Friday, April 24, 2015

Superpower: variant detection

Warning! Highly technical post today!
This week is pretty calm at the lab, with nearly half our staff on vacation. That means 4 of the 5 diagnostic techs. Since we're pretty far behind in signing out cases, I've been picking up some of the slack in reading sequences. All our sequences are read twice, to maximize the chances of actually seeing any variants that are there, and getting them onto the summary table with the right name and all.
You'd think with all the sequence alignment software that's been out there for so very, very long, that you'd barely need one person to read all the sequence, let alone two. Differences should just pop out. Gotta be blind not to notice one bold red base in a whole field of grey.
Weeeeellll, it's not always so easy as that.
This particular case we signed out last year as having no mutation. But it bugged me, because the patient's tumor had all the signs of mutation in this one gene. Not everyone with all the signs has an inherited mutation, but most of them do. All the analytical steps checked out, though. No mistakes in testing the right sample or anything.
These days with every new analytic series, we include a rerun of something we've run before. So in the current series I reran this sample.

Because 80% of my techs are out on vacation, there wasn't anyone to do the second read of the sequences this week, so I've been doing some of that, both to stay in shape and to move things along. But what's this? Oh, look, a little secondary signal when there should only be one.
The software missed it, because it's just a little bump, far below the 35 % threshold to be flagged. The technician missed it. Really, the second signal accounts for less than 20 % of the signal at that position. Maybe just 10 %. Sometimes we have little peaks at around 10 % in all the samples, though only on one of the two strands of the DNA. Just artifacts when they're systematically present like that. Certain exons are notorious for this sort of thing - you can't get rid of it, and it means nothing.
This one is different. The second peak is only in this one sample out of the 47 in the series. It's present on both strands. The rest of the sequence is nice & clean. And the kicker is that the normal signal is diminished (which doesn't happen with artifacts). The two signals added together equal the intensity of the normal signal in all the other samples.
Hmmm. Normally, inherited mutations are 50-50, because you have one mutated and one normal copy. The patient's family history clearly indicates an inherited mutation, not one that is new in her and doesn't affect all of her cells. No reason for me to look for something 85-15.
So I have my computing guy dig up the archive of this bit of sequence for the first time we ran the sample. This time you really have to look for it, but it's there. No way anyone would have caught that.

There can be technical reasons why a mutation might not show up 50-50. In special genetics cases it might just not be 50-50, like when you have a chimera (a patient whose cells do not all come from a single fertilized egg, but who is a mix of two, only one of which carried the mutation in question) or a mosaic (where the mutation was not present in the egg or sperm, but happened in the early rounds of cell division so that only a subset of the cells in the body have it). Then there's the technical case where there is a second sequence variant farther on which just happens to sit within the sequence of the primers we used to amplify the bit in question - and the amplification of the mutant copy is much less efficient than for the normal copy.
We'll be looking into all that. A chimera will have this lopsided ratio at any sequence variant in the gene, and there are plenty of neutral variants to look at. A mosaic we should be able to rule out, based on family history. And the technical case we can resolve by designing new primers farther away and capturing that second variant. Oh, and the bugaboo of DNA labs, the sample itself is contaminated with a second person's DNA. We'll be looking at an independent sample from the same person, just to be sure.

No comments: