Evolutionary and drug resistance landscape of human tyrosine kinases

The National Institutes of Health National Human Genome Research Institute awarded The Pennsylvania State University $806,828 on February 16, 2026, under the Human Genome Research program (CFDA 93.172) to conduct comprehensive functional prediction and experimental validation of mutations across human tyrosine kinases.

The research employs a hybrid experimental and computational approach to predict the effects of point mutations across tyrosine kinases in the presence and absence of xenobiotics. In Aim 1, the recipient generates mutational scanning data across at least 15 diverse tyrosine kinases alongside 3 ancestral reconstructions and develops family-wide models of mutational effects using Gaussian process regression and transformer-based protein language models. Aim 2 completes benchmarking of CRISPR-Cas9 base editors for variant annotation and studies segregating variants in kinases and their interactions with xenobiotics. Aim 3 investigates higher-order epistatic interactions inferred from kinase super-family sequence analysis.

The award period runs through January 31, 2030, with place of performance in Pennsylvania. The total obligated amount is $806,828.

Cold Spring Harbor Laboratory serves as a sub-grantee, collaborating on the evolutionary and drug resistance landscapes of tyrosine kinases under sub-award R01HG013692-S008002-DHHS, awarded June 22, 2026.

Abstract

Tyrosine kinases cause mendelian disorders, cancer, contribute to polygenic risk scores in common diseases, and have led to 62 approved drugs for cancer, inflammatory disorders, and lung fibrosis. However, the function of every kinase domain residue across the large number of existing germ-line and somatic variants in the kinome remains largely unknown. While current techniques allow US to comprehensively measure the effects of every possible mutation in a single gene, we still do not know how to scale residue-by-residue functional discovery across a gene family. We hypothesize that the shared evolutionary and functional characteristics across a protein family can be leveraged to make comprehensive functional predictions of mutations, both in the context of the reference genome and in the context of other segregating and somatic variants. Here we propose a hybrid experimental/computational approach to comprehensively predict and test the effects of point mutations across all tyrosine kinases in the presence and absence of xenobiotics. This joint approach could dramatically decrease the experimental scale of residue annotation by exploiting the phylogenetic understanding that mutational effects across gene families are highly – but not completely – conserved. In aim 1 we will generate mutational scanning data across at least 15 diverse tyrosine kinases alongside 3 ancestral reconstructions and investigate family- wide models of mutational effects using gaussian process regression and transformer-based protein language models. In aim 2 we will complete a benchmarking effort for crispr CAS9 base editors as a tool for variant annotation and in parallel we will use our existing approach to study all editable segregating variants in kinases and their interactions with xenobiotics. In aim 3 we investigate higher-order epistatic interactions inferred from kinase super-family multiple sequence alignments and test our epistasis predictions with medium throughput experiments. The completion of these independent but synergistic aims will provide a comprehensive understanding of the genotype-phenotype relationship across human tyrosine kinases with respect to natural kinase function, xenobiotic drug phenotypes and protein coding variants relevant to hereditary disease. Moreover, IT will provide a proof-of-concept and a generalizable workflow for extending residue discovery genome-wide one protein family at a time.