To The Top

Data-Driven Patients Stratification and Drug Repositioning


Zainab Al-Taie







Cancer is a heterogeneous disease and represents a great example of the need for selecting patient-centric rather than disease-centric treatment. De novo drug discovery is a time-consuming and high-cost process with a low success rate. Drug repositioning (DR) reduces the time, cost, and risk of developing new drugs because it recommends new uses for drugs already declared safe for human use. This study uses a novel data mining approach, exploratory data mining (EDM), and network analysis for patient cohort stratification and DR. This approach enables identifying homogeneous subgroups within a heterogeneous disease population and recommends drugs based on subgroup-specific molecular alterations.¬†¬†Colorectal cancer patients (CRC) were used as a case study for this work. Mining patients’ data and querying a drug knowledgebase then generated subgroups with significant contrast from the rest of the population. The difference between each subgroup and the rest of the population was then assessed based on the differences in genotypic, phenotypic, and drug-related biomedical entities. The result of this approach formed a network with heterogeneous biomedical components, including different drugs related to a given subgroup. For each subgroup’s network, drugs were ranked based on their connectivity to each subgroup’s gene signature and other biomedical entities like disease, organs, side effects, and others. The analysis was done using a big data ecosystem, where apache-spark was used to implement our data mining method, and network analysis was done using neo4j. We found that most top-recommended drugs for each subgroup are currently utilized in other malignancies or in CRC. Our method demonstrates that combining data mining with network analysis has the potential to discover unique patient subgroups and identify existing drugs to target these subgroups. Future work will focus on confirming these findings in tumoroid and xenograft models mimicking molecular alterations in humans.

Please contact Robert Sanders ( for Zoom link.