Realities and Limitations of Coverage in Current "Whole''-Exome Sequencing Capture Approaches

New technologies such as ‘‘whole’’-exome sequencingcapture methods have led to renewed excitement aboutdiscovering more rare, Mendelian, high-risk susceptibilitygene variants in humans. To assess ‘‘whole’’-exomesequence capture approaches with respect to coverage,we evaluated the content and the empirical performance ofthe currently available ‘‘whole’’-exome sequence capturemethods [NimbleGen Sequence Capture 2.1 M HumanExome Array; Agilent SureSelect Human All Exon Kit] onthree sequencing platforms [454 FLX Ti (4 runs); ABISOLiD (1 quadrant); Illumina GA II (2 lanes)]. The proteincoding sequences (CDS) reported in the RefSeq database(build 36.3) served as the gold standard. NimbleGencapture probes currently target 77% of CDS bases andAgilent capture probes target 83% of CDS bases. Weobserved 21.5 Mbps (65%), 25.0 Mbps (76%) and 23.4 Mbps(71%) of the 33.0 Mbps of CDS with ?8x sequence depth forNimblegen/454, Agilent/SOLiD, and Agilent/Illumina,respectively. Since identification of rare gene variantsrequires high per-gene coverage, we also computed theproportion of genes with 490% of CDS bases covered with?8X sequence depth. Only 42%, 55%, and 45% of geneswere covered by Nimblegen/454, Agilent/SOLiD, andAgilent/Illumina, respectively. Coverage of entire geneCDS is presently incomplete, thus negative results must beinterpreted cautiously since it is not currently possible tofully exclude most genes from consideration of harboringcausal mutations.