High-throughput sequence analysis was performed according to a customized bioinformatic pipeline for tracking sequence data, aligning reads, calculating coverage, calling variants, annotating variants with respect to functional effect, filtering out benign variation and flagging candidate rare, pathogenic mutations. Briefly, BWA version 0.5.7 (ref. 3) was employed to align reads to the human genome (reference build hg18). Consensus and variant base calls were made with SAMtools version 0.1.7 (pileup), filtered for quality (mapping quality >10 for insertions and deletions, and >25 for SNPs), and loaded into a MySQL database for storage and further processing, including annotation of the predicted consequences (noncoding, coding synonymous, coding nonsynonymous or frameshift, splice site) of each variant using GMCC [43] (Genomic mutation consequence calculator). Candidate mutations were identified by starting with a list of all variants, removing those present either in dbSNP130 or the 1000 Genomes Project database, and selecting for coding nonsynonymous, frameshift or splice site changes. Sequence data were visualized using either the UCSC Genome Browser or the Broad Institute Integrated Genome Viewer. All genomic base positions are presented in reference to the human genome NCBI build 36 (hg18). The functional effect of the mutation on the protein was assessed using PolyPhen-2 [44].
