Toward comprehensive functional analysis of gene lists weighted by gene essentiality scores

Rui Fan, Qinghua CuibioRxiv 2021 · DOI

TL;DR

This paper introduces WEAT, a novel weighted gene set enrichment analysis tool that accounts for the varying "essentiality" of genes. By incorporating gene essentiality scores, WEAT can uncover biological insights missed by traditional methods, as demonstrated in case studies of aging and lung cancer.

What Problem Does This Paper Address?

Gene functional enrichment analysis is a cornerstone of bioinformatics, helping researchers understand the biological roles of gene lists. However, existing methods, like the hypergeometric test, treat all genes equally. This is problematic because genes aren't created equal in biological processes; some are far more critical than others. Imagine a team working on a project: a senior engineer's contribution is likely more impactful than an intern's, even if both are "working on the project." Traditional methods fail to capture this differential impact, potentially leading to missed discoveries. This paper addresses this limitation by proposing a way to weigh genes based on their essentiality, thereby providing a more nuanced and accurate functional analysis.

Methods

The core innovation is the Weighted Enrichment Analysis Tool (WEAT). WEAT builds upon the established hypergeometric test but introduces a crucial modification: it weights genes based on their "essentiality scores." Instead of simply counting genes, WEAT sums the weights of the genes that fall within a given functional category.

To handle the continuous nature of these essentiality scores, which can violate the discrete assumptions of the hypergeometric distribution, the authors employ the Gamma function. This mathematical tool naturally extends factorial calculations to non-integer values, allowing for a more accurate probability estimation.

WEAT also incorporates score normalization to ensure consistency between weighted and unweighted methods. This normalization process scales the essentiality scores so they have a mean of one, making the results comparable.

Furthermore, the paper explores different types of gene essentiality scores, including:

  • Conservation scores (e.g., phastCons, phyloP)
  • Gene Importance Calculator (GIC) scores
  • DepMap scores (lineage and disease-specific)
  • Inverse Document Frequency (IDF) scores
  • Protein-Protein Interaction (PPI) scores
  • Gene expression profiles from various tissues (normalized using log proportion)

Datasets

The study utilized several publicly available datasets for gene sets and gene essentiality scores:

  • Gene Sets: Kyoto Encyclopedia of Genes and Genomes (KEGG), Gene Ontology (GO), BioCarta, and Reactome.
  • Gene Essentiality Scores:
    • Conservation scores from UCSC.
  • GIC scores.
  • DepMap data.
  • STRING database for PPI.
  • GTEx (Genotype-Tissue Expression) project for tissue-specific gene expression.
  • Various other RNA-seq datasets from BioProject.

The authors also used specific gene expression datasets for their case studies:

  • Aging-associated periventricular white matter lesions (PVL) dataset (GEO: GSE157363).
  • Lung Squamous Cell Carcinoma (LUSC) gene expression data from The Cancer Genome Atlas (TCGA).

Key Insights and Ideas

  • Genes are not created equal: The fundamental insight is that biological genes have varying levels of importance or "essentiality" within biological processes, a fact largely ignored by traditional enrichment methods.
  • WEAT uncovers hidden signals: By weighting genes, WEAT can identify significant functional terms that traditional methods miss, especially when a few highly essential genes drive the enrichment.
  • Different scores reveal different biology: The choice of gene essentiality score significantly impacts the identified functional terms, suggesting that different scores capture distinct biological aspects. For instance, tissue-specific expression scores highlight tissue-specific functions, while conservation scores point to evolutionarily conserved pathways.
  • A web tool for broader accessibility: The development of the WEAT web server democratizes this advanced analysis, making it accessible to a wider research community.

Key Results

The paper demonstrates the utility of WEAT through two case studies:

  • Aging and Prion Pathway: WEAT identified the "Prion pathway" as statistically significant in age-associated periventricular white matter lesions (PVL), a finding missed by conventional methods. This was largely due to the high essentiality score of the GFAP gene in brain tissue, which traditional methods treated with equal importance as other genes in the pathway.
  • Lung Squamous Cell Carcinoma (LUSC): WEAT identified several novel terms related to LUSC, such as dilated cardiomyopathy and focal adhesion, using different essentiality scores (phastCons and GIC). These terms were not identified by conventional methods, highlighting WEAT's ability to find biologically plausible associations missed by simpler counting approaches.

The paper also shows that different essentiality scores have distinct preferences for functional terms, with tissue-specific expression scores favoring tissue-specific functions and conservation scores highlighting evolutionary pathways.

Limitations and Future Work

  • Knowledge Base Expansion: The authors suggest expanding the knowledge base with more gene sets and gene essentiality scores, and annotating more species.
  • Negative Scaling Factor: While the current WEAT algorithm uses a positive scaling factor (Q), the authors note that a negative Q could be useful for exploring "trivial" functional terms when using IDF scores.
  • Profile-Level Analysis: The paper proposes that the WEAT concept could be extended to profile-level two-phenotype gene set enrichment analysis.
  • Inter-organ Interactions: The ability to use gene expression essentiality scores from tissues different from the input case could reveal functional terms involved in crosstalk between organs, offering a new avenue for studying inter-organ interactions.

Summary updated

Abstract

ABSTRACT Gene functional enrichment analysis represents one of the most popular bioinformatics methods for annotating the pathways and function categories of a given gene list. Current algorithms for enrichment computation such as Fisher’s exact test and hypergeometric test totally depend on the category count numbers of the gene list and one gene set. In this case, whatever the genes are, they were treated equally. However, actually genes show different scores in their essentiality in a gene list and in a gene set. It is thus hypothesized that the essentiality scores could be important and should be considered in gene functional analysis. For this purpose, here we proposed WEAT ( https://www.cuilab.cn/weat/ ), a weighted gene set enrichment algorithm and online tool by weighting genes using essentiality scores. We confirmed the usefulness of WEAT using two case studies, the functional analysis of one aging-related gene list and one gene list involved in Lung Squamous Cell Carcinoma (LUSC). Finally, we believe that the WEAT method and tool could provide more possibilities for further exploring the functions of given gene lists.
BibTeX
@misc{fan2021toward,
  author = {Rui Fan and Qinghua Cui},
  title = {{Toward comprehensive functional analysis of gene lists weighted by gene essentiality scores}},
  year = {2021},
  doi = {10.1101/2021.04.26.441450},
  howpublished = {bioRxiv},
}

Loading related content...

Figure