A team from the University of Barcelona’s Faculty of Medicine and Health Sciences and the Institute of Neuroscience (UBneuro) applied advanced artificial intelligence techniques to better understand why Huntington’s disease can develop at very different ages in patients. This inherited neurodegenerative disease causes movement, cognitive, and psychiatric disorders and is caused by neurodegenerative disease mutations. HTT A gene encoding the huntingtin protein.
Mutations in this gene produce a series of CAG repeats that alter the properties and function of the huntingtin protein in the brain. The length of the CAG repeat is HTT Although genes influence the age at which first symptoms appear, this factor does not completely explain the wide variation observed in disease onset among patients. The study is currently analyzing what additional genetic factors may play an important role in determining when the disease begins in affected individuals.
The article was published in the 20th conference proceedings.th Machine Learning in Computational Biology Conference (MLCB, 2025). One of the most internationally recognized scientific forums exploring the frontiers of knowledge between machine learning and computational biology. The study was led by the research group of Ramon y Cajal researcher Jordi Avante (UB-UBneuro), and master’s student Caterina Fuse was the first author of the paper.
This study is a pioneering and innovative application of artificial intelligence language models to genomic information to predict phenotypes from multimodal genotypes. This study outlines a new framework for investigating complex genetic diseases and demonstrates how multimodal machine learning can help uncover biologically meaningful patterns that are difficult to detect using traditional methods.
Beyond traditional statistical methods
In this study, the researchers used nonlinear machine learning models such as tree-based models and graph neural networks (GNNs) to identify genetic modifiers, or genes that slow or accelerate the onset of the disease, depending on a patient’s genetic background. Unlike traditional statistical approaches, these models can detect complex interactions between genes and reveal effects that depend on the length of the CAG triplet expansion.
To make the analysis more efficient and easier to interpret, the team also developed a method to compress genetic information using gene-specific neural networks. This reduced computational cost without losing predictive power. Additionally, it incorporates gene expression changes predicted and generated by state-of-the-art genomic language models. This innovation allowed researchers to link regulatory DNA variants to changes in gene activity in brain regions affected by disease.
As part of the study, researchers analyzed genetic data from more than 9,000 Huntington’s disease patients. This allowed the team to identify both previously known modifiers associated with DNA repair and new candidate genes involved in processes such as transcriptional regulation and cellular metabolism. Of note, this result shows for the first time that different biological mechanisms may influence disease development in patients with shorter disease. versus Longer CAG extensions reveal the context dependence of these genetic effects.
This study shows that the genetic factors that modify Huntington’s disease are not universal and are highly dependent on genetic context. Nonlinear, multimodal machine learning can reveal interactions that are essentially invisible using traditional approaches. ”
Jordi Avante, principal investigator of the study, professor in the UB School of Medicine and Health Sciences, and UBneuro member
“This approach could be applied to other genetic and neurodegenerative diseases, opening new avenues of research and potentially enabling more personalized treatment strategies in the future,” Avante concluded.
