Ravi Madduri

The paper analyzes gut microbiome metadata from the NCBI SRA to assess its structure, completeness, and reusability for large-scale studies and AI model development.

The paper introduces ARCHER-LD, a GPU-accelerated method to compute long-range linkage disequilibrium at biobank scale, achieving significant speedups over existing tools.

This paper develops an open-source LLM-based pipeline to identify patients with rare genetic aortopathies using clinical notes, achieving high accuracy in recommending genetic testing.

This study compares pooled analysis and meta-analysis for multi-ancestry GWAS, finding pooled analysis offers better statistical power while controlling population structure.

The paper introduces a closed-loop framework for single-cell foundation models that incorporates perturbation data to improve prediction accuracy, identifying therapeutic targets and pathways in RUNX1-familial platelet disorder.

The paper identifies pleiotropic loci across 1,167 traits using GWAS meta-analysis and shows they are enriched for functional annotations and improve polygenic score predictions.

scPrediXcan improves transcriptome-wide association studies by integrating deep learning for cell-type-specific gene expression prediction, identifying more causal genes and explaining more GWAS loci in type 2 diabetes and systemic lupus erythematosus.

The paper develops a multi-task learning approach to create context-specific gene network representations, applying it to colorectal tumors to predict therapeutic targets.

Population-stratified GWAS of 2,068 traits in 635,969 Million Veteran Program participants, 29% genetically non-European, yields 98,485 associations and 57,601 fine-mapped signals with largely shared cross-population architecture.

This paper optimizes the SAIGE algorithm using GPUs to accelerate GWAS on large datasets, achieving 20x speedup on the Million Veteran Program data.

The paper identifies monocyte-endothelial interaction alterations in TET2 CHIP and shows that targeting CXCR2 and ICAM1 can restore normal interactions.

This study identifies 447 new genetic variants associated with prostate-specific antigen levels across multiple ancestries and shows improved cross-ancestry prediction of PSA, with implications for personalized prostate cancer screening.

This paper introduces RNA-Cascaded-Diffusion-Model to generate synthetic whole-slide image tiles from RNA-Seq data, improving machine learning model performance in low-data settings.

This paper conducts a multi-ancestry GWAS meta-analysis of suicide attempt, identifying 12 genome-wide significant loci and finding shared genetic risks with ADHD, smoking, and risk tolerance.

This study identifies genetic variants associated with severe COVID-19 and finds shared genetic architecture with complex diseases using EHR data from the Million Veteran Program.

This paper shows that reducing tractography streamlines by two orders of magnitude does not significantly impact connectome identifiability, suggesting computational optimizations are feasible.

MaPPeRTrac streamlines and accelerates large-scale diffusion MRI tractography by automating the pipeline and leveraging HPC resources.

The paper develops a workflow to identify transcription factor binding sites across 27 human tissues using DNase-seq data, resulting in a comprehensive footprint atlas with high-quality predictions.

The paper presents tools for ensuring FAIRness in big biomedical data analysis through a case study on creating a transcription factor binding site atlas from ENCODE data.

BDQC identifies data outliers and anomalies in large-scale genomic pipelines without domain knowledge, improving data quality and enabling new insights.