Artificial intelligence-mediated drug synergy prediction and validation in patient-derived ex vivo tumor organoid models

In precision oncology, single drug treatment  often leads to acquired resistance and tumor relapse. Novel drugs to overcome resistance are in development. Unfortunately, there is a substantial number of animals used, costs involved and high failure rates in the development of new drugs. Drug repurposing is increasingly becoming an attractive proposition because it involves ready to use compounds. Over the past decades we learned that the combination of multiple drugs can increase drug efficacy and reduce the development of drug resistance. However, to test any kind of drug combination is simply not possible from an experimental approach, as the number of drug compounds is extremely large and the effectiveness of drug combinations varies, depending on genotype-phenotype relations of the patient’s tumor. For that reason, in silico prediction of drug synergies can tackle this problem. 

Advancements in high-throughput screening with the availability of larger datasets have enabled machine and deep learning model development  (e.g. DeepSynergy, Matchmaker) for improved drug combination synergy predictions. However, all these methods lack biological explainability, hindering  clinical application. User-friendly platforms to interpret their results and ex vivo validation are so far completely missing. 

We propose to develop a novel and user-friendly Deep Learning algorithm, enabling us to predict and understand novel drug synergies. Drug-Drug interaction analysis will reduce the number of potential toxic combinations. Results will be  validated  ex vivo in patient derived tumor organoids for a robust clinical implementation. Our platform will allow drug repurposing, reduce toxicity and therefore reduce and replace animals.