New directions for machine learning

Applications such as spam filtering, economic forecasting, and Netflix recommendations are powered by algorithms that allow computers to learn from and make predictions based on vast collections of data. But even as machine learning has purged e-mail inboxes of Vi@gra, the technology has been slower to infiltrate chemistry.

As described in this collection of stories, however, the gap is starting to fill. Some researchers are pushing the boundaries of computer science and chemistry, applying the most sophisticated version of machine learning yet—deep learning—to drug discovery. A poster child of this effort is a nascent partnership between pharma firm GlaxoSmithKline and government labs that seeks to use screening data on a large number of compounds to find promising therapeutic candidates. Meanwhile, publishers and others haven’t given up on more traditional machine-learning algorithms: They’re using them to develop tools to capitalize on information catalogued in journals, patents, and elsewhere.

CORRECTION: This story was updated on Jan. 23, 2017, to correct the number of government labs with which GlaxoSmithKline is partnering. It is more than two.