DESCRIPTION: Two new studies have identified common developmental pathways in eight rare genetic forms of autism. In eleven people with no known risk variants, the same investigation found two notable transcripts. Why this convergence is so valuable for drug discovery and what it does not reveal about the majority of diagnoses.

Two genes in eleven people

In January 2026, a paper in *Nature* described how eight rare genetic forms of autism converge on common developmental pathways in the laboratory. The headlines read ‘Hundreds of genes, one pattern’. The same paper contains a second figure that hardly anyone has written about. In the eleven people with idiopathic autism – that is, without any known risk variants – who were studied, the model identified two notable genes. This article explores the gap between these two findings and asks who stands to gain from this confusion.

Two transcripts

The study was conducted by Aaron Gordon, Se-Jin Yoon, Lucy Bicks and colleagues led by Sergiu Pasca and Daniel Geschwind. They used human cells (induced pluripotent stem cells) and measured their gene activity on days 25, 50, 75 and 100 (in cortical organoids). Following quality control, 70 cell lines from 55 individuals remained.

Eleven of these individuals had a diagnosis of autism without any of the rare variants under investigation. For this group, the study identified two genes with significantly different regulation across the entire experimental design: PRRC2C on day 50 and the long non-coding RNA RP11-114H21.2 on day 100.

Two! The authors attribute this to polygenic inheritance in cases of autism without a major risk mutation and note that far larger sample sizes would be required.

This group represents the norm. In most people with a diagnosis of autism, no single rare variant can be identified. What the study reveals therefore applies to a small, precisely defined part of the spectrum.

Where convergence holds

For the other cell lines, the results are different. Eight mutations associated with autism were investigated: the 22q11.2 deletion, the 22q13.3 deletion (Phelan–McDermid syndrome), the 15q13.3 deletion, the 16p11.2 deletion, the 16p11.2 duplication, Timothy syndrome, the PCDH19-associated disorder and the SHANK3 point mutation R522W. It would be inaccurate to speak of eight syndromes, as only two of these are named syndromes.

Early in development, these eight forms appeared markedly different. As maturation progressed, their changes converged on common processes: gene regulation, chromatin dynamics, the development of neural progenitor cells, and the maturation of synapses.

This is a narrowly defined and interesting finding. It becomes incorrect as soon as its scope is disregarded.

What the classification has to say about this

To understand this classification, it is worth looking at how official systems categorise autism today. Kanner’s and Asperger’s syndromes, as well as atypical autism, are no longer separate diagnoses since the DSM-5 and the ICD-11. They have been replaced by a classification based on intellectual development, language and support needs. There is no classification according to ‘types of autism’ in these systems.

The new typology emerging from laboratory research is therefore not a return to the old classifications. It draws a different line, and the ICD-11 already has a category for this line. Code 6A02 covers primary, idiopathic forms. Autism resulting from another genetic disorder or damage is separated from this and classified as a secondary neurodevelopmental syndrome under code 6E60. The eight constellations identified in the *Nature* study are all named genetic constellations and, in essence, belong to this second category.

(This allows us to describe the result precisely: the increasing similarity – convergence – applies to the genetically determined forms.) On the other hand, in the case of the idiopathic forms, only two genes have been identified.)

Who benefits from convergence?

A second study shows why this detailed finding nevertheless attracted considerable attention. At the end of August 2026, a group led by Nevan Krogan and Matthew State at the University of California, San Francisco, published a map of ‘protein functions in autism’ in *Science*: 100 high-ranking risk genes, over 1,800 so-called protein interactions – 87 per cent of which had not previously been described – and 54 variants derived from patient samples. The study explicitly focuses on profound autism, i.e. autism with severe intellectual disability and absent or minimal spoken language.

State describes the logic bluntly: if several mutations converge on the same process, this is a strong indication that the process in question can be treated with medication. This is where the value of convergence lies. Without it, a separate active substance would have to be sought for every rare variant. With it, a common target becomes conceivable and, with it, a medicine that could be suitable for more than just a handful of people.

