Can autism be diagnosed with a brain scan? The lead author tells the Guardian that the test will be 90% accurate. No, says Carl Heneghan later in the Guardian – the actual chances that someone with a positive result would have autism is 4.5%, or around 1 in 22. The authors reply in the comments, and Heneghan replies back. Meanwhile, Dorothy Bishop discusses why it’s actually very difficult to set up a screening test.

Resistance to resistance is futile. Antibiotic-resistant bacteria carrying a gene called NDM-1 can shrug off all but two of our antibiotics, leading some scientists to warn about the end of the antibiotic era. Maryn McKenna has the best analysis at Superbug, while