Clinical trials are essential in the process of new drug development and repositioning. As clinical trials involve significant investments of time and money, it is crucial for trial designers to carefully investigate trial settings prior to launching it. In the 356,282 trials registered on ClinicalTrials.gov , one can search similar trial setting with the current trial of interest and identify prior or ongoing trials that share similar patient’s population, genetic characteristics, intervention means, etc. It is a wise strategy to learn from successful trials and to avoid repeating mistakes from failed trials. For example, in our computational drug repositioning project for colorectal cancers, given the complexity of cancer pathogenesis and diversity of cancer therapeutics, retrieval of relevant cancer trial documents with phase transition is nontrivial. Recently, MatchMiner provides a mechanism to extract genomic eligibility criteria from trial documents in addition to clinical information. However, the scope of such extraction mechanism is limited by the predefined data model and by the scarcity of genomic information described in trial documents. In this work, we represent cancer trial documents based on validated drug-gene interaction and eligibility criteria information for document retrieval. In addition, we explored learning-to-rank approaches for retrieval of cancer trial documents given a query document. Matching a new trial to the trial collection of ClinicalTrials.gov through document retrieval, we developed a relevance measure and evaluated the rank-based retrieval performance. In the future, we plan to further develop genetic and clinical information panels to facilitate trial design to achieve better trial success rate.
Please contact Robert Sanders (email@example.com) for Zoom information.