There has been tremendous progress in identifying the genetic basis of autism spectrum disorder (ASD). However, even in causal genes, individuals can carry novel mutations, or variants of uncertain significance (VUS), that lack a definitive diagnosis. Classifying VUS is essential for accurately diagnosing ASD and developing targeted treatments.
Many VUS are missense mutations that introduce single amino acid changes. A proven approach for classifying VUS is to make many mutations in a protein prospectively and measure how these mutations impact function with high-throughput assays. For example, in deep mutational scans, each position of the protein is mutated to all 19 other amino acids. This process produces lookup tables for identifying mutations that disrupt protein function, which can refine diagnoses.
Max Staller’s lab applies deep mutational scanning, rational mutagenesis and machine learning to study transcriptional activation domains, the regions of transcription factors that bind coactivator complexes and activate transcription. The SFARI Gene database contains 114 transcription factors, but these proteins contain only 39 annotated activation domains, reflecting the fact that activation domains are under-annotated. The Staller lab has computationally predicted and experimentally verified eight activation domains, seven of which were novel1.
In this project, Staller and his colleagues plan to perform deep mutational scans and targeted mutagenesis on activation domains of five transcription factors associated with autism, namely C2D1A, vitamin D receptor (VDR), NFE2L3, SRCAP and AHDC1, to build lookup tables and help refine ASD diagnoses. These data will train computational models to predict which mutations in activation domains disrupt function. These models hold the potential to predict the effects of any individual’s mutation, not just those cataloged, and thereby lay a foundation for molecular diagnostic models that will power precision medicine.