Largest ever "map" of autism may hold clues for new targeted therapies (CNN, 2026-08-27)
Jacqueline Howard, CNN Health, on a Science paper (doi 10.1126/science.ady4523) from
nevan-krogan's group at UCSF's Quantitative Biosciences Institute and the Gladstone Institutes.
Parked in the hub _inbox on 2026-08-28 (CNN geo-blocks the fetcher; read via firecrawl then) and
moved here at spin-out. The Science primary is still unread.
The finding
Instead of studying autism's 250+ risk genes one at a time, the team measured how the mutations rewire protein–protein interactions in the developing brain: affinity-purification mass spectrometry run in the presence of the mutations, yielding more than 1,800 interactions, 87% of them previously unseen. AlphaFold ranked which mutations to follow up; the shortlist was tested in lab-grown brain organoids. Krogan's framing — the gene list is "a parts list", the interaction map "the wiring diagram" — and three therapeutic programmes are said to be running off it.
Caveats in the piece
Most directly relevant to profound autism (~30% of people with ASD); Autism Speaks' Andy Shih warns the path to therapy is long. The article ends on a US research-policy dispute: Alison Singer (Autism Science Foundation) says the draft IACC federal plan "deemphasizes genetics, neurodevelopment, and brain biology in favor of a broad injury model."
Why it is here and not in psychology-wiki
The object of study is a proteome, not a mind: no behaviour is measured, no person assessed. The park record recorded that boundary on 2026-08-28 and it holds. AlphaFold appears as an instrument, not as the subject — the biological-foundation-model question is whether such instruments become the substrate, which is what israel-ai-bio-consortium is funding.
Tier T2 — named-reporter piece at a major outlet on a Science paper, with the primary unread.
Related
nevan-krogan · biological-foundation-model · hippocampus-aging-epigenome · synthesis