AT CAPACITY - Laksshman Sundaram gives Special YCC/CSEI Seminar- REGISTRATION CLOSED

Abstract:

This talk presents our approach to virtual cell modeling by building computational representations of core cellular processes—transcription, translation, and cell state dynamics—using specialized models: Epinet for transcription, NVCell for single-cell multiomic state modeling, and CodonFM for translation. In this talk, we focus on CodonFM and introduce a suite of codon-resolution language models trained on over 120 million coding sequences from 22,000 species. Unlike traditional protein models that ignore synonymous codon variation, cdsFM captures the functional and regulatory impact of codon choice, learning the structure of the genetic code and offering new insights into how coding sequences shape protein expression in the cell.