Artificial intelligence-aided clinical annotation of a large multi-cancer genomic dataset

Abstract

To accelerate cancer research that correlates biomarkers with clinical endpoints, methods are needed to ascertain outcomes from electronic health records at scale. Here, we train deep natural language processing (NLP) models to extract outcomes for participants with any of 7 solid tumors in a precision oncology study. Outcomes are extracted from 305,151 imaging reports for 13,130 patients and 233,517 oncologist notes for 13,511 patients, including patients with 6 additional cancer types. NLP models recapitulate outcome annotation from these documents, including the presence of cancer, progression/worsening, response/improvement, and metastases, with excellent discrimination (AUROC > 0.90). Models generalize to cancers excluded from training and yield outcomes correlated with survival. Among patients receiving checkpoint inhibitors, we confirm that high tumor mutation burden is associated with superior progression-free survival ascertained using NLP. Here, we show that deep NLP can accelerate annotation of molecular cancer datasets with clinically meaningful endpoints to facilitate discovery.

Fig. 1. Example of a clinico-genomic analysis based on outcomes ascertained using natural language processing models: Association between TMB and progression-free survival after initation of immunotherapy.

High tumor mutational burden defined as >=20 mutations per megabase. Results in this figure represent unadjusted Kaplan-Meier curves. Events were recorded using the “PFS-I-and-M” endpoint, defined as the earlier of death, or the time by which both a medical oncologist note and an imaging report had described cancer progression/worsening. Progression/worsening was defined using natural language processing models applied to imaging reports and medical oncologist notes. Survival curves were not adjusted for left truncation, since progression events were possible prior to genomic testing and cohort eligibility.