The Predictive Ability of a Cancer Subtype Ontology Assignment

The following article features coverage from the American Association for Cancer Research (AACR) 2019 meeting. Click here to read more of Cancer Therapy Advisor‘s conference coverage.

Results of a study that evaluated the use of newly developed large cancer subtype resource, the Cancer Subtype Ontology (CSO), for patients included in The Cancer Genome Atlas (TCGA), as well as those enrolled in clinical trials of immune checkpoint inhibitors, provided evidence for the predictive ability of CSO cancer subtype assignment. The findings from this study were presented at the American Association for Cancer Research (AACR) Annual Meeting 2019.

Although many cancer subtypes have been identified based on pathologic and molecular characteristics, the CSO has expanded this list to include a total of 840 cancer subtypes, which subcategorize 40 different cancer histologies based on models developed from data on cancer genomics, transcriptomics, epigenetics, and proteomics, as well as information related to immune infiltration. Furthermore, the CSO uses a novel machine learning framework to assign clinical and preclinical samples to a particular cancer subtype.

In this study, researchers used the CSO to assign cancer subtype to subsets of patients included in the TCGA or enrolled in clinical trials of immune checkpoint inhibitor therapy, and compared cancer subtype assignment with treatment responses and  clinical outcomes.

A finding from studies involving the TCGA was that new model-based CSO subtype assignment technique accounted for “approximately 60% of the most significant outcome associations across genomic and molecular subtypings.”

With respect to studies of patients enrolled in 101 clinical trials of immune checkpoint inhibitors, the researchers showed that half of these clinical trials included cancer subtypes that were predictive for survival or response, and that most of these were novel CSO cancer subtypes.

For example, assignment of CSO subtype to patients with locally advanced or metastatic urothelial cancer who participated in the IMvigor 210 phase 2 trials of the programmed cell death ligand 1 (PD-L1) inhibitor, atezolizumab, (ClinicalTrials.gov Identifiers: NCT02951767;  NCT02108652) showed that patient subgroups with disease characterized by 2 particular cancer subtypes were significantly more likely to achieve a partial response (P =.0002; OR =3.87) or complete response (P =.002; OR =6.86) compared with patients assigned to other cancer subtype groups. Furthermore, one of those cancer subtypes was a novel cancer subtype identified through the CSO.

Disclosure: The presenters have disclosed financial ties to Data4Cure, Inc.

Read more of Cancer Therapy Advisor‘s coverage of AACR 2019 meeting by visiting the conference page.

Reference

1. Ronen R, DeBoever C, Kluge B, et al. The Cancer Subtype Ontology: Predicting response to therapy in clinical and translational research using 840 cancer subtypes. Presented at: American Association for Cancer Research Annual Meeting 2019; March 29-April 3, 2019; Atlanta, GA. Abstract LB217/9.