Genome-wide association studies (GWAS) identify genomic loci linked to disease risk in cats but rarely pinpoint causal variants. Marker SNPs used in current genetic screening tests often fail to generalize across cohorts and breeds, limiting diagnostic accuracy and mechanistic value. This project will develop and apply AI-based models that identify candidate causal non-coding variants underlying immune-related traits and diseases in cats by learning how individual SNPs alter cis-regulatory element (CRE) activity in relevant immune cell types. By integrating cell-type-resolved regulatory genomics with existing GWAS data, this work will translate association signals into mechanistic insight relevant to feline health, breeding decisions, and clinical diagnosis.

Aim 1 will train AI models predicting how non-coding genetic variation affects transcription initiation and chromatin accessibility across multiple immune cell types. We will collect PRO-cap and PRO-seq data from primary feline blood cells, use these data to train CLIPNET deep learning models predicting how DNA sequence changes affect regulatory function, and apply these models to identify candidate causal variants underlying blood cell traits.

Aim 2 will identify candidate causal SNPs underlying feline chronic gingivostomatitis (FCGS), a painful inflammatory oral disease. We will generate PRO-cap and PRO-seq data from diseased oral mucosa tissue, integrate these profiles with single-cell ATAC-seq data, and train CLIPNET models to identify candidate causal variants and the cell types responsible for FCGS.

Together, these aims will establish a generalizable framework for identifying causal variants from GWAS data in cats, applicable to additional immune-mediated conditions such as feline infectious peritonitis, eosinophilic keratoconjunctivitis, and inflammatory bowel disease, yielding more accurate genetic markers and new mechanistic understanding of feline immune and blood disorders.