Cancer relevance of human genes

Tao Qing, Hussein Mohsen, Vincent L. Cannataro, Michał Marczyk, Mariya Rozenblit, Julia Foldi, Michael F. Murray, Jeffrey P. Townsend, Yuval Kluger, Mark Gerstein, Lajos PusztaibioRxiv 2021 · DOI

TL;DR

This study proposes that genes interacting with known cancer genes in a protein-protein network are also functionally relevant to cancer. They found that proximity to core cancer genes correlates with increased importance in cell viability, higher mutation rates, and stronger evolutionary pressure, suggesting a much broader set of cancer-relevant genes than previously appreciated.

What Problem Does This Paper Address?

Identifying all genes that contribute to cancer development is a monumental task. While hundreds of "driver" genes are known, it's unclear how many other genes, through indirect interactions, play a role in the complex process of malignant transformation. This paper aims to map the functional landscape of cancer-relevant genes by exploring the concept of "neighborhood" within protein-protein interaction networks, hypothesizing that genes physically or functionally linked to known cancer drivers are also likely to be important.

Methods

The researchers utilized a protein-protein interaction (PPI) network from the STRING database to define "neighborhoods" around core cancer genes (CCGs). They categorized genes based on their shortest distance (number of steps) from a CCG: 1-step (direct interactors), 2-steps, 3-steps, and >3 steps. They then assessed the functional relevance of genes in these categories using four independent metrics:

  • Network Connectivity: How many other genes a gene interacts with.
  • In vitro Cancer Cell Viability: Measured by gene dependency scores from the Cancer Dependency Map (DepMap) project, indicating how essential a gene is for cell survival when silenced.
  • Somatic Mutation Frequency and Cancer Effect Size: Analyzed from TCGA data to see how often genes are mutated in cancer and the impact of those mutations.
  • Germline Selection Pressure: Assessed using measures of negative selection against protein-truncating variants in healthy populations, indicating evolutionary conservation.

Datasets

  • STRING Database: A comprehensive database of protein-protein associations used to build the network and define gene neighborhoods.
  • Cancer Dependency Map (DepMap) Project: Provided gene dependency scores from CRISPR and RNAi screens across numerous cell lines to assess gene essentiality for cancer cell viability.
  • The Cancer Genome Atlas (TCGA): Used for somatic mutation data (missense, non-sense, frameshifting, etc.) across 32 cancer types to determine mutation frequencies and cancer effect sizes.
  • Exome Sequence Data (from UK Biobank and Exome Aggregation Consortium): Used to derive germline selection pressure metrics (Sh and pLI scores).
  • CancerMine Database: A text-mining resource to cross-reference genes with known cancer driver, oncogene, or tumor suppressor roles.

Key Insights and Ideas

  • The "Cancer Neighborhood" Hypothesis: The core idea is that genes don't operate in isolation. Their functional importance in cancer is strongly influenced by their proximity to known cancer drivers within the cellular interaction network.
  • Ubiquity of Cancer Relevance: A significant portion of human genes (around 70%) are found to be within 1 or 2 steps of core cancer genes, suggesting a much larger universe of potentially cancer-relevant genes than previously thought.
  • Proportionality of Importance: The study consistently shows a decreasing trend in functional importance as the distance from CCGs increases, across all four measured metrics. This provides strong evidence for the neighborhood effect.
  • Multiple Lines of Evidence Converge: The fact that four distinct metrics (connectivity, cell viability, somatic mutations, and germline selection) all point to the same conclusion strengthens the findings and provides a robust picture of gene relevance.

Key Results

  • Approximately 70% of human genes are either directly interacting with (1-step) or one intermediary away from (2-steps) core cancer genes.
  • Genes closer to CCGs exhibit higher network connectivity.
  • Silencing genes closer to CCGs leads to greater impairment of cancer cell viability.
  • Genes closer to CCGs show higher somatic mutation frequencies and larger cancer effect sizes.
  • Genes closer to CCGs are under stronger negative germline selection pressure, indicating their fundamental importance in normal cellular function.
  • The study provides a putative "cancer relevance annotation" for 16,932 human genes.

Limitations and Future Work

  • Tissue Specificity: The authors acknowledge that the functional importance of genes can be highly tissue-specific, and their analysis pooled data across many genes and cancer types, which might diminish the effect size for individual genes.
  • Discordant Scores: For individual genes, the four metrics (dependency score, mutation prevalence, effect size) don't always correlate perfectly, suggesting complex regulatory mechanisms and context-dependent roles.
  • Future Work: The study's findings provide a framework for prioritizing novel genes for further functional studies in cancer research. The annotated list of cancer-relevant genes can guide future investigations into the roles of these genes in various cancer types and their potential as therapeutic targets.

Summary updated

Abstract

Abstract Background It is unclear how many of genes contribute to the biology of cancer. We hypothesize that genes that interact with core cancer gene (CCG) in a protein-protein interaction network (PPI) may have functional importance. Methods We categorized genes into 1- (n=6791), 2- (n=7724), 3- (n=1587), and >3-steps (n=362) removed from the nearest CCG in the STRING PPI and demonstrate that the cancer-biology related functional contribution of the genes in these different neighborhood categories decreases as their distance from the CCGs increases. Results Genes closer to cancer genes manifest greater connectedness in the network, show greater importance in maintaining cell viability in a broad range of cancer cells in vitro, are also under greater negative germline selection pressure in the healthy populations, and have higher somatic mutation frequency and cancer effect. Conclusions Approximately 70% of human genes are 1 or 2 steps removed from cancer genes in protein network and show functional importance in cancer-biology. These results suggest that the universe of cancer-relevant genes extends to thousands of genes that can contribute functional effects when dysregulated.
BibTeX
@misc{qing2021cancer,
  author = {Tao Qing and Hussein Mohsen and Vincent L. Cannataro and Michał Marczyk and Mariya Rozenblit and Julia Foldi and Michael F. Murray and Jeffrey P. Townsend and Yuval Kluger and Mark Gerstein and Lajos Pusztai},
  title = {{Cancer relevance of human genes}},
  year = {2021},
  doi = {10.1101/2021.02.04.429823},
  howpublished = {bioRxiv},
}
Figure