CTA: You’ve expressed concerns about WES and its limitations compared to WGS, including diagnostic yield and bias.2 What are your concerns?
Dr Gerstein: Basically, the appeal of WES is it’s the cheaper version of WGS. I think almost anybody would say, if costs were equal, you’ll obviously get more from WGS. The issue is that the price of WGS is dropping, it’s really going down. I mean, we’re approaching the $1000 genome and [it] will be even less than that.
A few years ago, it was prohibitive to do WGS. It was more expensive and there was a strong argument for studying cancer using 5 WES for the cost of 1 WGS. But now the landscape is changing; it’s much cheaper to sequence the whole genome. That’s one aspect.
Another aspect is that even if you’re exclusively interested in the exome — studying a rare disease or cancer and want to focus on a specific genetic mutation in protein-coding genes, or to identify clinically-relevant mutations — even then, it is still the case that WGS is potentially better.
That’s because introducing the capturing reagent can bias [WES] reads. The recording instrument can slightly alter things. If you really want the best coverage and to be able to assess it most cleanly, then it’s better to sequence everything native; to get the whole-genome sequence. People have shown that the variant-identification quality of WGS is superior to WES.
What that means is that if you have a really important sample, and you’re interested only in particular exome region sequences, WGS is still better for identifying variants. Moreover, people are realizing more and more that there are mutations in regulatory regions outside of genes.
The final thing I’ll say is while there’s been a lot of focus on the canonical SNP variants, there are other genomic variants that are much more complicated. Structural variants, INDEL [insertion/deletion] variants. These variants are thought to have much stronger impacts than [SNPs] because a chunk of genome is removed.
One of the issues is that it’s very hard to assess structural variances based on exome sequencing because a structural variance itself might span the protein-coding region and part of a noncoding region, or half of a protein-coding gene and half of a noncoding region. If you’re only interrogating the protein-coding region, you’re not going to be able to characterize the structural variant completely.