Algorithmic matching of genomic profiles to precision cancer medicine clinical trials at DFCI.

Abstract

6620 Background: Genomic profiling and access to precision medicine clinical trials are now standard at leading cancer institutes and many community practices. Interpreting patient-specific genomic information and tracking the complex criteria for precision medicine trials requires specialized computational tools, especially for multi-institutional basket studies such as NCI-MATCH and TAPUR. Methods: To address this challenge we have developed an open source computational platform for patient-specific clinical trial matching at Dana-Farber Cancer Institute (DFCI) called MatchMiner, which aides in both patient recruitment to precision medicine trials, as well as decision support for oncologists. Trial matches are computed based on genomic criteria, including mutations, CNAs, and SVs, as well as clinical and demographic information, including cancer type, age, and gender. A formal standard called clinical trial markup language (CTML) to encode complex clinical trial eligibility criteria has also been created. Results: MatchMiner is now available at DFCI. Currently 123 precision medicine clinical trials have been transformed into CTML and 13,000 patient records are available, with over 88% of current patients having at least 1 match (average 2.6). A total of 103 genes are specified as criteria for at least 1 trial. KRAS, TP53, PTEN, PIK3CA and BRAF are the genes driving the most number of matches. General usage statistics and trial enrollment rates are currently being monitored to determine the system effectiveness. As this is an open source initiative, the software is also now publically available at https://github.com/dfci/matchminer. Conclusions: We have developed an open source computational platform that enables patient-specific matching and recruitment to precision medicine clinical trials at DFCI. We are actively seeking collaborators and plan to make CTML a multi-institution standard for encoding complex clinical trial eligibility in a computable form.

Establishing the Automatic Identification of Clinical Trial Cohorts from Electronic Health Records by Matching Normalized Eligibility Criteria and Patient Clinical Characteristics

“…Earlier studies have focused on parsing clinical trial eligibility criteria into computerinterpretable representations [15][16][17] to facilitate trial protocol design, automated cohort selection, and collaborative clinical research [18][19][20][21][22][23][24][25] . Expression and query languages such as Arden Syntax 23,26 , Guideline Expression Language Object-Oriented (GELLO) 27 , ECLECTIC 28 , and Clinical Trail Markup Language 29,30 , use a syntax similar to the computer programming languages for representing eligibility criteria in a computer interpretable format. Template-based approaches such as Eligibility Rule Grammar and Ontology (ERGO) 31 and Eligibility Criteria Extraction and Representation (EliXR) 32 , transform eligibility criteria into computable representations.…”