ERC analysis: web-based inference of gene function via evolutionary rate covariation

Abstract

Summary: The recent explosion of comparative genomics data presents an unprecedented opportunity to construct gene networks via the evolutionary rate covariation (ERC) signature. ERC is used to identify genes that experienced similar evolutionary histories, and thereby draws functional associations between them. The ERC Analysis website allows researchers to exploit genome-wide datasets to infer novel genes in any biological function and to explore deep evolutionary connections between distinct pathways and complexes. The website provides five analytical methods, graphical output, statistical support and access to an increasing number of taxonomic groups.

Availability and implementation: Analyses and data at http://csb.pitt.edu/erc_analysis/

Contact: nclark@pitt.edu

1 Introduction

Genes carry out their functions in a complex network of interactions, both physical and genetic. These interactions have effects on gene evolution for multiple reasons. First, genes performing a common function will experience shared adaptive and conservative evolutionary pressures, so that fluctuations in these pressures affect genes throughout the network. Second, changes in one gene product can affect the function of physically interacting gene products and thereby encourage compensatory changes. Both effects are hypothesized to cause genes participating in a common function to experience evolutionary rates that covary over time. This signature of co-functionality is quantified by the evolutionary rate covariation (ERC) statistic.

ERC measures the gene-by-gene correlation of evolutionary rates over a phylogeny of species, allowing the extraction of genes that have experienced parallel evolutionary histories. Generally, ERC values are significantly elevated between genes that encode proteins in a common complex, metabolic cascade or genetic pathway, and this observation extends across divergent taxonomic groups including prokaryotes, fungi, insects and mammals (Clark et al., 2013; Findlay et al., 2014; Juan et al., 2008). In practical application, ERC can be used to discover new functionalities via a sort of ‘guilt by association’. Recently, ERC signatures were used to expand the gene network controlling mating response in Drosophila (Findlay et al., 2014). In this study, ERC signatures between known pathway genes and an initial pool of 664 candidates were used to identify six new genes in the pathway. A similar study used ERC in yeast and Drosophila to discern the functional roles of homologous recombination genes (Godin et al., 2015). For human genes, ERC has been shown to efficiently identify causal disease genes out of chromosomal regions or other large sets of candidate genes (Priedigkeit et al., 2015). These successful applications make a compelling case for public access to ERC datasets so that geneticists and evolutionary biologists may similarly infer functional relationships between genes.

We have generated genome-wide ERC datasets focused on three taxonomic groups including humans and popular genetic models. The size of these datasets prohibits casual browsing and analysis, and requires custom computational tools and statistical tests in order to be used effectively. The aim of the ERC Analysis website is to allow users to quickly perform custom ERC-based inferences for specific gene sets of interest.

2 Implementation of ERC analysis

2.1 Datasets: scope and generation

The website, written in PHP and Perl, provides rapid, custom analysis of ERC datasets in three taxonomic groups—The mammalian dataset employs 33 species, Drosophila 12 species and yeasts 18 species, and these numbers will be augmented as more genomes become available. Protein-coding gene models and annotations were centered on a focal species for each taxonomic group. Mammals were based on human ‘UCSC genes’ as annotated at the UCSC Genome Browser; Drosophila melanogaster annotations were as in FlyBase and yeast genes as for Saccharomyces cerevisiae in Saccharomyces Genome Datanase (Cherry et al., 2012; Dos Santos et al., 2015; Karolchik et al., 2014). The website provides statistical analysis with rapid, memory mapped retrieval of data and pre-computed permutation test P-values fit to each application, dataset and sample size.

The underlying ERC datasets were calculated as previously described (Clark et al., 2012; Findlay et al., 2014; Priedigkeit et al., 2015). Generally, each protein was compared with its orthologs within its taxonomic group to estimate the rate of amino acid divergence along each branch of its phylogenetic tree. Each protein has a unique pattern of fast and slow branches relative to other proteins. Using these characteristic patterns, we calculated each rate covariation value (ERC) as the Pearson correlation of branch-specific rates between a pair of proteins (Sato et al., 2005).

2.2 Analysis tools

Analytical tools were written to perform the tasks most commonly requested by geneticists and molecular, cellular and evolutionary biologists. Broadly, these tasks (i) describe evolutionary correlations between co-functional genes, (ii) discover new candidate genes for a specific biological process and (iii) prioritize candidate genes for experimental validation (Fig. 1). Each of these analyses, detailed below, provides an easy-to-interpret results page complete with statistical support. When possible, output is available as a hyperlinked table or tab-delimited text to accommodate downstream computational analyses.

Fig. 1.

The ERC Analysis website offers five functions and provides results complete with tables and statistical analysis for each custom query

2.2.1 Group analysis

This tool allows the user to quickly measure the strength of the ERC signature within a group of genes known or suspected to be functionally related. This analysis is an important first step to assess the underlying strength of ERC in a given function before proceeding to subsequent analyses below. The results provide the pairwise matrix of ERC values between all listed genes along with a permutation-based P-value reflecting the elevation of the mean ERC within the gene set compared with random sets of matched size. An additional option can be chosen to cluster correlated genes within the matrix.

2.2.2 Multiple groups

The user submits multiple sets of genes, and the tool creates a matrix of all pairwise ERC values for each, their mean and median, their permutation P-value, and a Wilcoxon test to assess significant differences between groups.

2.2.3 Top genes

This popular analysis reveals the most correlated genes for a given input gene. It provides the biologist with a list of candidate genes enriched for functional association with the input gene. The tool returns an ERC-ranked list of genes along with their empirical rank-based P-value and a functional description of each gene.

Top genes will search either genome-wide or within a user-defined set of query genes. The ability to search within a specific set of query genes based on secondary criteria has proven highly beneficial for ERC applications (Findlay et al., 2014).

2.2.4 Top functions

Similar to top genes, this tool takes an input gene and returns the set of complexes or pathways associated with it via the ERC network. This tool is useful to provide candidate functions for poorly characterized genes.

2.2.5 Gene prioritization

When a biomedical study results in a list of candidate genes—from a mapping or association study for example—it is often difficult to experimentally test all candidates. This tool prioritizes those candidates for functional validation using ERC signatures with genes known to affect the phenotype in question. The user submits a set of training genes—from a particular disease or pathway for example—and a set of candidates. The tool then provides a ranked list of those candidates and statistics of their association with the training set. This type of analysis was first demonstrated in the context of human disease gene mapping (Priedigkeit et al., 2015).

3 Conclusion

The evolutionary history of a gene can be a powerful predictor of its function and can rapidly provide novel candidates for genetics studies. The ERC Analysis website provides this information to all biomedical researchers in a simple interface. Furthermore, because the power of ERC analysis generally improves with more species, the website is continually updated with increasingly powerful datasets gained by incorporation of newly sequenced genomes.

Funding

This work was supported by a Charles E. Kaufman New Investigator research grant from The Pittsburgh Foundation (KA2014-73920) and by a Pilot Grant from the NIH-supported Pittsburgh Center for Kidney Research (P30 DK079307).

Conflict of Interest: none declared.