Pharma groups might screen millions of compounds before deciding on contenders to test in preclinical trials, and most companies have brought various software tools into the mix to automate the discovery process. Still, drug discovery sucks up a lot of time and money, plus compounds entering clinical trials have an abysmal 1-in-10 shot at success. Predictive modeling based on chemical structure, disease/target and other data offer a faster way to home in on drug candidates.
"Our view has been that, for every single project, design huge virtual libraries of 100 [million or more] compounds and then use these predictive models to search those libraries for the compounds that are going to have the highest likelihood of success against all of our criteria," Guido Lanza, CEO of Numerate, told FierceBiotechIT.
Numerate, which has revealed tie-ups with drugmakers Boehringer Ingelheim and Merck ($MRK), builds virtual assays or predictive models that are designed with the drug target and treatment goals in mind. Then the company unleashes the models to analyze massive virtual libraries of compounds that amount to terabytes and potentially petabytes of data. It spent months to accomplish in an HIV drug project what took a similar effort at a Big Pharma company years, which was to test and synthesize anti-HIV compounds, Lanza says.