Nobody wants to publish negative results. It's a familiar story to anyone who (like me) spent years as an experimental scientist, tearing their hair out in frustration as yet another reaction failed to work for inexplicable reasons, leaving that world-beating publication just as far away as ever.
The result is that lots of valuable scientific knowledge remains in lab books and never sees the light of day. "Unpublishable" results could be hiding a mass of information that would open up whole new fields of research.
Of course, nobody wants to trawl through old lab books looking back at past failures. But this new machine-learning approach does that, and it's already beaten humans at finding new and improved methods for crystallisation based only on the information hiding in old, handwritten notes.
From an IP perspective, this could offer an exciting new way to unearth all sorts of "hidden" inventions - new synthetic routes, improved variations on old reactions, or other surprising things buried in "useless" old data - and, what's more, this potential IP goldmine doesn't require any more research effort than you've already put in to find the "flagship" results.
Who knows, the key clue to the next Nobel-winning, multi-billion-dollar, patentable drug might be languishing in an old and unloved lab notebook somewhere. Time to dust off your old PhD notes and give your patent attorney a call?
“Failed reactions contain a vast amount of unreported and unextracted information,” says Alex Norquist, a materials-synthesis researcher at Haverford College in Pennsylvania, who is part of the team that has reported the work in Nature. “There are far more failures than successes, but only the successes generally get published.”