Harnessing AI to Help Discover New Medical Treatments

By November 12, 2014
Keywords : e-Health, AI, America, Asia, EMEA
AI pour les médicaments

Medical research can be a very lengthy process. Now researchers are drawing on artificial intelligence (AI) techniques to speed up drug development and open up new uses for known molecules.

According to Canadian firm Chematria, it currently takes around fifteen years and close to €2 billion to develop a new therapeutic drug and bring it to market. The Toronto-based startup has set out to streamline all this.  To do so, the four founders are harnessing artificial intelligence techniques. Working in association with the University of Toronto, they have begun to develop computer projections capable of simulating the kind of experiments than are usually carried out in a laboratory. The tool they have developed is able to analyse millions of data points that have been gathered over past years on molecular reactions in order to predict future reactions and combinations.

Speeding up the discovery process

Other researchers have applied similar methods in the past, but what makes Chematria really stand out from the crowd is the tool the startup is using.  The team have obtained access to the most powerful classical supercomputer in Canada, a 32,767-core IBM BlueGene/Q, which means they can achieve results 150 times faster than would be possible using the mainstream approach. Saving time also leads to a drastic cut in costs, which can multiply their options. The series of tests they carry out no longer take place in test-tubes but inside the computer. The first step is to test the therapeutic effect on a given medical condition of molecules whose properties are already known. Then statistical models are applied to the molecules’ binding activity in order to predict how strongly they will bind to a biological target.  Following this approach, the Chematria programme is likely to find new applications for therapeutic substances that already exist.

Focus on Ebola

Chematria’s technology is driven by a ‘virtual brain’ that studies millions of datapoints on how drugs designed to treat a particular disease or medical condition have worked in the past. Armed with this information, the software can apply the observed patterns to predict the potential effectiveness of hypothetical drugs and come up with new uses for existing drugs. Very recently Chematria CEO Dr Abraham Heifets announced that the company intended to focus near-term research on finding new treatments for the Ebola virus. The speed of response which this current medical emergency demands is precisely Chematria’s strong point. AI has already been used in a more general way in the health sector, to improve patient data management.  In fact AI-related techniques have a wide range of potential applications but medical research would appear to be one of the most promising.

Legal mentions © L’Atelier BNP Paribas