A new study of more than 15,000 Latin American individuals has found that autism risk genes identified in that population overlap substantially with those found in European ancestry cohorts — leading the researchers to conclude that the core genetics of autism are universal across ancestries. The finding, published Monday in Nature Medicine, is one of the largest genomic studies of autism in a non-European population. But two experts not involved in the work say the conclusion goes further than the data warrants.
The study, led by first author Marina Natividad Avila and senior author Joseph D. Buxbaum of the Seaver Autism Center for Research and Treatment at Mount Sinai, examined more than 18,000 genes across roughly 4,700 individuals with autism and 10,300 without, all from Latin American populations across 10 collection sites spanning Brazil, Colombia, Costa Rica, Mexico, Peru, and sites in California, Florida, and New York. The team identified 35 genome-wide significant autism-associated genes — and found that 19 of those 35 were also strongly associated with autism in a separate European ancestry dataset, suggesting the genetic architecture of autism is largely shared across populations. "Our results indicate that the core genetic architecture of autism is shared across ancestries," Buxbaum told GEN News. "This suggests that the biology underlying autism is universal."
That framing has roots in the Autism Sequencing Consortium, which Buxbaum co-founded in 2010 to address a persistent problem in autism genomics: nearly all large genetic studies had been conducted in European ancestry populations, leaving clinicians with limited reference data when interpreting genetic tests for non-European patients. The new study was conducted through the Genomics of Autism in Latin American Ancestries (GALA) Consortium, assembled in part to close that gap.
The practical case for more diverse reference data is not in dispute. As GEN News reported, the lack of ancestral diversity in existing datasets has contributed to higher rates of inconclusive genetic test results among non-European individuals — because the tools used to interpret variants were built on data that skews European. The study also found that one class of conservation metrics used to prioritize candidate autism genes in clinical analysis may overestimate conservation overall, a limitation the authors trace back to European-ancestry bias embedded in the metrics' reference datasets.
But whether the data supports a claim of universal biology is where the scientific consensus thins.
Maria Chahrour, a neuroscientist at UT Southwestern who has studied ancestry-specific inherited variants in East African autism cohorts, reviewed the preprint version of the paper last year. Her takeaway, as she told The Transmitter in February 2025, was that the universality conclusion was "a bit of an overreach." The logic, she explained, is that even if a core set of genes overlaps across populations, the way those genes are expressed — or the degree to which they produce autism traits — may vary by genetic background. "Even if core gene sets overlap, the expression or penetrance may be modulated by population-specific genetic backgrounds," she said.
Dennis Lal, a neuroscientist at UT Health Houston, is more sympathetic to the study's core claim but draws a line at what it applies to. The 35 genes the paper highlights are linked specifically to autism with high support needs — sometimes called profound autism — cases involving significant intellectual disability and minimal verbal ability. That focus matters, because it leaves open whether ancestry and environment might play a larger role in the genetics of autism that manifests less severely. "That leaves open the possibility that ancestry and environment could have a greater effect on other forms of autism," Lal told The Transmitter.
There is also a methodological note worth flagging for anyone using these results to build clinical tools: the 19 genes that replicated across ancestries are the ones most likely to replicate, because the analysis was designed to find genes with effects strong enough to appear across diverse genetic backgrounds. That's a reasonable scientific strategy — and a reasonable clinical priority — but it is not a neutral description of all autism genetics.
Buxbaum, for his part, has made no secret of where he lands. In commentary on the preprint last year, he offered a characteristically blunt summary of his view: there is no ancestry-specific genetic architecture of autism. "We've shown there is no cat," he said, as The Transmitter reported — a remark that was intended as a dismissal of the possibility that autism genetics varies meaningfully by ancestry.
The question is whether that confidence is warranted. The GALA cohort is an important correction to a field that has historically excluded non-European populations. That contribution stands on its own regardless of how the universality claim is framed. Whether the biology of autism is genuinely universal — or whether this study found that the most severe, high-support-needs forms of autism are genetically similar across ancestries — is a meaningfully different question, and the distinction matters for how the findings get used.
One gene flagged in the study, MARK2, is worth watching: it has not been definitively linked to autism but appeared in severe developmental delay datasets, and the researchers flagged it as a candidate worth following up. That kind of lead is exactly what more diverse cohorts are supposed to produce. The reference data problem is real, and fixing it matters for diagnosis today — not just for the science of tomorrow.
