In the last 5 years, billions of dollars have been invested in the AI-for-drug-discovery space. Companies like insitro, Recursion, Atomwise, and BenevolentAI have made great strides to open the top of the drug discovery funnel. Pharma as an industry is better than ever at finding hit compounds that are likely to help patients. Figuring out which of these hits to optimize and develop, and determining which patients they will work best for, remains a more elusive problem.
Matching the right cancer drugs to the proper patient has confounded physicians and scientists for decades. Today, only 50% of late-stage cancer patients show any response to their first treatment. Additionally, current pharma efforts to find more effective drugs stall, with just 3.4% of cancer drugs making it from Phase I to patients.
Much of this failure can be attributed to the fact that cancer is best managed as a collection of many diseases with different treatment sensitivities. Patient-to-patient variation in treatment response is inevitable. Human variation is one reason the oncology drug development field is pivoting to find the best responders for a particular drug, rather than find one drug to treat every patient. This new strategy is critical for the success of clinical trials; drugs with a companion diagnostic to identify responder patient groups pass Phase II and Phase III clinical trials at double the rate of drugs without companion diagnostics. Unfortunately, many drugs do not have a clear biomarker predictor of success when entering clinical trials, and ultimately are not successful in treating cancer in the selected patient population.
Known Medicine’s platform allows us to understand which cancer patients will respond to which drugs and why. We can assess the “which patients” portion through our micro-tumors– we have shown a high correspondence between our platform responses and actual patient responses in the clinic. These functional outcomes allow us to essentially run a clinical trial for efficacy without dosing a single person. The “why” comes from patient biomarker profiles, where we can identify thousands of factors that could be predictors of patient responses to drugs. We then focus on just those biomarkers that correspond to expected responses to our new drug. As our dataset, we will move into discovering drugs of our own based on platform insights.
Our ability to capture patient-to-patient variability allows us to narrow the drug discovery funnel to identify a best-in-class molecule. We can match the best drugs to the ideal patient without having to expose a single person to ineffective therapies. Using our dense dataset of treatment responses and biomarker profiles, we can identify exactly which patients will benefit most from a new treatment, as well as detect common complex collections of biomarkers from those patients. By the end of 2021, we collected >3 million images of patient-specific microtumor organoids and their response to multiple pre-established cancer therapies. We anticipate our database will increase to >20 million images by the end of 2022.
Known Medicine is the perfect collection of data scientists, engineers and cancer biologists to identify new drugs and match them to the proper patient. Our novel platform has already been able to identify and match clinical drug responses to therapies in actual patients. We can provide the necessary images and -omic data to predict which patients will fail or respond to which cancer treatment, taking the experiment out of the patient. Ultimately, our imaging and -omic technologies can be used to avoid the astronomical costs of failed clinical trials. We have the expertise and technology in-house to continue to develop this new technology and find the best treatment for every cancer patient.