Identification of rare variants from exome sequence in a large pedigree with autism

Abstract

We carried out analyses with the goal of identifying rare variants in exome sequence data that contribute to disease risk for a complex trait. We analyzed a large, 47-member, multigenerational pedigree with 11 cases of autism spectrum disorder, using genotypes from 3 technologies representing increasing resolution: a multiallelic linkage marker panel, a dense diallelic marker panel, and variants from exome sequencing. Genome-scan marker genotypes were available on most subjects, and exome sequence data was available on 5 subjects. We used genome-scan linkage analysis to identify and prioritize the chromosome 22 region of interest, and to select subjects for exome sequencing. Inheritance vectors (IVs) generated by Markov chain Monte Carlo analysis of multilocus marker data were the foundation of most analyses. Genotype imputation used IVs to determine which sequence variants reside on the haplotype that co-segregates with the autism diagnosis. Together with a rare-allele frequency filter, we identified only one rare variant on the risk haplotype, illustrating the potential of this approach to prioritize variants. The associated gene, MYH9, is biologically unlikely, and we speculate that for this complex trait, the key variants may lie outside the exome.

Figure 1

Schematic representation of the pedigree affected by autism. Symbols illustrate relationships between the six affected sibships, represented by filled diamonds in the bottom generations. Only unaffected subjects required to illustrate the relationships between the affected sibships are drawn, all others are excluded from the diagram.

Figure 2

Linkage analyses shown relative to physical map positions for chromosome 22 for Stage 1 STR markers (solid line), Stage 2 STR + linkage SNPs (dashed line), and Stage 3 STR + phasing SNPs (dash-dot line). Crosses show locations of each marker. Vertical dotted lines direct the reader towards boundaries of the risk haplotype shared among the 8 affected subjects inferred to carry the risk variant. Bars underneath the lod score curves represent chromosomal regions sharing the same founder genome label (FGL) as D22S683 (solid bars), not sharing this FGL (white bars), and ambiguous regarding FGL sharing (hatched bars). Stage 2 results are in black; Stage 3 results are in grey. The haplotype sharing patterns among the 8 subjects are identified by unique labels after the subject’s corresponding sibship label from Figure 1. Marker names are in Supplemental Table 1.

Figure 3

Number of imputed exome variant alleles under the default threshold for (A) all 147 variant sites, and (B) for the variants with rare minor allele frequencies.

Figure 4

Probability of calling an exome variant allele as a function of minor allele frequency for the 147 variant sites base on (A) the default threshold, and (B) deterministic threshold.

Figure 5

Exome variant call rates in the 42 subjects with no direct exome genotyping when exome data for only the four subjects at the bottom of the pedigree are used vs. when exome data for all five subjects are used, for (A) the default threshold, and (B) deterministic threshold.