Non-phylogenetic identification of co-evolving genes for reconstructing the archaeal Tree of Life

Abstract

Assessing the phylogenetic compatibility between individual gene families is a crucial and often computationally demanding step in many phylogenomics analyses. Here we describe the Evolutionary Similarity Index (IES) to assess shared evolution between gene families using a weighted Orthogonal Distance Regression applied to sequence distances. This approach allows for straightforward pairing of paralogs between co-evolving gene families without resorting to multiple tests, or a priori assumptions of molecular interactions between protein products from assessed genes. The utilization of pairwise distance matrices, while less informative than phylogenies, circumvents error-prone comparisons between trees whose topologies are inherently uncertain. Analyses of simulated tree datasets showed that I_ES was more accurate and less susceptible to phylogenetic noise than existing tree-based methods (Robinson-Foulds and geodesic distance) for assessing evolutionary signal compatibility. Applying IES to a real dataset of 1,322 genes from 42 archaeal genomes identified eight major clusters of co-evolving gene families. Four of these clusters included genes with a taxonomic distribution across all archaeal phyla, while other clusters included a subset of taxa that do not map to generally accepted archaeal clades, indicating possible shared horizontal transfers by co-evolving gene families. We identify one strongly connected set of 62 co-evolving genes occurring as both single-copy and multiple homologs per genome, with compatible evolutionary histories closely matching previously published species trees for Archaea. An I_ES implementation is available at https://github.com/lthiberiol/evolSimIndex.

The Evolutionary Origins of Extreme Halophilic Archaeal Lineages

“…A similar transfer has also been suggested for the Nanohaloarchaea, which would complicate the interpretation of the ATPase phylogenies. To shed light on whether this is the case, we regressed and correlated the pairwise distance matrices of the gene families in the Nanohaloarchaea and the Thaumarchaeota (using a modification of the approach described in Rangel et al [2019 , 2021 ]; see Materials and Methods for details). Two analyses were performed with this gene family distance method, with different marker sets.…”

“…Using pairwise distance matrices built from sequence comparisons avoids relying on reconstructing phylogenetic trees where the placement of our groups of interest are already uncertain. This correlation of pairwise distance matrices of gene families is based on coevolution implementations found in Gueudré et al (2016) and Rangel et al (2021) . These pairwise distance matrices were regressed (in the sklearn Python module) against all other distance matrices in each respective data set (i.e., each gene vs. the 281 other genes, or 1 gene vs. 121 genes).…”