Nona - A Novel Multimodal Masked Modeling Framework for Functional Genomics

Learn about Nona, a groundbreaking multimodal masked modeling framework for functional genomics, in this 51-minute conference talk from the Broad Institute's Models, Inference and Algorithms series. Discover how this neural network model operates on both DNA sequence and epigenetic tracks including DNase-seq, ChIP-seq, and RNA-seq at base-pair resolution, utilizing a flexible masking strategy to predict any subset of masked DNA and tracks from unmasked data. Explore how Nona supports existing sequence-to-function models and their applications such as variant effect prediction, while enabling novel application modes including context-aware predictions, functional language modeling, and functional genotyping. Gain insights into this versatile framework that extends sequence-to-function and masked language modeling to innovative applications in regulatory genomics, presented by Surag Nair, Senior ML Scientist at ReLU, BRAID, Genentech.