Genome-Wide Ortholog Inference from Gene Trees

A gene tree is the canonical representation of the evolutionary relationships between the genes in a gene family. Thus, ortholog inference from gene trees is an important goal. However, no automated software tools are available that provide genome-wide ortholog inference from gene trees. A number of challenges had to be addressed to enable this. These included the efficient partitioning of genes into small, non-overlapping sets such that all orthologs of a gene are contained in the same set as the original gene; scalable and accurate inference of gene trees from these gene sets; automatic rooting of these gene trees without a user-provided species tree; and robust ortholog inference in the presence of imperfect gene tree inference. The OrthoFinder workflow was designed to address each of these challenges and is described in detail below.
By default, OrthoFinder infers orthologs from the orthogroup trees (a gene tree for the orthogroup) using the steps shown in Fig. 2. Input proteomes are provided by the user using one FASTA file per species. Each file contains the amino acid sequences for the proteins in that species. Orthogroups are inferred using the original OrthoFinder algorithm [10 (link)]; an unrooted gene tree is inferred for each orthogroup using DendroBLAST [24 (link)]; the unrooted species tree is inferred from this set of unrooted orthogroup trees using the STAG algorithm [33 ]; this STAG species tree is then rooted using the STRIDE algorithm by identifying high-confidence gene duplication events in the complete set of unrooted orthogroup trees [22 (link)]; the rooted species tree is used to root the orthogroup trees; orthologs and gene duplication events are inferred from the rooted orthogroup trees by a novel hybrid algorithm that combines the “species-overlap” method [31 ] and the duplication-loss-coalescent model [32 (link)] (described below); and comparative statistics are calculated. All major steps of the algorithm are parallelized to allow optimal use of computational resources. Only the orthogroup inference was provided in the original implementation of OrthoFinder [10 (link)]; all other subsequent steps are new and described below.