Machine Learning Reveals Autism-Linked Variants in Microexon Regulation
Published in Artificial Intelligence, Bioinformatics, Brain/Neurology, Journals.
Published October 15, 2026.
A study published in Nature on October 14 uses machine learning to investigate how tiny gene segments called microexons are regulated in neurons. The work identifies genetic variants associated with autism that are predicted to disrupt the inclusion of these segments in RNA.
Researchers expanded the catalogue of microexons and built a sequence-based model of their regulation. Their family-based analysis covered 25,132 people from 7,391 families. Rare variants predicted to interfere with microexon splicing showed the strongest estimated effects.
The team also tested selected variants in human induced neurons using reporter assays, confirming that these examples altered splicing. The affected genes included known neuronal and synaptic genes as well as genes not previously linked to autism.
Why it matters: the study provides tools for investigating genetic variation outside the regions typically emphasized in gene-disruption studies.
Limits: genetic association and predicted molecular effects do not establish that every identified variant causes autism. This research does not provide a standalone diagnostic test or a demonstrated treatment.
Source: Parada and colleagues, Nature research article, published October 14, 2026.