A large sequencing study found substantial overlap in rare autism-associated genes between Latin American participants and earlier cohorts dominated by European ancestry. The result strengthens evidence that important parts of autism's genetic architecture are shared across populations, while exposing how much genomic research still depends on narrow reference data.
The Genomics of Autism in Latin American Ancestries, or GALA, analysis included 15,427 people: 4,717 participants with an autism diagnosis and 10,710 controls. The researchers combined exome and genome sequencing data from cohorts across North, Central and South America and published the findings in Nature Medicine.
The Study Tested Rare Coding Variation
Autism has a complex genetic architecture. Common variants account for much of population-level genetic liability, while rare variants, including changes that arise de novo rather than being inherited from either parent, can make a substantial contribution for some individuals. The GALA study focused on rare deleterious coding variants rather than claiming to capture every genetic or non-genetic contributor.
Researchers analyzed more than 18,000 genes with a Bayesian association method. They identified 35 genes associated with autism at a false discovery rate below 0.05; 16 met a threshold below 0.01 and eight met a threshold below 0.001. Genes involved in regulation of gene expression, neuronal communication and cytoplasmic functions were represented among the results.
The Signals Overlapped With Earlier Cohorts
The 35 genes showed substantial overlap with genes identified in sequencing studies composed largely of people with European ancestry. Signal was concentrated in highly constrained genes, where damaging changes are rare because the genes perform important biological functions. Very rare de novo variants drove much of that association.
The study also added evidence for emerging autism-associated genes, including MARK2, YWHAG, PACS1, RERE, SPEN, GSE1, GLS, TNPO3 and ANKRD17. These are statistical gene-disease associations across a cohort. They do not mean that a variant in one of these genes determines whether any individual is autistic.
The authors concluded that they detected no ancestry-related difference in the core rare coding-variant architecture examined. That is narrower than saying every genetic influence is identical across all populations. The analysis did not test every type of variant, every ancestry or every environmental and social factor that shapes diagnosis and lived experience.
Latin American Diversity Made the Test Stronger
Much autism gene discovery has relied on European-ancestry samples. That imbalance can make variants harder to interpret in patients whose ancestry is underrepresented in reference databases. Latin American populations often combine Indigenous American, African and European ancestry, making them especially informative for testing whether established gene signals generalize.
GALA brought together multiple recruitment sites and existing cohorts. Autism diagnoses were based on expert clinical evaluations using DSM-5 criteria and available standardized assessments. The study excluded people with a known genetic condition such as fragile X syndrome from its primary analyses.
Not every GALA site contributed sequenced samples to this analysis, and some samples had appeared in previous large sequencing efforts. The paper reports that 1,613 samples, including 707 autism probands, were analyzed for the first time here. Those details prevent the total cohort size from being mistaken for an entirely new or uniformly recruited population.
Conservation Scores Were Reliable at the Strongest End
The researchers also examined gene-constraint metrics, which estimate how strongly evolution has selected against disruptive variation in a gene. Some widely used measures may overstate constraint overall because European reference populations contain less genetic diversity than African populations.
For the most constrained genes, however, the metrics remained stable across the ancestral comparisons. That matters because many high-confidence neurodevelopmental-disorder genes fall in this strongly constrained group. Adding more diverse sequence data should improve interpretation of less-constrained genes and reduce uncertainty around variants that are rare in existing databases.
This is a reference-data result as much as a gene-discovery result. A variant may appear unusually rare when the comparison database does not adequately represent the patient's ancestry. Broader sampling helps laboratories distinguish truly unusual changes from population variation that was simply missing from an older dataset.
The Findings Support Testing, Not Genetic Certainty
The authors say the overlap supports the use of established clinical genetic approaches across ancestries, particularly for highly deleterious variants when ancestry-specific allele frequencies are incorporated. It does not turn sequencing into a complete autism diagnosis. Autism remains a clinical diagnosis, and many autistic people will not have a single rare coding variant that explains their traits.
The study also did not test a treatment, measure whether sequencing changed care or show that every reported gene is ready to guide an intervention. Genetic findings may sometimes inform counseling or identify a related condition, but their meaning depends on the exact variant, the evidence for that gene and the individual's clinical picture.
Precision Medicine Fails When Its Reference Population Is Narrow
The hard conclusion is methodological. Researchers cannot claim universal precision from databases built around a minority of the world's ancestry. GALA shows that many of the strongest rare autism gene signals do travel across ancestral backgrounds, but it also shows why that proposition had to be tested rather than assumed.
Expanding diverse sequencing is not a symbolic add-on to genomic medicine. It is how false rarity is reduced, associations are challenged and uncertain results become more interpretable. The study strengthens the shared-biology case without promising a universal test. That restraint is the point: better evidence comes from widening the population and narrowing the claim to what the data can actually establish.