Decoding the molecular grammar of repressive condensates using spatial-functional screens

The National Institute of General Medical Sciences awarded The Leland Stanford Junior University $158,696 on February 19, 2027, under the Biomedical Research and Research Training program (CFDA 93.859) to decode the molecular grammar of repressive condensates using spatial-functional screens.

The research integrates rational deep variant scanning, a protein language model trained on intrinsically disordered regions (RAVEN), image-enabled cell sorting, and pooled reporter-based functional genomics to systematically study how sequence features in heterochromatin protein 1A (HP1A) intrinsically disordered regions encode nuclear condensate formation, spatial organization, and transcriptional repression. Aim 1 uses machine-learning-guided variant libraries to perturb charge patterning, aromatic clustering, and motif organization across HP1A intrinsically disordered regions. Aim 2 maps how these variants alter condensate formation and nuclear localization through high-throughput single-cell imaging. Aim 3 determines how intrinsically disordered region sequence grammars encode transcriptional repression using a chromosomally integrated reporter system. The work compares spatial and functional grammars to establish whether condensate formation and gene silencing arise from shared or separable determinants.

Work is performed at Stanford's Palo Alto, California campus. The period of performance runs from February 19, 2027, through August 17, 2029. This is a Project Grant, the standard assistance type under NIGMS's mission to support basic research that increases understanding of biological processes and lays the foundation for advances in disease diagnosis, treatment, and prevention.

Abstract

Intrinsically disordered regions (IDRS) make up nearly one-third of eukaryotic proteomes and are especially enriched in transcriptional regulators, where they drive the formation of nuclear condensates and control gene expression through multivalent interactions. Yet despite their central role, we lack a predictive understanding of how IDR sequence features encode condensate formation, spatial organization, and transcriptional repression in cells. Heterochromatin protein 1Α (HP1Α), a major effector of gene silencing, provides an ideal model: ~50% of the protein consists of three IDRS whose collective sequence properties shape chromatin compaction, heterochromatin condensates, and transcriptional repression. This proposal will define the molecular grammar of HP1Α IDRS by integrating rational deep variant scanning (DVS), an IDR-trained protein language model (RAVEN), image-enabled cell sorting (ICS), and pooled reporter- based functional genomics. Aim 1 will use machine-learning-guided DVS libraries to systematically perturb charge patterning, aromatic clustering, and motif organization across the HP1Α IDRS. Aim 2 will map how these variants alter condensate formation and nuclear localization using high-throughput single-cell imaging. Aim 3 will determine how IDR sequence grammars encode transcriptional repression using a chromosomally integrated reporter system. By comparing spatial and functional grammars, this work will establish whether condensate formation and gene silencing arise from shared or separable IDR determinants. The proposed research is significant because IT develops a predictive, mechanistic framework for how IDRS encode nuclear organization and gene regulation. IT is innovative because IT unites protein language models with high-throughput spatial phenotyping and functional screening in living cells. Together, these studies will transform our ability to interpret IDR sequence variation and will provide generalizable principles for how disordered regions govern genome function.