Structural models of genome-wide covariance identify multiple common dimensions in autism.

Common genetic variation has been associated with multiple phenotypic features in Autism Spectrum Disorder (ASD). However, our knowledge of shared genetic factor structures contributing to this highly heterogeneous phenotypic spectrum is limited. Here, we developed and implemented a structural equation modelling framework to directly model genomic covariance across core and non-core ASD phenotypes, studying autistic individuals of European descent with a case-only design. We identified three independent genetic factors most strongly linked to language performance, behaviour and developmental motor delay, respectively, studying an autism community sample (N = 5331). The three-factorial structure was largely confirmed in independent ASD-simplex families (N = 1946), although we uncovered, in addition, simplex-specific genetic overlap between behaviour and language phenotypes. Multivariate models across cohorts revealed novel associations, including links between language and early mastering of self-feeding. Thus, the common genetic architecture in ASD is multi-dimensional with overarching genetic factors contributing, in combination with ascertainment-specific patterns, to phenotypic heterogeneity.

Detecting selection using extended haplotype homozygosity (EHH)-based statistics in unphased or unpolarized data.

Analysis of population genetic data often includes a search for genomic regions with signs of recent positive selection. One of such approaches involves the concept of extended haplotype homozygosity (EHH) and its associated statistics. These statistics typically require phased haplotypes, and some of them necessitate polarized variants. Here, we unify and extend previously proposed modifications to loosen these requirements. We compare the modified versions with the original ones by measuring the false discovery rate in simulated whole-genome scans and by quantifying the overlap of inferred candidate regions in empirical data. We find that phasing information is indispensable for accurate estimation of within-population statistics (for all but very large samples) and of cross-population statistics for small samples. Ancestry information, in contrast, is of lesser importance for both types of statistic. Our publicly available R package rehh incorporates the modified statistics presented here.

The neutral frequency spectrum of linked sites.

We introduce the conditional Site Frequency Spectrum (SFS) for a genomic region linked to a focal mutation of known frequency. An exact expression for its expected value is provided for the neutral model without recombination. Its relation with the expected SFS for two sites, 2-SFS, is discussed. These spectra derive from the coalescent approach of Fu (1995) for finite samples, which is reviewed. Remarkably simple expressions are obtained for the linked SFS of a large population, which are also solutions of the multi-allelic Kolmogorov equations. These formulae are the immediate extensions of the well known single site θ∕f neutral SFS. Besides the general interest in these spectra, they relate to relevant biological cases, such as structural variants and introgressions. As an application, a recipe to adapt Tajima's D and other SFS-based neutrality tests to a non-recombining region containing a neutral marker is presented.

rehh 2.0: a reimplementation of the R package rehh to detect positive selection from haplotype structure.

Identifying genomic regions with unusually high local haplotype homozygosity represents a powerful strategy to characterize candidate genes responding to natural or artificial positive selection. To that end, statistics measuring the extent of haplotype homozygosity within (e.g. EHH, iHS) and between (Rsb or XP-EHH) populations have been proposed in the literature. The rehh package for r was previously developed to facilitate genome-wide scans of selection, based on the analysis of long-range haplotypes. However, its performance was not sufficient to cope with the growing size of available data sets. Here, we propose a major upgrade of the rehh package, which includes an improved processing of the input files, a faster algorithm to enumerate haplotypes, as well as multithreading. As illustrated with the analysis of large human haplotype data sets, these improvements decrease the computation time by more than one order of magnitude. This new version of rehh will thus allow performing iHS-, Rsb- or XP-EHH-based scans on large data sets. The package rehh 2.0 is available from the CRAN repository (http://cran.r-project.org/web/packages/rehh/index.html) together with help files and a detailed manual.