Special Focus on Biomedical Data Science

Abstract

JAMIA has documented the evolution of biomedical informatics through its dissemination of original research and applications, brief communications and case studies, thought-provoking perspectives, and insightful reviews. The number and diversity of data-driven models have increased substantially in the past few years. From the first developments in machine and statistical learning that were applied to health sciences decades ago, our field has flourished to include biomedical data science as one of its important components, which is possible only because of other informatics work that allows data to be standardized, integrated, and used in various learning models. This issue is focused on biomedical data science and illustrates a broad range of techniques and application areas in this field; articles submitted in response to a specific request for papers are featured in an editorial by Brennan et al. (p. 2). In addition to the articles covered in the editorial, this issue highlights tools and applications of data science in a variety of domains, all of which use clinical text as a source of data: Trivedi (p. 81) presents an interactive tool for processing clinical text, Luo (p. 93) uses convolutional neural networks to classify relations in clinical notes, and Bejan (p. 61) introduces an approach to find homelessness and adverse childhood experiences described in clinical narratives. Additionally, nonclinical text is increasing in importance for health care and public health. Xie (p. 72) uses recurrent neural networks to find e-cigarette adverse events in social media posts, while Vigo (p. 88) describes a method to collect seasonal allergy symptoms for the British population. New types of structured data and new ways to integrate them are also continuously being produced: Doostparasti (p. 99) describes a novel approach for integrating -omics data to enhance phenotype classification performance, and Yu (p. 54) introduces a phenotyping algorithm that does not depend on expert-labeled observations. The articles listed above are only a few examples of the scope of informatics activities covered in JAMIA. Starting with this January issue, readers will be able to easily group articles into themes based on technologies used or application areas. This grouping is made possible by JAMIA’s change in frequency and format (to monthly online), which will allow for more frequent indexing. Readers will be able to compare approaches and discover solutions that are best suited to their problems. Stay tuned for additional data science articles in future monthly issues, as well as articles focused on clinical informatics systems (including clinical decision support), clinical research systems, translational bioinformatics, global public health informatics, and many other subfields of informatics that help us, through information technology, understand and address human health and disease.