New AI Method Refines Drug Target Discovery for Brain Disorders

For two decades, investigators have relied on genome-wide association studies (GWAS) to identify single nucleotide polymorphisms (SNPs) associated with complex diseases. While these studies have been instrumental in mapping the genetic architecture of various conditions, translating their findings into clinical therapies has been a sizeable hurdle. A key limitation is that the GWAS process typically identifies millions of genetic regions, many of which do not fall within a specific gene or instead span noncoding sequences. As a result, determining the gene affected by a given variant and its relevance to disease often involves a protracted process that slows progress toward new therapies.

To address these challenges, a Cleveland Clinic research team has developed a new genomic analysis framework called GenT that uses artificial intelligence (AI) to identify disease-associated genes and potential drug targets. Recently published in Nature Communications, GenT offers a sophisticated alternative to traditional methods of interpreting DNA sequences, with the potential to advance laboratory discoveries, drug development and clinical diagnostics.

From SNPs to gene-level inference

In traditional GWAS, gene-level inferences are typically drawn by assigning the so-called lead SNP, the variant with the smallest P value in a locus, to its closest physical gene. However, functional genomic investigations show that only about one-third of lead SNPs accurately tag the actual causal genes, and a mere 5% of lead SNPs are likely to be causal themselves.

This traditional approach requires strict statistical corrections for all genome-wide tested SNPs, which puts significant limits on statistical power and undercuts the ability to identify “druggable genes,” or those whose protein products can be targeted by existing or investigational ligands.

GenT takes a different approach. Instead of testing millions of variants individually, it groups variants around different genes into sets. By performing a joint test on these gene-specific SNP sets, the framework takes advantage of the shared heritability of multiple variants.

“I don’t use the word innovation lightly, but this method can really change how people analyze genetic data,” says the study’s senior and corresponding author, Feixiong Cheng, PhD, Director of the Cleveland Clinic Genome Center. “GWAS gave us the map, and now GenT helps find the landmarks — the genes that matter for disease pathogenesis, progression and drug development.”

Leveraging multi-omics at global scale

Beyond gene-based testing, the framework includes specialized extensions to capture the complexity of human biology:

• The MuGenT tool integrates ancestry data to harness shared heritability across various populations to identify shared genetic drivers that might be missed in single-population studies.
• The xGenT tool incorporates multi-omics data, such as brain-specific protein levels and gene expression. By weighting genetic variants based on their functional impact in specific brain tissues, xGenT helps prioritize the most likely driver genes in crowded genomic regions.

Using these advanced tools, Dr. Cheng’s team analyzed over 18,000 genes across several complex conditions. GenT identified a significant number of high-confidence candidate genes for several neurodegenerative and psychiatric disorders, including 16 for Alzheimer’s disease, 15 for amyotrophic lateral sclerosis, 35 for major depression and 83 for schizophrenia. Notably, many of these genes were found in novel locations.

Validation in Alzheimer’s disease

A notable example is the gene SYK, which is highly expressed in brain tissue but failed to reach significance in previous studies. GenT identified it as a high-confidence driver of Alzheimer’s disease risk. Biologically, SYK works with the TREM2 receptor in microglia to help clear amyloid-beta, a primary hallmark of Alzheimer’s pathology.

The study’s potential clinical relevance was further bolstered by experimental validation of NTRK1, a gene identified by xGenT as a potential druggable target for Alzheimer’s disease. Although NTRK1 is known to be vital for nerve cell survival, any direct role for it in Alzheimer’s risk had previously lacked strong genomic support.

The researchers treated patient-derived neurons with a selective NTRK1 inhibitor and observed a significant, dose-dependent reduction in tau hyperphosphorylation. Given that neurofibrillary tangles composed of hyperphosphorylated tau are a primary driver of neurodegeneration, these findings suggest that targeting NTRK1 could be a viable strategy for alleviating tau-related pathology, the researchers note.

Dr. Cheng and his team are now using GenT to pinpoint new drug targets for Alzheimer’s disease by analyzing datasets from the National Institute on Aging’s Alzheimer’s Disease Sequencing Project. He says that pairing advanced AI tools like GenT with emerging human model systems, including brain organoids, could improve Alzheimer’s treatments in the near future.

The Cheng Laboratory has published the GenT software and results online, inviting other genetic researchers to explore their new method for drug discovery.

This work was supported by the National Institute on Aging (U01AG073323).