AI Maps Mitochondrial Protein Interactions to Reveal Disease Targets

Researchers at Texas A&M AgriLife Research have used artificial intelligence to map likely interactions among human mitochondrial proteins, offering a faster route to understanding disease-linked molecules whose functions have long remained unclear. The work, published in Nature Communications, applies Google DeepMind’s AlphaFold-Multimer, an AI system originally developed in 2022 to predict protein complex structures, as a classifier to screen possible pairings among roughly 1,100 mitochondrial proteins. From more than 630,000 candidate interactions, the team identified 2,895 likely protein partnerships, more than half of which had not previously been reported. The study addresses a major gap in human biology: many mitochondrial proteins lack known biochemical roles, yet mutations in about 30 percent of mitochondrial proteins have been associated with diseases, including cancer, Parkinson’s disease, Alzheimer’s disease and rare inherited metabolic disorders. Because protein function is often inferred from interaction networks, linking unknown proteins to partners with more established roles provides clues about their biological activities and potential involvement in disease pathways. Led by AgriLife Research scientist Abhinav Swaminathan in the laboratory of Vishal Gohil, a professor and Chancellor EDGES Fellow in the Texas A&M Department of Biochemistry and Biophysics, the project began as a way to evaluate difficult-to-assay protein relationships in the mitochondrial copper pathway. Encouraged by initial results, the team expanded the approach to the entire human mitochondrial proteome. Laboratory experiments in yeast and human cells, including work conducted with researchers at Washington University School of Medicine in St. Louis, confirmed several AI-predicted interactions over a roughly two-year validation period. The findings suggest that AI-based interaction maps can guide experimental researchers toward the most promising hypotheses, reducing the time and cost of probing unknown proteins. The team has released its predictions through MitoMatch, a free online resource that allows scientists worldwide to identify likely partners for mitochondrial proteins in their own studies. By assigning probable interactions to 85 previously uncharacterized proteins, the study also supplies new starting points for basic science and disease research. While treatments for mitochondrial disorders remain years away, the work demonstrates how biochemistry and genetics laboratories can use advanced AI tools creatively to tackle large-scale biological questions. The mitochondrial interactome now serves as a practical roadmap for future experiments aimed at identifying diagnostic markers and therapeutic targets.