimodal deep learning framework to predict Coronary Artery Disease (CAD) risk, integrating imaging, genetics, and metabolic data from 65,000 UK Biobank participants.
- Demonstrated superior predictive accuracy (C-index 0.75) over established clinical tools, identifying a patient subgroup with a 15-fold elevated risk.
- Engineered a genomic foundation model by pre-training a GPT-2 architecture on the 1000 Genomes Project.
- Achieved a validation loss of 0.05 for next-variant prediction and improved performance over traditional HMM-based methods.
- Developed and applied a multi-locus genotype statistic (G12) on ancient DNA from 708 individuals spanning 7000 years.
- Identified 14 candidate regions of selection and showed how demographic processes obscured early adaptive signals.
Selected Publications
In Revision at Radiology: Artificial Intelligence.
npj Digital Medicine (2026).
Circulation, 152(Suppl-3):A4342257 (2025).
Science, 388, eadq1521 (2025).
Nature Communications, 15, 9772 (2024).
Service & Teaching
BioML Society | The University of Texas at Austin
Founding Member (Jan 2023 - Present)- Designed and conducted an eight-week seminar series on Bioinformatics and ML tools, attracting over 100 attendees.
- Developed seminar materials that were subsequently adopted as the curriculum for a new graduate course at UT Austin.
Principles of Computational Biology | UT Austin
Teaching Assistant (Aug 2024 -
Present)
Preparing and grading assignments and delivering tutorials on genomic data analysis using Python and R.