AI drug positioning engine
Applying AI to match a drug asset with new indications or subpopulation of existing indications
Challenge
90% of blockbuster medicines only work for 30-50% of patients
- Lack understanding of TME and drug mechanism of action
- Incomplete knowledge of clinical heterogeneity
- Testing all possible drug combinations is impractical
Solution
Owkin’s DrugMATCH
- DrugMATCH is a suite of end-to-end tools that uses knowledge graphs, graph machine learning, biomedical literature and multimodal patient data to identify novel disease indications and associated subgroups for a given drug
DrugMatch is Owkin's AI drug positioning engine - designed to help us uncover how existing drugs could benefit new patient populations [02:59]
Drug positioning engine
How does DrugMATCH work?
Best-in-class data
Broad data access
We access patient data for hundreds of disease indications from public datasets.
We integrate this data with biomedical literature and multimodal patient data from our network of partners
Knowledge graph
Drug asset in broad disease space
Our knowledge graph represents known interactions between drugs, targets and indications.
- Tens of thousands of nodes
- Millions of connections between nodes
AI drug-disease matching
Prioritization of best drug-disease pairs
Our graph ML methods learn the patterns that made drugs/targets effective during past clinical trials.
We use AI to predict the best disease indications to potentially target with the drug of interest.
AI subgroup matching
Positioning of drug on patient subpopulations
Deep dive into a given drug-disease pair, to characterize the patient subtypes most likely to respond to that drug.
- Owkin data access
- Multimodal patient embeddings
- ML subpopulation optimization
Discovery
- Top indications potentially affected by the drug
- Matched patient subpopulations to increase treatment efficacy
AI drug positioning
Why work with Owkin?
Carefully constructed knowledge graphs
Expert team of biologists, medical experts and data scientists combine multiple data modalities to build quality knowledge graphs and disease maps.
World-class graph machine-learning expertise
We use graph neural networks & advanced interpretable AI to identify key regulators and high/low responder subgroups.
Unmatched access to patient data
Our AI is powered by patient data to capture disease heterogeneity and the TME for a multiscale understanding of disease biology.
Subgroup discovery
Multimodal AI-powered biomarkers
We combine cutting-edge machine learning and biology to identify biomarkers.
TargetMATCH
A new approach to target discovery
A suite of end-to-end tools that uses multimodal patient data as the input, and outputs the top candidate targets and paired patient subgroups that would most benefit from therapeutic intervention on these targets.