More than a thousand genes have been linked to autism. That diversity has long puzzled researchers: how can so many different genetic changes lead to overlapping traits? A study published in Science on 17 September 2026 offers part of an answer. A viral post summed it up as "hundreds of autism mutations converge on just two molecular patterns". That is broadly accurate, with important limits.
The Study
The team of Professor Kim Eunjoon at the Institute for Basic Science in South Korea studied 17 lines of mice, each carrying a different mutation linked to autism. Some affect synapses, the connections between neurons, others the way genes are switched on and off, others the signals inside cells. The researchers measured which genes were active in the prefrontal cortex, a brain region involved in social behaviour and planning. They analysed more than 1,000 brain samples and about one million individual cell nuclei from 205 male and female mice.
Two Opposite Brain States
Despite the very different starting mutations, the brains fell into two groups:
- Group 1: genes that build and run synapses were less active, while genes that organise DNA and process RNA were more active. Neurons and support cells changed more broadly.
- Group 2: the mirror image, with synapse genes more active and DNA and RNA programs less active, and more moderate changes in cells.
In other words, the brain seems to have a limited number of ways to respond to very different genetic changes. For research, that matters: instead of studying 1,000 genes one by one, scientists may be able to study a few shared brain states.
Three Surprises
- Sex mattered. In 7 of the 17 mouse lines, males and females carrying the same mutation ended up in different groups. The same gene does not act the same way in every body.
- Groups can change with age. Some lines kept their group as the mice developed, others switched.
- Drug responses differed. When young mice received fluoxetine or lithium, gene activity in Group 1 moved more consistently toward typical patterns than in Group 2. The authors stress that a change in gene activity does not mean a change in behaviour, and no behaviour was measured.
What About Humans?
The researchers also looked at existing data from the prefrontal cortex of 40 autistic people and 17 non-autistic people, analysed after death. They found two opposing subgroups there too. But the human patterns were not the same as in mice, showed more changes linked to the immune system, and could not be connected to specific mutations.
What the Study Does Not Say
The authors are clear on its limits, and so should we be:
- These are not two "types" of autism that a doctor could diagnose.
- They cannot predict a person's traits, needs or strengths.
- They do not tell anyone which medication would work. Fluoxetine and lithium were research tools here, not a recommendation.
- Most of the work was done in mice, which are models, not autistic people.
This study fits well with another 2026 finding, published in Nature, showing that many autism genes disturb the developing brain at the same moments in time. You can read our summary of that one here: Hundreds of autism genes, one shared brain pattern. Together they point in the same direction: autism is genetically diverse, but the brain's responses may be far less so.
What This Means for You
For now, nothing changes in how autism is recognised or supported. Autism is identified through traits and life experience, not through genes or brain samples. If you wonder whether you might be autistic, a validated questionnaire is a useful first step before talking to a professional: try the AQ-10 autism screener or the more detailed RAADS-14.
Source: Kim E, Bae M, et al. Institute for Basic Science, Center for Synaptic Brain Dysfunctions. Science, 17 September 2026. DOI 10.1126/science.adz6688. Study mainly in mice (17 mouse lines, 205 mice); human data from post-mortem prefrontal cortex of 40 autistic and 17 non-autistic people.