OneThree Biotech was born out of the largest precision medicine institute in America with the purpose of decreasing the failure rate of bringing therapeutics to market using biology driven artificial intelligence. Our technology was built in a lab setting; allowing us to engineer our methodologies around real-world results. While working with major biotech and pharma partners; we’ve validated our predictions through work in target discovery, new indication positioning, compound and combination therapy identification, toxicity prediction, biomarker, and patient selection, and much more. In this white paper, we’ll discuss our methodologies behind patient selection through biomarker and indication identification.
Beginning with the data ingestion engine, the OneThree Biotech (OTB) approach starts by integrating over 40 distinct data types from 160+ different public and proprietary data sources. The backend OneThree database is fed into 7 different modules that make up the OneThree’s AI platform – each designed to address a different stage in the drug development pipeline. In this module overview we describe how the OTB Biomarker Identification pipeline can be used to identify specific predictive biomarkers for a given asset and how these biomarkers can be used to identify new indications for future development. This workflow was used in our recently announced partnership with Jubilant Therapeutics where we used it to:
- Generate new data to support the dual inhibition (LSD1 and HDAC6) to drive inhibition
- Pinpoint molecular markers that could be used to identify patients who might benefit from the dual inhibition
- Determine additional target indications for future trials
Cancer by nature is a heterogeneous disease, which can lead to highly variable patient responses for targeted therapies – even within the same clinically defined cancer type or subtype. Identifying drug sensitivity biomarkers, patient-specific traits that are highly correlated with a positive response has proven to be an effective strategy to identify positioning opportunities, and selecting patients. However, existing methods are refined to investigating either one type of biomarker (ie mutation-based, patient sex, etc.) or one type of mechanism of action, such as synthetic lethality or pathway dependencies. Here, we describe the creation of a flexible workflow that enables the identification of actionable biomarkers across various mechanisms of action.
Biomarker Identification Overview
The OTB Biomarker Identification module is an ensemble workflow, which leverages a flexible algorithmic architecture to enable the identification of a single biomarker as well as a collection of biomarkers that can be used as a sensitivity signature for indication positioning or patient selection (Figure 1). We work with our partners to ingest any available drug efficacy data (in vitro or in vivo results) matched with the baseline data of their choice. The type of data input that is fed into the workflow is often dependent on what types of biomarkers partners are most interested in (e.g. patient meta-data, genomic aberrations, gene expression, methylation, etc.). To properly capture the diverse mechanisms a therapy can work through we use diverse algorithm types to model distinct underlying biology. For example, linear models better identify genetic interactions, whereas non-linear models can better uncover more complex pathway based modes of action. By combining multiple different model types, we’ve found that we not only have better accuracy at capturing specific biomarkers, but we can also better understand the mechanism driving a given biomarker or signature. Once all models are recursively run to identify the most concise biomarkers that are predictive of sensitivity a final sensitivity signature is created.
The sensitivity signatures created can be used for patient selection in a cancer type of interest or indication positioning. When an asset is already being investigated for a specific cancer type, the sensitivity signature can be used within the clinic for patient selection to improve clinical trial outcomes. The flexibility of our approach enables the selection of actionable biomarkers, therefore ensuring clinical applicability. Additionally, OTB can leverage our backend clinical data, spanning tens of thousands of cancer samples, to rapidly identify top cancer subtypes that fit the previously identified sensitivity signatures. This approach enables us to not only identify new treatment populations but also predict the percent of distinct patient populations that will be sensitive or resistant, allowing for more informed downstream decision making.