.

.

Monday, November 25, 2013

13 % is Not Good

Sometimes it is better to just bang one's head against the wall.
Years ago, when L was hired, we knew she didn't know how to read DNA sequence, so we spent considerable time showing her how. At the time, though, we weren't very good at documenting all the different training sessions and tutorials and certifications.
Now we are.
Since forever, however, all our sequences have been read, independently, by two different technicians, and every disagreement is judged by a third. So if somebody misses something, there's a backup, and we really don't miss very much.
In the early summer we found some problems in some archived data, and we decided to give all the techs a test of 100 sequences to read, a new training session, with a quiz afterward, and then another test of 100 sequences. So today I should be able to count on all of the techs to read sequence. Even the newest tech has been on the job for at least 3 years.
There will always be things skipped over. Nobody's perfect, and even I have passed over variants that were right in front of my eyes. So I understand the problem. An error every 500 sequences seems reasonable.
So.
There's a series of samples just finishing up in the lab, and I took the page for the 2nd read myself just to get things going faster. Families are waiting for these results.
70 sequences, and the tech had either not- or mis-noted eight neutral variants.
Now, neutral variants have no effect on the results (they are, after all, neutral). But they are an indication of whether or not the tech is paying attention to the sequence she's reading. Any base change could be important. It's obligatory to note them all.
I made a tic mark next to each discordant result and put the tech's page back in her folder for her to finish her analysis (there were 7 sequences she hadn't gotten to yet). I haven't had the chance to talk to her yet, because some days she's not here, others I'm in meetings all day. So tonight I went to see her progress, and yes, she had correctly noted all the variants in the sequences I had pointed out. So she was aware that she had made a bucket-load of mistakes. And she was aware that it wasn't one of her buddies that found them, but the boss. So you'd think she'd take special care to finish those last 7 sequences.
Two additional errors.
Head meets concrete.

No comments: