The authors summed selection coefficients across genes to produce per-individual, constraint-weighted genetic loads (GLs). They computed these separately for rare protein-truncating variants (PTVs) and for damaging missense variants. The analysis used European-ancestry participants from the Adolescent Brain Cognitive Development Study (ABCD; 3,523 children) and the All of Us Research Program (AoU; 302,421 adults), and the report specified that ABCD Data Release 7.0 and the All of Us Controlled Tier Dataset v9 were the data sources. The study team did not report exhaustive lists of specific variants in the source text; those details were not provided.
In the ABCD cohort, investigators observed that each additional constrained gene harboring a PTV was linked to a 0.22 standard-deviation lower cognition score. The cohort showed an association between that same PTV GL increment and ASD (odds ratio 1.80), the authors reported. The effect on cognition was larger for PTVs than for damaging missense GLs in the same analysis, according to the source. The study did not report drug interventions, dosing, or clinical management details, because none were relevant or included in the manuscript.
In the much larger AoU sample, one additional constrained-gene PTV GL increment associated with raised odds of ASD (OR 1.34), schizophrenia (OR 1.51) and intellectual disability (ID; OR 2.02), investigators reported. The authors noted that carriers of a constrained PTV comprised 12% of AoU participants but accounted for disproportionate fractions of diagnosed cases: 31% of ID, 20% of schizophrenia and 17% of ASD cases. The report did not provide clinical diagnostic algorithms or case-ascertainment protocols in detail within the source text; those procedural specifics were not included in the preprint.
GLs derived from damaging missense variants showed smaller but concordant associations with cognition and diagnoses compared with PTV-derived GLs. The cohort data indicated that risk rose with increasing gene constraint; in the most constrained genes the investigators reported nearly a ninefold odds of ID. The authors framed the findings as a gradient: more constrained genes carried larger per-unit GL effects. The source did not publish the precise thresholds or full ranked gene lists in the public abstract; those elements were not provided in the text available on the medRxiv page.
The study team compared effects across ancestry groups and with standard polygenic scores. PTV GL retained between 71% and 125% of its European-ancestry effect size in African-ancestry participants, while matched polygenic scores retained only 23% to 62% of their European-ancestry effects, the authors reported. The investigators also observed that the GL metric was essentially uncorrelated with polygenic scores, indicating largely independent information content between constraint-weighted rare variant load and common-variant polygenic risk measures.
The authors declared employment and equity ties to Herasight, Inc., and one coauthor is an advisor to the company; funding was reported from Herasight, Inc. Ethical approvals were reported: the University of California, San Diego IRB (central IRB for ABCD) and the All of Us Institutional Review Board. The analysis used de-identified ABCD data under an NIH Brain Development Cohorts Data Use Certification and controlled-tier All of Us data under the program's data-use agreement. Study-level summary statistics are available from the authors on reasonable request, with access subject to ABCD and All of Us data-use and dissemination requirements. The source did not list explicit limitations or offer a separate limitations section in the medRxiv abstract; the preprint likewise did not specify variant-level catalogs, detailed case-ascertainment algorithms, or full analytic code in the text available on the medRxiv page.
The analysis reported here quantifies a measurable contribution of constraint-weighted rare exonic variation to cognitive and neuropsychiatric outcomes in population cohorts, using ABCD and All of Us datasets. The manuscript ends with availability statements rather than clinical recommendations; readers seeking variant-level results or analytic code should note that the source stated study-level summary statistics are obtainable from the authors, subject to the cohorts' data-use policies.