Extraction of Orthologs from Genome-Sequencing Data for Phylogenetic Analysis

Background

With huge advances both in evolutionary theories and sequencing technologies, phylogenetic analysis is entering a new era—phylogenomics. Current methods for phylogenomic inference can generally be categorized into two types: supertree and supermatrix methods [1–3]. The former approach obtains one supertree by combining inferred individual gene trees, each containing information from partially overlapped sets of taxa. Alternatively, the supermatrix approach analyzes the concatenated alignment of individual genes. Unavailable genes/loci are coded as missing data in the supermatrix [2,4]. Likelihood-based reconstruction methods are particularly suited for the analysis of supermatrices. These methods consider the heterogeneity across genes referring to evolutionary rates by using partitioned-likelihood models, which allow each gene to evolve under a different substitution model. According to the total evidence principle of using all the relevant available data, it is somewhat more popular as a strategy to adopt the supermatrix method, which is also used in the current pipeline of demonstrated examples.

The two crucial steps of standard phylogenetic inference are the identification of homologous sequences and tree reconstruction. Therefore, besides the accuracy of the tree-building method, the reliability of a phylogenomic tree also largely depends on the quality of homology, that is, the determination of paralogs and orthologs within and between genomes [5]. In contrast to paralogs, which are derived from gene duplication and should thus be excluded from phylogenetic analyses, orthologs are genes that are derived from speciation events; orthology, in this case, refers to the relationship between the corresponding genes in different species. So far, the most widely used methods for orthology inference can be classified into two groups [6]. One group infers pairwise relationships between genes in two species and then to multiple species, while the other identifies complete orthogroups (OGs), which are identified as the set of genes descended from a single gene in the last common ancestor of all of the species considered [5,7].

In the current pipeline, we use OrthoFinder, a popular method for inferring OGs of protein-coding genes. Starting from gene sequences (the input files), an advantage of using this program is that, by default, it infers OGs, orthologs, the complete set of gene trees for all OGs, the rooted species tree, and all possible gene duplication events. Furthermore, it also provides extensive comparative genomics statistics [5,7]. Despite the fact that OrthoFinder generates individual gene trees and the species tree, for customized tree building we recommend IQ-TREE (IQ-TREE 2 here) for subsequent analyses. In our laboratory, when working with multiple fungal genomes, IQ-TREE runs fast and provides automatic model selection, which also includes data partitioning, an efficient search algorithm for ML trees, ultrafast bootstrapping, and more [8]. With this protocol, we aim to demonstrate effective examples of orthology inference, ortholog extraction, paralog exclusion, individual gene tree reconstruction with the ML method, and data partitioning. This is done using five fungal genomes, with the four main members belonging to the genus Trichoderma. Trichoderma spp. are among the best studied groups of filamentous fungi due to their high value in applications from agriculture to industrial enzyme production [9]. The present protocol is simple and can also be easily adopted for genomic data from other organisms.