Language-network gene sets (v0.4): a pre-specified, literature-curated pathway resource for autism verbal-heterogeneity analysis

A hand-curated, literature-provenanced collection of 30 gene sets ("pathways") comprising 185 unique genes spanning the biology of the cortical language network (superior temporal gyrus and connected regions), assembled as a phenotype-blind feature space for pathway-level analysis of autism transcriptomes. Each gene was selected from independent published language-network biology — not tuned on any sample's verbal-ability phenotype — and annotated with its literature citation, evidence strength, curation axis, and a language-specificity score. Freeze provenance. This gene set was frozen prior to phenotype analysis. It is referenced as the pre-specified feature space in OSF pre-registration 10.17605/OSF.IO/3BC86 (registered 2026-05-25; under embargo), and version-controlled (repository commit 094bade, 2026-06-16); the artifact is reproduced here byte-for-byte. Scope. This is a curated feature-space resource, not a validated result. It encodes prior published biology so downstream pathway-subtyping analyses operate on an independently-specified, phenotype-blind gene space; whether any set recovers a real, replicable signal in a given cohort is an empirical question for the analysis that uses it. Curation axes: cortical-layer programs, region-specific enrichment, cell-type (PV/SST/VIP interneurons, OPC, oligodendrocyte), critical-period plasticity, language-disorder GWAS, auditory-language relay, and speech-motor/Broca. See README for the full file and axis manifest.