A drug molecule invented entirely by artificial intelligence marks a critical milestone for the role of machine learning in medicine.
In a sharp acceleration of the typical path to drug development, which can take about four and a half years, the AI-designed compound reached the point of entering clinical trials within just 12 months. Exscientia’s AI platform used a suite of algorithms to decide on the best chemical structure for the new compound, which is known as DSP-1181 and is targeted at a specific receptor in the brain involved in OCD. Together the algorithms were able to generate tens of millions of potential molecules, sift through the candidates and make a decision about which ones to synthesise and test.
“The AI can learn faster than conventional approaches, so we had to make and test only 350 compounds, a fifth of the normal number of compound candidates, which is record-breaking productivity,” said Andrew Hopkins, chief executive of the start-up and a molecular biophysicist. “The algorithms . . . can be applied to any drug targets, against a huge range of diseases in oncology, cardiovascular and rare diseases.” Toru Kimura, senior executive research director of Sumitomo Dainippon Pharma, added: “We are very excited with the results of the joint research . . . We will continue to work hard to make this clinical study a success so that it may deliver new benefits to patients as soon as possible.”
