Identifying variants

Run metrics are really important. And what I mean by run metrics is, for instance, how well is a particular variant covered, and what are the ratios if it’s a heterozygous case? Is it a balanced ratio, which is what we would be hoping for– so in other words, 50-50 of a reference to your alternative value. But of course, when you have a presentation like in this case, which is supposed to be an autosomal recessive, kind of characteristics looking for homozygous event or compound heterozygous event happening.

Once we have managed to collate all that data in relation to the variant, we then look at published evidence to suggest is there any prior evidence to suggest it’s been seen previously? And we look at population frequencies. So we resource a lot for information that’s been done, for instance, in the ExAC projects, or looking at the one key projects, or looking at dbSNP as well. But more important, that we found to be a very powerful tool as to look at our in-house data, to filter out information that might be just background polymorphisms.

When the evidence comes to light in regards to a gene being relevant to a particular phenotype, as clinical bioinformaticians, we’re not automatically allowed to go back and screen for any variants that might have occurred in this new gene. The reason for that is that there might be many decisions that patients might not want to know. And of course, that assessment needs to be done by the clinician or a genetic counsellor at that stage. And of course, and when you’re in the research sector, quite often we switch projects. New evidence is always informative for that kind of recursive or rapid recursive reanalysis of data to make new gene discoveries.

And of course, in diagnostics we’re not by the business of gene discovery, but the business of actually providing better patient care.