CheMeleon model adapts to predict molecular properties from scant data

Discovering new molecules for specific uses is a major driver of progress in a wide variety of industries, whether it's pharmaceutical companies looking for new drugs or electronics or energy companies seeking new materials with particular properties. But it's also largely a trial-and-error process, because there is not yet any way to perfectly predict how a given chemical compound will behave in terms of the myriad different properties and responses materials can exhibit.