Last month, PNGC had the pleasure of hosting Dr. Towfique Raj, PhD in partnership with the Perelman School of Medicine for the Biomedical Data Science Seminar Series. Dr. Raj is an Associate Professor of Neuroscience at the Icahn School of Medicine at Mount Sinai. His seminar addressed the post-GWAS challenges of variant and gene prioritization. By using pathogenic cell types, Raj has contributed to the identification of causal variants and genes, building toward a better understanding of the mechanisms and biological pathways underlying disease. Dr. Raj’s talk primarily focused on Alzheimer’s Disease (AD), and was divided into three parts.
First, he discussed the Microglia Genomic Atlas (MiGA), a genetic and transcriptomic resource generated using human myeloid cells from 255 individuals. Microglia, innate immune cells in the brain and spinal cord implicated in AD risk, are notoriously difficult to isolate, making MiGA the largest resource of its kind. The transcriptomic data he generated enabled the study of expression quantitative trait loci (eQTLs, variants correlated with a change in gene expression), especially age-associated changes in microglia gene expression. Additionally, he developed a computational approach, leveraging MIGA, to prioritize AD genetic variants and performed functional validation using iPSC microglia MPRA. This showed a significant overlap in eQTL effects and AD risk genes in microglia and peripheral monocytes.
Next, he discussed the long-read RNA-seq data his lab generated and how it can be used to study mRNA splicing. Long-read RNA-seq can be used to augment existing short-read transcriptomic datasets, allowing the study of full splicing QTL effects on loci of interest in AD or Parkinson’s disease. The novel full-length isoform structure he identified provides a great resource to the community to study novel isoforms and splicing events.
Lastly, he described collaborative work with David Knowles at Columbia University to build AI/ML-based cell type-specific models to help predict the effects of noncoding variants. New models are being developed to identify causal variants and quantify the risks of rare variants in disease. These models leverage integrative functional genomics from thousands of assays, spanning many measurement types.