KlickLabs at TREC 2018 Precision Medicine track

Lediona Nishani, Maheedhar Kolla, Gaurav Baruah

Abstract: Precision medicine aims to provide the most accurate course of treatment, personalized for each patient. The 2018 Precision Medicine (PM) track aims to build systems, that sift through biomedical articles to find relevant cancer treatments for patients, as well as search for clinical trials for which a patient might be eligible. Cancer, as a disease, can occur in various forms, be of different types, and affect multiple organs as well as bodily systems. Additionally, every patient–cancer combination presents with slight variations depending on genetics, age, sex, and other factors. As of 2017, 16% of PubMed articles related to cancer [5], and over 500,000 articles are added to PubMed each year, thereby, making quality information retrieval for PM an important problem to tackle, for the benefit of doctors and patients alike. In this work, we develop query-expansion methods, with an aim to compare their performance over the 2018 PM task. Our experiments indicate that medical ontology (NCIt) based expansion performs well for retrieving scientific abstracts from PubMed. Utilizing demographic information in queries improves the performance for clinical trial retrieval.