1 November 2024
Ayan Paul
The Institute for Experiential AI
Northeastern University
One of the core organizing principles of eukaryotic genomes is the modular organization of genes into multiple protein-coding exons separated by non-coding intronic sequences. For proteins to be translated from mRNA, these intronic sequences must be spliced out of immature transcripts before producing a mature mRNA transcript isoform. Nearly all human genes undergo alternative splicing, resulting in proteome diversity. The complex and nuanced process of splicing is far from being fully charted out with both cis- and trans-acting factors playing significant roles and often interacting in a highly non-linear manner to produce mature mRNA isoforms. With high-throughput short- and long-read RNA sequencing becoming a reality bringing about a wealth of data from a variety of tissues and cell lines, data-driven methods, like machine learning, are becoming instrumental in charting the process of alternative splicing that has eluded detailed dynamical modeling. In this "chalk" talk, I will discuss the intricacies of alternative splicing and how machine learning is being used to map transcriptomic sequences and actions of trans-regulatory factors to mRNA isoform diversity.