OncoTree: A Cancer Classification System for Precision Oncology

Abstract

Purpose: Cancer classification is foundational for patient care and oncology research. Systems such as International Classification of Diseases for Oncology (ICD-O), Systematized Nomenclature of Medicine Clinical Terms (SNOMED-CT), and National Cancer Institute Thesaurus (NCIt) provide large sets of cancer classification terminologies but they lack a dynamic modernized cancer classification platform that addresses the fast-evolving needs in clinical reporting of genomic sequencing results and associated oncology research.

Methods: To meet these needs, we have developed OncoTree, an open-source cancer classification system. It is maintained by a cross-institutional committee of oncologists, pathologists, scientists, and engineers, accessible via an open-source Web user interface and an application programming interface.

Results: OncoTree currently includes 868 tumor types across 32 organ sites. OncoTree has been adopted as the tumor classification system for American Association for Cancer Research (AACR) Project Genomics Evidence Neoplasia Information Exchange (GENIE), a large genomic and clinical data-sharing consortium, and for clinical molecular testing efforts at Memorial Sloan Kettering Cancer Center and Dana-Farber Cancer Institute. It is also used by precision oncology tools such as OncoKB and cBioPortal for Cancer Genomics.

Conclusion: OncoTree is a dynamic and flexible community-driven cancer classification platform encompassing rare and common cancers that provides clinically relevant and appropriately granular cancer classification for clinical decision support systems and oncology research.

FIG 1.

OncoTree homepage and tree structure. All cancer types are represented by a node on the tree. All subclassifications are connected to parent nodes through branches. The location of the cancer is based on the cell of origin and histologic architecture. Although all melanomas fall under the main type Melanoma, the location of each cancer on the tree is based on the organ of tumor origin. In this case, Mucosal Melanoma of the Esophagus is located under the Esophagus/Stomach node; Uveal Melanoma is located under Ocular Melanoma, which itself is nested under the Eye node; and Cutaneous Melanoma is located under Melanoma, which itself is nested under the Skin node. This structure of the tree allows connecting nodes across branches on the basis of histology.

FIG 2.

The OncoTree API. (A) Interactive documentation of the OncoTree API. (B) The API returns all curated attributes, which include cancer type full name, a unique OncoTree code of the tumor type, the tissue type, and the OncoTree code for the parent tumor type and any mapping to external sources. Historical information if applicable are also returned. API, application programming interface.

FIG 3.

Overview of the curation process in OncoTree. OncoTree relies on a variety of sources for new additions and modifications. All new suggestions and updates are reviewed by the OncoTree committee and if approved are applied to the next OncoTree release for downstream applications.

FIG 4.

Overview of the different tools and resources that use OncoTree as a cancer classification vocabulary. From the clinical aspect, OncoTree codes are assigned in each genomic patient report at MSK by a clinician. In research, projects such as AACR Project GENIE and cBioPortal use OncoTree to assign codes to all tumor samples in their cohorts. OncoKB also uses OncoTree diagnosis codes to curate the knowledge base and to annotate clinically actionable variants in the genomic patient reports at MSK. AACR, American Association for Cancer Research; GENIE, Genomics Evidence Neoplasia Information Exchange; MSK, Memorial Sloan Kettering.