the way they watched the process was to sequence the whole genome of each bacterial isolate. What they found were a total of 35 mutations, which developed sequentially as the treatment continued (and the levels of resistance rose). Here’s natural selection, operating in real time, under the strongest magnifying glass available. And it’s in the service of a potentially serious problem, since resistant bacteria are no joke. (Reading between the lines of the PNAS abstract, for example, it appears that the patient involved in this study may well not have survived).

The technology involved here is worth thinking about. Even now, this was a rather costly experiment as these things go, and it’s worth a paper in a good journal. But a few years ago, needless to say, it would have been a borderline-insane idea, and a few years before that it would have been flatly impossible. A few years from now it’ll be routine, and a few years after that it probably won’t be done at all, having been superseded by something more elegant that no one’s come up with yet. But for now, we’re entering the age where wildly sequence-intensive experiments, many of which no one even bothered to think about before, will start to run.