The postdoc and I are still/again working on his paper about DNA uptake specificity, and were getting deep into the Results. I had thought I understood the specificity matrices and sequence logos, but tonight I realized that I didn't even understand the simple stuff (the ratios of different categories of sequences in the sequence reads from the input pool and the recovered pool.
While struggling with that, I realized that we had been assuming that all the DNA fragments in the recovered pool had indeed been taken up by the competent cells and then reisolated. We hadn't properly controlled for the possibility that the recovered pool was contaminated with DNA fragments that had never been taken up, but that had the same sequence distribution as the input pool.
I started thinking about contamination because of what appeared to be an odd result. The postdoc's results show that four positions in the 31 bp DNA uptake sequence (USS) are particularly important for uptake; fragments with a non-consensus base at even one of these positions were rarely taken up by competent cells. But he has sequences of 10^7 fragments from the recovered DNA pool, and this large dataset does include quite a few fragments with such mismatches.
Surprisingly, when a set of fragments with one of these mismatches is examined (e.g. fragments that do have a mismatched base (A, G or T rather than C) only at position 7, the most important for uptake), the bases at all of the other positions of the USS appear to have made no contribution to uptake. This observation might be explained by some weird type of interaction, but it might also be a sign that the recovered DNA preparation is contaminated with DNA fragments that were never taken up.
In fact, we can use this observation to set limits on contamination, and to correct the database for possible effects of contamination, by considering the implications of two extreme hypotheses about contamination. The first is the hypothesis that there is no contamination - that every fragment in the recovered pool is there because it was taken up by a competent cell, including all the fragments with a mismatched base at position 7. The second is the hypothesis that fragments with A, G or T at position 7 are never taken up, and that all of the fragments with a mismatched base at position 7 are there because of contamination, not uptake. We can use this hypothetical value to calculate an upper limit to the contamination of the recovered pool, and then apply appropriate correction factors to all the analyses.
It's a bit embarrassing to realize that we've been neglecting this obvious issue with our data. But I'm really glad that we're catching it now and not leaving it for the referees to catch after we submit our manuscript.
What math can teach us about drug discovery and biology (and all of science, really)
11 hours ago in The Curious Wavefunction