The headline is sensationalized. They did a virtual screen of a large compound library. QSAR / QSPR / Virtual Screening has been around since the late 1950s. The secret sauce here was the large experimental dataset - on the order of 10^5 compounds - they generated to train the model. Maybe the explainability part is kind of novel for a neural network based approach; but not clear that you couldn’t identify substructu…
"The insight here was that we could see what was being learned by the models to make their predictions that certain molecules would make for good antibiotics," James Collins, professor of Medical Engineering and Science at the Massachusetts Institute of Technology (MIT) and one of the study’s authors, said in a statement.
...
"What we set out to do in this study was to open the black box. These models consist of very large numbers of calculations that mimic neural connections, and no one really knows what's going on underneath the hood," said Felix Wong, a postdoc at MIT and Harvard and one of the study’s lead authors.
> not clear that you couldn’t identify substructure classes better with a pure bioinformatic approach.
Lots of things aren't clear, but what supports your particular claim? Has it ever been done? And is it addressed in the article?