Advancing Genomic Interpretation for Chromatin Remodeling Enzymes

PROJECT SUMMARY/ABSTRACT The functional interpretation of human genetic variation needs to catch up to our ability to catalog the DNA sequences of individuals. Better tools are required to interpret mutations’ effects, improving our ability to diagnose and treat diseases. Our goal is to accurately predict which mutations cause human disease by

developing tools for three-dimensional and context-dependent interpretation of genetic information. We focus on genes that regulate the genome, known as epigenetic enzymes. We will specifically study the chromatin remodeler, SMARCA4, which enables the genome to be used in the right amount at the right time. SMARCA4 is

critical to study because 1) it is required for normal physiology and development, 2) germline mutations define rare and undiagnosed diseases that affect people of all ethnicities and genetic ancestries, 3) somatic mutations underly cancer development from many body tissues, making new findings of high biomedical value. Additionally,

epigenetic regulators are often associated with more than one medical condition. This data conveys epigenetic enzymes' powerful and multi-faceted functions and their ability to orchestrate many different physiologic changes that can differ across individuals. Further, the distinct physiological contexts that SMARCA4 acts within will vary

across body tissues and over time. SMARCA4 is the catalytic unit within the BAF complex, which also exists as different sub-complexes, necessitating the study of each. This concordance affords a pivotal opportunity to develop new procedures for helping more patients with diverse congenital and somatic diseases and better

understand this chromatin remodeler's normal functioning. Our central premise is that SMARCA4 mutations alter specific features of the encoded 3D protein in quantifiable ways using novel computational tools that generate mutation-specific functional calculations. This is important since current guidelines for genetic diagnosis are

primarily based on linear sequences rather than 3D features of gene products, and these guidelines fail to capture many pathogenic mutations that likely cause diseases. More importantly, our methodology applies to all diseases, independent of the affected cell or organ, and performs with the same high accuracy for cancers and

non-cancer diseases. Our proposal will generate specific deliverables that fill crucial knowledge gaps for advancing data science for genomics. Current data science methods in genomics are primarily based on sequence-based evolutionary conservation and population-level empirical observations. Thus, existing methods

fail to reveal genetic differences specific to individuals and fail to provide actionable mechanistic information. Using mutation-specific structural, dynamic, and systems-level annotations, our work offers a more powerful interpretive toolset across diverse clinical and research domains. We will benchmark our approach against

current tools, validate via clinical cases, and bring information of mechanistic value that can be tested experimentally. Finally, we will develop a scalable process that can, through future work, be applied to additional epigenetic enzymes.