OrthoDB v10: sampling the diversity of animal, plant, fungal, protist, bacterial and viral genomes for evolutionary and functional annotations of orthologs

1Genomic sequencing is the most comprehensive method for the molecular interrogation of organisms, with the potential to reveal the complete repertoire of genes and enable the study of cellular processes at the molecular level. Homology, the recognition of gene sequence similarities as evidence of shared ancestry, allows for hypotheses on a gene’s function when biological roles of related genes in other species are characterized. Homologs with a reference to a specific phylogeny radiation, i.e. descendants from a single gene of the last common ancestor, are termed orthologs and referred to below as ortholog groups or OGs. Such gene genealogies, pinned to particular ancestor genes, enable the most specific functional hypothesis for the descendant genes. Orthology is also the cornerstone for comparative evolutionary studies. The large-scale delineation of gene orthology is a popular but challenging task as evidenced by numerous proposed approaches.

2OrthoDB is one of the largest resources of orthologs. Beyond the benchmarks of the underlying algorithm, the accuracy of our methodology earned its reputation through many comparative genomic studies (e.g. ), particularly in the i5K initiative. The concept of orthologous groups is inherently hierarchical, as each phylogenetic clade or subclade of species has a distinct common ancestor; OrthoDB has explicitly emphasized this aspect since its inception. The ortholog delineation procedure is applied at each major radiation of the species taxonomy to produce more finely resolved groups of closely related species and to allow users to select the most relevant level.

3OrthoDB provides tentative functional annotations of groups of orthologs and mapping to functional categories by summarizing functional gene annotations, extensively collected from other public resources. Annotation of genes is complicated and contains errors. Although in many cases OrthoDB makes such errors in the underlying data apparent, discordant annotations should be considered with caution. The evolutionary annotations of the orthologs remain another distinguishing feature of OrthoDB (Figure 1). In this update (v10) we further increased the coverage of organisms, adjusted the underlying algorithm, and improved usability of the web interface.

4A substantial fraction of widely inherited genes evolve under the single-copy control. These are the easiest to predict and are the basis for our BUSCO tool. The BUSCO acronym stands for Benchmarking Universal Single-Copy Orthologs. The software aims at quantitative assessment of completeness of genome assemblies, gene sets or transcriptomes, based on evolutionarily informed expectations of gene content, complementing the technical metrics like N50. Although we derive these BUSCO marker genes from OrthoDB, OrthoDB also strives to resolve the challenging cases of gene duplications and losses. Sampling of species diversity was shown to be a major factor affecting accuracy of inferred gene orthology, besides the quality of the underlying genomes and their annotations. OrthoDB thus strives to cover as much sequence diversity as practical with our computational resources. When there are multiple genomes available with greater than about 96% identity using MASH estimates, we sample the best annotated representatives with the most complete gene sets according to BUSCO metrics.

5OrthoDB v10 now covers 1271 eukaryotes, 5609 bacteria, 404 archaea and 6488 viruses, as detailed in Table 1 (with the figures from the other resources as of September 2018). Overall, OrthoDB v10 covers over 37 million genes, classifying them into over 8.5 million tentative groups of orthologs at 624 levels of granularity. The orthology-levels, referring to the last common ancestors from which extant orthologs evolved, are defined according to the NCBI Taxonomy. Protein-coding gene translations for this release were retrieved mostly from RefSeq and NCBI complete genomes and the genome assembly ID is referenced now in the browsable taxonomy of organisms.

6The OrthoDB computational pipeline for delineation of orthologs is based on assessments of pairwise gene homology between complete genomes and their subsequent clustering (Figure 1). The pipeline has previously been described, and our software is freely available from https://orthodb.org/software.

7The latest adjustments to the OrthoDB algorithm include: (i) optimizations of the underlying data structures to reduce the memory usage, enabling us to increase the number of organisms covered, (ii) use of MMseqs2 for homology searches to speed up the computations growing as square of the total number of genes, (iii) introduction of an additional species overlap criteria for merging seed clusters to improve accuracy of inferred orthology and (iv) splitting of frequently mispredicted gene fusions to avoid spurious association of different orthologs.

8Gene functional annotations were collated from the major resources including Uniprot and NCBI gene records, as well as InterPro and Gene Ontology (GO). All data were processed to assemble consolidated and non-redundant per-gene annotation records, presented as short one-line descriptions that are click-expandable to immediately access the complete annotation record. The relative amount of available annotation data per gene is indicated by the size (one to five chevrons) of the click-expandable widget. Notably, we collated and made searchable references such as to KEGG pathways and to Online Mendelian Inheritance in Man (OMIM®) linking to human diseases.

