Robot overlords

Researchers at Carnegie Mellon University are training computers to decide which preclinical experiments will result in the best predictions for how compounds interact with proteins, cells and genes. Having shown the algorithms can improve efficiency for a high volume of experiments, the team has spun out Quantitative Medicine LLC, and is actively seeking partners for licensing.

The idea is to use a process of "active machine learning" instead of passive in silico- or in vitro-based screening, which are often constrained by the range of possible experiments that can be included...