Comprehensive Benchmarking and Integration of Tumor Microenvironment Cell Estimation Methods

Cancer Res. 2019 Dec 15;79(24):6238-6246. doi: 10.1158/0008-5472.CAN-18-3560. Epub 2019 Oct 22.


Various computational approaches have been developed for estimating the relative abundance of different cell types in the tumor microenvironment (TME) using bulk tumor RNA data. However, a comprehensive comparison across diverse datasets that objectively evaluates the performance of these approaches has not been conducted. Here, we benchmarked seven widely used tools and gene sets and introduced ConsensusTME, a method that integrates gene sets from all the other methods for relative TME cell estimation of 18 cell types. We collected a comprehensive benchmark dataset consisting of pan-cancer data (DNA-derived purity, leukocyte methylation, and hematoxylin and eosin-derived lymphocyte counts) and cell-specific benchmark datasets (peripheral blood cells and tumor tissues). Although none of the methods outperformed others in every benchmark, ConsensusTME ranked top three in all cancer-related benchmarks and was the best performing tool overall. We provide a Web resource to interactively explore the benchmark results and an objective evaluation to help researchers select the most robust and accurate method to further investigate the role of the TME in cancer ( SIGNIFICANCE: This work shows an independent and comprehensive benchmarking of recently developed and widely used tumor microenvironment cell estimation methods based on bulk expression data and integrates the tools into a consensus approach.

Publication types

  • Research Support, Non-U.S. Gov't

MeSH terms

  • Algorithms
  • Computational Biology / methods
  • Datasets as Topic
  • Gene Expression Profiling / methods*
  • Humans
  • Models, Genetic*
  • Neoplasms / genetics*
  • Neoplasms / immunology
  • Neoplasms / pathology
  • Transcriptome / genetics
  • Tumor Microenvironment / genetics*
  • Tumor Microenvironment / immunology