AI Accelerates Discovery of Genes Linked to Neurodevelopmental Disorders

Researchers, led by a team from Baylor College of Medicine, have developed a new artificial intelligence (AI) approach that has accelerated the identification of genes contributing to neurodevelopmental disorders (NDDs) such as autism spectrum disorder, epilepsy, and developmental delay. Study details, published in the American Journal of Human Genetics, show the new models can play an important role in fully characterizing the NDD genetic landscape, information that can improve diagnosis and the development of new targeted therapies.

“Although researchers have made major strides identifying different genes associated with neurodevelopmental disorders, many patients with these conditions still do not receive a genetic diagnosis, indicating that there are many more genes waiting to be discovered,” said first author Ryan S. Dhindsa, MD, PhD, an assistant professor of pathology and immunology at Baylor College of Medicine and principal investigator at the Jan and Dan Duncan Neurological Research Institute at Texas Children’s Hospital.

Despite significant progress in this area, the team noted in their research summary that thousands of NDD-associated genes remain to be discovered. For their work, rather than sequencing the genomes of large patient populations, the investigators developed a machine-learning approach, training their model on single-cell RNA sequencing data.

“We demonstrate that models trained solely on single-cell RNA sequencing data can robustly predict genes implicated in autism spectrum disorder (ASD), developmental and epileptic encephalopathy (DEE), and developmental delay (DD),” the researcher wrote.

Based on some of their early findings, Dhindsa said the team wanted to take their work one step further, to enhance their ML-base models. To do this, the researchers integrated more than 300 additional biological features, including gene intolerance to mutations, interactions with other disease-associated genes, and functional roles in biological pathways.

“These models have exceptionally high predictive value,” Dhindsa said. “Top-ranked genes were up to two-fold or six-fold, depending on the mode of inheritance, more enriched for high-confidence neurodevelopmental disorder risk genes compared to genic intolerance metrics alone.”

The study also found significant differences in gene expression patterns between genes with monoallelic (one copy) and bi-allelic (both copies) inheritance patterns in the developing human cortex. By refining these models, the researchers were able to improve their ability to predict genes associated with NDDs based on inheritance type.

The research offers potential benefits in diagnosing and treating NDDs. “We see these models as analytical tools that can validate genes that are beginning to emerge from sequencing studies but don’t yet have enough statistical proof of being involved in neurodevelopmental conditions,” Dhindsa said. This could expedite the discovery process and improve diagnostic accuracy for patients, particularly those who are currently undiagnosed despite extensive genetic testing.

The models developed by the researchers provide predictions for high-confidence NDD risk genes that could complement large-scale gene discovery efforts. As Dhindsa noted, “We hope that our models will accelerate gene discovery and patient diagnoses, and future studies will assess this possibility.”