Twenty Mouse Models
Twenty mouse models. Each one carries a different genetic mutation linked to autism in humans. Researchers at the Italian Institute of Technology scanned the brains of all twenty, then mapped the connectivity patterns onto brain scans from 940 autistic individuals, according to a study published in Nature Neuroscience. The comparison group: more than 1,000 neurotypical brain scans.
The mouse models were not decorative. They were the reverse-engineering tool. Each mutation produces a different biological change in the brain. Scan them all, catalog the connectivity signatures, then look for those same signatures in human data. If the patterns match, you have traced the biology to the imaging.
Two Clusters, Not a Spectrum
The human scans split into two groups. One subtype showed reduced communication between brain regions. The second showed unusually strong communication between different brain areas. Not scattered across a continuum. Two distinct biological patterns, according to the study.
The reduced-connectivity group traced to synaptic pathways, the signal-sending machinery between brain cells. The increased-connectivity group traced to immune-related biological systems. Two different mechanisms, two different connectivity signatures.
The Replication Test
The same two-subtype pattern appeared across multiple independent datasets from research centers worldwide. Not an artifact of one lab's imaging protocol or one population's genetics. The division held.
Gene expression analyses confirmed the split. Regions showing lower connectivity expressed synaptic genes. The biology matched the imaging, according to the study.
The Denominator Problem
940 autistic brains divided by two subtypes. Every clinical trial testing autism drugs to date has been dosing a mixed population. One subtype has a synaptic problem. The other has an immune problem. A drug targeting synaptic function will do nothing for the immune group, and vice versa.
The failure rate of autism drug trials suddenly has a mathematical explanation. You cannot average two different biological conditions and expect one treatment to work. The denominator was wrong.
What the Mouse Models Bought
The mouse-to-human mapping is the mechanism that made this possible. Without the twenty models, each representing a known genetic mutation, researchers would have had 940 human scans and no way to trace connectivity patterns back to specific biological pathways. The mice were the Rosetta Stone.
The study did not identify which individual falls into which subtype based on behavior or diagnosis. It identified the subtypes based on brain connectivity, then traced those patterns to biological pathways. The division is not clinical. It is structural.
Two Pathways
Two pathways. Two subtypes. 940 divided by 2.
The next clinical trial will need to divide its participants before the first dose is administered.
Precision medicine for autism is no longer theoretical. It has a connectivity map, a biological mechanism, and a denominator that finally makes sense.
