Speakers: Jay Unruh, PhD

Jay Unruh, PhD

Director of Scientific Data
Stowers Institute for Medical Research

 

Intrigued by the molecular underpinnings of life, Jay Unruh pursued a B.S. degree in molecular biology at John Brown University and later a Ph.D. in chemistry from the University of Kansas. He continued his molecular studies in living cells during his postdoctoral studies in the Laboratory of Fluorescence Dynamics at the University of California, Irvine before joining the Stowers Institute as a research specialist in 2008. At Stowers, he extended his exploration of the molecular details influencing the cellular and organismal diversity. In 2010, Unruh took on the role of research advisor, then in 2015 he was appointed co-head of the Microscopy Center. In 2019, Unruh co-directed the Microscopy, Imaging, and Big Data group and in 2022, he accepted the role of Director of Scientific Data.

A Protein Language Model to Identify Signatures of Meiotic Transverse Filament Proteins

The synaptonemal complex is a genome-scale proteinaceous ladder-like structure that pairs and organizes homologous chromosomes during meiosis.  The core structural proteins (called transverse filaments) form elongated coiled-coil structures which are poorly conserved but manage to form very similar structures across the tree of life as demonstrated by light and electron microscopy.  Because of poor sequence conservation it has been difficult to identify unique features of these proteins and trace them across evolution.  Deep learning protein language models (like ESM2) have been available for a number of years but their use in mainstream biology has been limited.  This is likely due to the requirement for complex post-processing of the predictions to gain meaningful insight and identify regions of proteins important for function.  We have developed several strategies for easily adapting ESM2 outputs to predict complex protein features and demonstrated those strategies with a simple model to predict transverse filament sequences.  The resulting model identifies a novel conserved charge bias in the coiled-coil sequence among this class of proteins.