Peer reviewed AI-assisted

A machine-learning microexon splicing code links rare variants to autism in 7,391 families

A gradient-boosted classifier trained to recognise neuronally spliced microexons from their surrounding sequence was used to score variants in whole genomes from three autism cohorts, prioritising 72 variants associated with autism within families that occur in 2.5% of autistic individuals and in none of their unaffected siblings; minigene assays in human induced neurons confirmed that representative variants alter microexon splicing.

Model
Gradient-boosted decision trees (XGBoost) microexon code-model
Field
Biology
Date
2026-10-14
Human collaborators
Guillermo E. Parada, Hannes Bretschneider, Quaid D. Morris, Benjamin J. Blencowe

What was found

Microexons are 3 to 27 nucleotide exons, mostly in neuronal genes, whose splicing is often disrupted in autistic brains. The study first profiled bulk and single-cell RNA-sequencing data to find 1,514 previously unknown human microexons and 1,115 mouse ones, then trained the classifier to separate sequences around true neural microexons from microexon-like background, so that a variant's effect on splicing can be scored as a change in the model's output. Applied to the SSC, SPARK and MSSNG cohorts (25,132 individuals in 7,391 families), rare and ultrarare variants had the largest predicted impact, the most affected microexons recurred across all three cohorts, and the 72-variant set comprises 28 ultrarare variants prioritised by this model and 44 also prioritised by DeltaSplice, an independent splicing predictor. About 82% of carriers had no previously detected candidate variant, and the variants concentrate in synaptic genes.

Novelty check

Europe PMC, 2026-11-12: microexon AND autism AND ("machine learning" OR "deep learning" OR model) returned 25 records; this paper is the only one linking variant-level microexon splicing predictions to autism status within families, the others being microexon function studies, reviews and an oligonucleotide therapy proposal. The paper's own discussion cites earlier machine-learning work on splicing variants in neurological disorders and studies of exons differentially spliced in autism, so the advance is the microexon-specific code and its cumulative-impact analysis. No earlier entry in this registry covers splicing or microexons.

Caveats and known objections

Peer-reviewed in Nature, open access, by the developers; the authors declare no competing interests, and there is no independent replication. The genetic result is an association: variants are predicted to disrupt splicing and their burden is tested within families, while the functional evidence is minigene reporter assays on a few representative variants, not editing of the endogenous genes or any phenotype. DeltaSplice alone did not reach significance, and 44 of the 72 variants required both models to agree. The manuscript was received in May 2024 and published in October 2026. Field is biology, not medicine: there is no clinical endpoint and the model is not a diagnostic. Autonomy is ai-assisted: the model is a conventional gradient-boosted classifier, and the study design, cohorts, statistics and experiments were human-driven, with the model scoring variants inside that analysis.

No independent check is recorded on this entry yet.

Reading the primary source closely enough to say whether it supports the claim counts as a check, and you are credited on the entry.

Or on GitHub: submit a check challenge the grade send a correction or send a pull request

Entry history (1 event)
  1. AddedEntered the registry graded Peer reviewed and AI-assisted.

Entries are never deleted. A grade that does not hold up is downgraded on the record, with the reason beside it.

Graded peer reviewed for verification and ai-assisted for autonomy. What these mean.

Cite this entry

Plain text
whataifound.org. (2026). A machine-learning microexon splicing code links rare variants to autism in 7,391 families. whataifound.org: A Registry of AI Scientific and Mathematical Discoveries. https://whataifound.org/finding/2026-10-14-microexon-code-autism-variants
BibTeX
@misc{whataifound-universityoftoronto-2026-variants,
  title        = {A machine-learning microexon splicing code links rare variants to autism in 7,391 families},
  author       = {{whataifound.org}},
  year         = {2026},
  howpublished = {whataifound.org: A Registry of AI Scientific and Mathematical Discoveries},
  note         = {Result by University of Toronto. Verification: Peer reviewed. Autonomy: AI-assisted.},
  url          = {https://whataifound.org/finding/2026-10-14-microexon-code-autism-variants}
}

Related findings

← All biology findings in the registry