Our platform offers cloud-scale artificial intelligence supportfor every design decision at each stage of drug discovery
Unlocking Emerging and Challenging Biology
Numerate’s learning algorithms can work with less data, noisier data, and more biased data than any alternative approach. The result is that, for example, an academic program that has uncovered exciting new biology with a few, non-drug-like “hits” and a phenotypic assay is sufficient for us to bootstrap into lead optimization / candidate identification within a matter of months, not years.
We follow the signals that scientists and clinicians want to follow, even if they are from high-content, low-throughput phenotypical assays that are poorly adaptable to high-throughput screening, structure-based design or traditional computational methods. This very quickly jumps programs ahead – to the lead design and optimization stage, where we then leverage the ability of our platform to look at extremely large universes of chemistry.
This same approach can be leveraged where there are larger amounts of data in more established drug programs. Here, in addition to greatly increasing the efficiency of lead design and candidate identification, our multi-parameter optimization allows us to greatly reduce the inefficiency and number of the design-test cycles in the lab.
Beyond Activity Prediction (ADME/Tox)
Because our AI algorithms have been designed to handle very noisy and biased data, we are able to apply our model building capabilities to all publicly and privately available data. Using our model building platform we have developed a suite of over 6,000 mechanism-of-action based models for more than 2,500 protein targets.
Using a custom designed AI approach, we have also been able to develop predictive ADME models. Our approach can capture the non-linearities inherent in ADME due to the interplay of physical and biological assay structure, biochemical interactions, and physicochemical properties. We can model many of the most important ADME properties, including intestinal absorption, active efflux pumping and metabolic stability. Our ADME models out-perform the best methods in the literature, increasing accuracy as the drug program moves forward.
We have invested more than $40 million in building, validating and applying our platform. This includes both traditional venture investment and support from the U.S. Department of Defense, where we have been extending and applying our platform to anti-chemical and bio-warfare countermeasure research.
Our platform has been applied successfully in the context of multiple pharma, biotech and academic collaborations. Our approach has been tested and refined over 10 years in more than 25 projects across multiple therapeutic areas, target families, types of modulation, and stages of program. In every case, these programs were extremely challenging, where the traditional process either was not applicable or had run into dead ends, from a chemical design perspective.