Back

Rule-out test for autism using machine-learning analysis of molecular temporal dynamics in hair - a multicenter study

Midya, V.; Bello, G. A.; Gomez, L. A.; Marin, M. R.; Piyankarage, S. C.; Elhlou, S.; Chumber, J.; Jaramilo, J.; Dessalle, S.; Yitshak Sade, M.; Cantoral, A.; Wright, R. J.; Wright, R.; Nakayama, S.; Bennett, D. H.; Schmidt, R. J.; Bolte, S.; Arora, M.

2025-11-20 health informatics
10.1101/2025.11.19.25340581 medRxiv
Show abstract

Absence of autism risk-stratification tools under 18 months hampers early intervention. In a multinational sample of 1697 participants, aged one month and older, we provide proof-of-concept that temporal molecular dynamics can stratify autism likelihood. Using laser-ablation-inductively-coupled-plasma-mass-spectrometry, we measured elemental intensities along growth increments of single hair strands at [~]800 timepoints. We developed a first-stage model to stratify individuals into a lower autism probability group and applied a second-stage model to the remaining participants, stratifying them into intermediate- and high-probability groups. Models were trained, ensembled, and tuned on participants from California and Sweden, then tested on 580 participants (within- and external-population replication in New York, Mexico, and Japan). Likelihood ratios (95%CI) for autism in low-, intermediate-, and high-probability groups were 0.18(0.15-0.23), 1.09(0.99-1.20), and 2.62(1.55-4.00), respectively. Low-probability classification (first-stage) had sensitivity of 96%(0.91-0.98), and high-probability classification (second-stage) had specificity of 90%(0.86-0.92). Our data support that elemental biodynamics can objectively stratify autism likelihood.

Matching journals

The top 3 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.