It takes approximately 1.3 Bi USD and a median 8.3 years to bring new drugs to market because approximately 97% of all drug candidates will fail during drug discovery and development. Importantly, approximately 90% of the time, cost, and failures occur during clinical trials for which approximately 60% fail due to lack of efficacy. Supplementing or replacing clinical trials with virtual patients (i.e., digital representations of human disease biology, e.g., digital twins in medical devices) is a viable solution that is gaining market and regulatory support for example in rare diseases where few patients exist or in patient populations with ethical concerns such as young children or pregnant mothers.
Virtual patient (Digital twin) VP/DT technologies takes drug asset data such as mechanisms of action, parameters of target engagement, pre-clinical data, and any other relevant data generated during drug discovery along with data describing the different disease patient endotypes as input, and computes a ranking of drug efficacy according to patient endotype along with a mechanistic rational that can be experimentally validated and interpreted by domain experts for the computation as output. VP/DTs model human biology at the individual patient level, which enables building virtual patient cohorts of sufficient heterogeneity and diversity to simulate clinical trials so that (non)-(super)-responder populations and their differentiating biomarkers can be identified.
While there has been progress in predicting clinical efficacy for best-in-class assets (i.e., drugs for which there are already drugs with the same mechanisms of action with clinical data), there is no VP/DT technology that can accurately and reliable predict clinical efficacy for first-in-class assets (i.e., drugs targeting novel mechanisms of action without any clinical data). Note that first-in-class assets are what lead to novel, life changing therapies for which there is no existing or satisfactory therapy. We are developing VP/DT technology using a combination of classic systems biology methods and new Agentic AI and Foundation Model methods in partnership with Sanofi that aims to overcome the limitations of existing VP/DT technology to predict the clinical efficacy of first-in-class assets and reduce the 90% failure rate of clinical trials so that more novel, life changing therapies can get to patients faster.