9OG functional annotations are consequent aggregations from their corresponding gene-level annotations, aiming to provide the user with an overview of the possible functions of the member orthologs. The compilation of one-line orthologous group descriptors to briefly but precisely outline functional knowledge in a human-readable language is a non-trivial task. We achieved this by identifying the best scoring single phrase found in any part of available annotation for all genes in the orthologous group. All these phrases were matched using a full-text search engine educated with a list of biological stop words, against a body of all annotation records of all genes in the group. This querying was performed separately for subsets of the annotations, partitioned according to data provenance. An empirically evaluated weight factor was used to choose the best phrases from each source, with the data source precedence as follows: Uniprot to ENSEMBL to NCBI to Interpro and then GO.

10Functional categories of COG, GO, and KEGG pathways were assigned to OGs whenever possible. Such high-level functional descriptors are informative for comparative studies, e.g. in metagenomics.

11Evolutionary annotations are computed for each OG from the available genomic data and sequence alignment scores. As detailed earlier, these metrics include: ‘phyletic profile’ that reflects gene universality (proportion of species with orthologs) and duplicability (proportion of multi-copy versus single-copy orthologs), ‘evolutionary rate’ that reflects the relative constraints on protein sequence conservation or divergence and ‘sibling groups’ that reflects the sequence uniqueness of the orthologs. The universality of a gene family hints on a function that is widely necessary and basal, for example Orco is an essential co-receptor for insect odorant-sensing, while a lineage-restricted genes may underlie the lineage-specific adaptations, for example the Drosophila gland-specific peptide 26Ab. Duplicability is also indicative of the type of molecular functions, e.g. members of a signal-transduction pathway or a protein complex may evolve under the single-copy control, both of the examples above show this pattern (and the handful of ‘duplications’ or ‘missing’ of Orco actually point to deficiencies of the underlying genomes or their annotations). ‘Evolutionary rate’ is a relative measure, where slower than genome average evolution may indicate stronger purifying selection, like in the case of the Orco gene, while the faster than average evolutionary rate may indicate positive selection, like in the case of the gland-specific peptide 26Ab. Although specific function of the 26Ab peptide is still unknown, this protein is transferred from male to female during mating and may act as key player in species-specific male reproductive success and thus, may be subject to rapid evolution resulting from sexual conflict and competition. The ‘sibling groups’ allow navigation to gene families possibly having similar molecular functions. These annotations providing an evolutionary perspective remain unique to OrthoDB.

12The OrthoDB resource is public. The optional registration allows the authenticated users to upload their own data for performing online BUSCO analysis and for mapping to current OrthoDB OGs, enabling the user to explore mapped functional annotations and to generate user-tailored comparative charts depicting the total gene count, the fraction of common genes, the fraction of the most conserved single-copy genes, etc. The growing number of available genomes may hamper the user experience while browsing the data. Therefore, by default, orthologs from only user-selected or reference species are shown, and users may choose to toggle a check-box to view orthologs from all available species.

13The growing number of sequenced genomes increases the power of comparative analyses, but it also presents challenges regarding scalability of methods and data presentation to end users. OrthoDB strives to informatively sample the available genomic space and to refine the accuracy of ortholog delineations.

14As for the previous versions of OrthoDB we provide data files for bulk download, one file per level of orthology; as well as the underlying amino acid gene translations. To retrieve substantial subsets of data from OrthoDB or to access it programmatically we provide a REST API, documented at https://www.orthodb.org/v10/?page=api, that returns data in JSON, FASTA or TAB formats. All data are distributed under the Creative Commons Attribution 3.0 License from https://www.orthodb.org/.

15The RDF SPARQL interface was introduced in the previous OrthoDB v9.1 and it is gaining momentum, being compatible with both UniProt and Ensembl SPARQL endpoints. Adopting Uniform Resource Identifier (URI) of UniProt proteins and Ensembl genes, it provides the possibility for very elaborate queries and a number of clickable links to Ensembl Genomes, NCBI, Interpro and GO resources.

16Users can also navigate to OrthoDB records by following links from FlyBase ‘Orthologs’ section, UniProt ‘Phylogenomic databases’ section or NCBI ‘Additional links/ Gene LinkOut’ section.