T19. DECOMPOSING PHENOTYPIC DISPERSION IN 16P11.2 COPY NUMBER VARIATION ACROSS TRAITS AND INDIVIDUALS

Abstract

Background Rare recurrent copy number variants (CNVs) are associated with risk for developmental and neuropsychiatric disorders but demonstrate considerable clinical heterogeneity at the individual level. It remains unclear how between-person variability within CNV groups compares to the general population, differs across traits, and on which outcomes carriers converge or diverge. In this study, we examine the profile of dispersion effects of 16p11.2 (BP4-BP5) CNVs. Methods Data from Simons Searchlight were analyzed using Gaussian location-scale mixed models to estimate deletion (n=127) and duplication (n=68) effects on the mean and dispersion of developmental, psychosocial, and neuropsychiatric traits, using neurotypical controls (n=62) as a common reference. Models adjusted for age and sex and included family_id as a random intercept. Trait-level dispersion estimates were then modeled as a function of smoothed mean effect size and variable skewness, and positive residuals were interpreted as evidence of excess phenotypic variability beyond expected scaling and distributional effects. In deletion carriers, we fit a weighted stochastic block model to a bipartite patient–symptom graph, with edges defined as symptom rating deviations from each individual’s average symptom severity, capturing relative strengths and weaknesses independent of global impairment. This co-clustering approach jointly identified patient and symptom blocks. We examined symptom profiles, demographic characteristics, and adaptive functioning for each patient block. Results After adjusting for mean effects and skewness, the strongest evidence for excess dispersion was seen for measures of thought disturbance, restricted and repetitive behaviors, social communication, and social cognition in both deletion and duplication carriers. Conversely, aggressive behavior and oppositional defiance were less variable than expected in duplication carriers. In deletion carriers, speech articulation and syntax structure were more homogeneous than expected. Across all traits, a small proportion of standardized phenotypic variance was explained by between-subject differences (ICC = 9.9% and 13.7% in deletion and duplication carriers, respectively), indicating substantial within-subject heterogeneity. To explore potential patterns in within-subject heterogeneity, bipartite weighted stochastic block modeling partitioned deletion carriers into three groups with distinct severity-centered profiles: group 1, internalizing/mood-enriched and language/social communication-spared; group 2, executive/externalizing-enriched and internalizing/mood-spared; and group 3, language/social communication-enriched and executive/externalizing-spared. Group 2 individuals were younger than the other groups and had the lowest mean daily living skills, although this difference did not remain significant after Tukey correction. Discussion We found domain-specific heterogeneity in 16p11.2 CNV carriers, with increased variability in core autism-related traits and copy number-dependent decreases in variability for externalizing and speech-related traits. Distinct subject-level clinical profiles may contribute to this group-level heterogeneity across trait domains. These findings highlight the importance of cross-disorder studies of dimensional traits and may guide investigations into mechanisms driving within-group heterogeneity, such as gene-gene or gene-environment interactions.