This brings economic interests into play. In August 2026, Krogan’s institute received 46 million dollars from the Aligning Research to Impact Autism initiative, funded by Google co-founder Sergey Brin. And Daniel Geschwind, the last author of the aforementioned Nature paper, describes the new map as “an unprecedented resource for the field”, which will be used by many with an interest in autism and drug development.

There is nothing disreputable about this. Research requires hypotheses about where interventions might be possible, and funders want to know what they are investing in. What is noteworthy is where the line is drawn: at the point where something can be identified and potentially treated. That is why the headline reads ‘Hundreds of genes, one pattern’ and not, more accurately, ‘We find nothing in most cases’.

What the models are – and what they are not

‘Cortical organoids’ replicate selected processes of early brain development. They are not brains, nor are they a model of a person’s life. Gene activity was measured in cell cultures at four time points, across 55 individuals.

Such measurements reveal nothing about communication, perception, masking, overload, meltdown or autistic burnout. Nor do they allow for a ranking of autistic ways of life: similar changes in a cell culture may be associated with very different developmental and life trajectories.

What this implies for a diagnosis

An autism diagnosis is made clinically, based on developmental history, sensory processing, communication, interests, challenges and support needs. The detection of a protein in a cell culture neither proves nor rules out such a diagnosis.

Anyone who does not identify with any of the eight genetically defined forms belongs to the group in which the cited study found two gene products. This is a statement about the scope of a cell model in eleven people. It says nothing about you, the reader.

This does not resolve the question of causation. It becomes more precise: for which genetic patterns can a common developmental process be demonstrated in which model? Which remain distinct? And who is left out of the narrative when a narrow set of test results is turned into a general theory of autism?

Key points in brief

•            A study published in *Nature* in January 2026 examined human cortical organoids from 70 cell lines taken from 55 individuals, at four time points up to day 100.

•            In eight genetically defined forms, initial variations in the course of the model’s development converged towards common processes of gene regulation and neuronal maturation.

•            In the same study, two notable molecules were identified in eleven individuals with idiopathic autism: PRRC2C and the long non-coding RNA RP11-114H21.2. The authors attribute this to polygenic inheritance and the need for much larger sample sizes.

•            Official classifications no longer recognise distinct types of autism. However, the ICD-11 distinguishes between primary, idiopathic forms (6A02) and secondary, syndromic forms (6E60). The convergence applies to the syndromic category.

•            A study published in *Science* in August 2026 mapped over 1,800 protein interactions for 100 risk genes, 87 per cent of which had not previously been described. The focus is on profound autism.

•            Convergence is the prerequisite for a common therapeutic target to become conceivable. This explains why it is being researched and funded, and does not imply any conclusions about the majority of people with autism.

Sources

·        Gordon, A., Yoon, S.-J., Bicks, L. K. et al.: Developmental convergence and divergence in human stem cell models of autism, Nature 651 (2026), 707–719: https://www.nature.com/articles/s41586-025-10047-5

·        Wang, B., Vartak, R., Hennick, K. et al.: Autism mutations rewire protein interaction networks to drive neurodevelopmental pathology, Science 393 (2026), eady4523, DOI 10.1126/science.ady4523: https://www.science.org/doi/10.1126/science.ady4523

·        Huang, P.: Scientists map key protein interactions linked with profound autism, NPR, 4 September 2026: https://www.npr.org/2026/09/04/nx-s1-5954823/profound-autism-proteins-genes-map-treatments

·        World Health Organisation, ICD-11 for Mortality and Morbidity Statistics, Version 2026-01, 6A02 Autism spectrum disorder and 6E60 Secondary neurodevelopmental syndrome: https://icd.who.int/browse11/l-m/en

·        American Psychiatric Association, Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition, Text Revision (DSM-5-TR), 2022, Autism Spectrum Disorder

Related

• ADHD and autism: Diagnosis within the spectrum of neurodiversity

• Autism and ADHD: Overload, meltdown, shutdown, panic attack – what’s what with AUDHS?

• Autism and the double empathy problem: Autistic people are not mind-blind or unempathetic

• Autism in adulthood

• Neurodiversity Lite: How Neurodiversity Became the Dominant Discourse