Identification of orthologous genes across the genomes of Paracoccidioides spp

Files containing annotated protein coding sequences (CDS genomes) of the Paracoccidioides isolates were downloaded from NCBI, using the following accession numbers: Pb18 (RefSeq# GCF_000150735.1), Pb03 (GenBank# GCA_000150475.2), Pb300 (GenBank# GCA_001713645.1), PbCnh (GenBank# GCA_001713695.1), and Pb01 (RefSeq# GCF_000150705.2). The CDSs from these genomes were compared against each other, in order to identify all groups of orthologous genes (OGs) shared by two or more of the isolates, with the aid of the software OrthoFinder [34], using the software´s default parameters. Paralogous genes present within the same OG group were compared by multiple alignment, using Clustal Omega 1.2.4 [35]. The input parameters were set as follow: Output guide tree: false; Output distance matrix: false; Dealign input sequences: false; mBed-like clustering guide tree: true; mBed-like clustering iteration: true; Number of iterations: 0; Maximum guide tree iterations: -1; Maximum HMM iterations: -1. The Nexus-formatted matrix generated by Clustal Omega was then used to estimate genealogical relationships with the aid of Bayesian inference, using Mr. Bayes 3.0 [36]. The analysis involved 1,000,000 iterations, with savings at every 100^th tree, 1,100,000 generations, in four heated Monte Carlo Markov chains (MCMCs), with 0.5 annealing temperature, 100 000 MCMC generation burn-in and a 16-category C distribution. A consensus tree was generated after burn-in, using a 50% majority rule, which allowed discrimination between orthologous and co-orthologous genes in the different OG groups. This list of orthologues was then used as a guide to ensure consistent annotation of equivalent genes throughout the five Paracoccidioides isolates, during the reannotation process (see below).

Several genes that transcribe non-coding RNAs (ncRNAs) have been identified and annotated in the genomes of Pb01 and Pb18 (the only formal datasets available for ncRNAs in Paracoccidioides spp.). Thus, their sequences were downloaded from the Ensembl Fungi database ftp site [21] and orthologues for each of these ncRNA genes were mapped in the genomes of Pb03, PbCnh and Pb300, using Bwa-MEM, version 0.7.17.1 [37], running in a local Galaxy environment [38], using the default software parameters. The resulting BAM alignment files were converted to BED files, with the aid of BAM-to-BED Converter, version 2.27.1 [39,40], also using default software parameters, which facilitated organizing and comparing the predicted ncRNAs across all Paracoccidioides spp. isolates. Finally, information regarding these ncRNAs was incorporated into GFF3 files (see below), with the aid of BED-to-GFF Converter [41], version 2.0.0, also using default parameters. All ncRNAs (along with their respective annotations) received identification codes consistent with the ones currently employed to describe Gene_IDs in each Paracoccidioides genome, but containing the designation NC (for non-coding) as a suffix. Thus, ncRNAs mapped in the genome of Pb18 received Gene_IDs starting from PADGNC_00001, while ncRNAs for Pb01, Pb03, Pb300 and PbCnh received Gene_IDs starting from PAAGNC_00001, PABGNC_00001, ACO22NC_00001 and GX48NC_00001, respectively. Genes responsible for transcribing additional ncRNAs, including tRNAs and rRNA genes (18S, 28S and 5S rRNAs) had been previously described in the original annotations of the Paracoccidioides spp. genomes [18] and sequences for such elements were available from the "rna_from_genomic" fasta files downloaded from GenBank/RefSeq [20]. These genes were also matched to their respective orthologues, using the same procedure described above, but we chose not to change their respective gene IDs (i.e.: they were not labeled with the designation NC), in order to respect their current GenBank/RefSeq IDs.

Functional reannotation

The overall process employed for reannotating the Paracoccidioides genomes is schematically shown in S1 Fig. Initially, all CDSs from Pb18 were individually submitted to comparative BLAST analyses against InterPro, Pfam and Swiss-Prot. Next, these CDSs were BLASTed against the manually-curated fungal databases SGD (Saccharomyces Genome Database), CGD (Candida Genome Database) and AspGD (Aspergillus Genome Database) [27,28,29]. All BLAST analyses employed high stringency criteria, which included as cutoff, an E-value < e^-10, to identify orthologous genes containing functional descriptions among these databases. Information derived from all these BLAST analyses were compiled in a spreadsheet, along with Gene Ontology (GO) data regarding Pb18 genes, obtained from the Database for Annotation, Visualization and Integrated Discovery (DAVID), version 6.8 [42]. GO data were downloaded through the GO Direct option, in order to reduce the redundancy of terms, typically observed in GO analyses. Next, the metadata regarding each CDS was independently analyzed by three expert reviewers, in order to determine a consensus annotation term (ParaDB Annotation) for each CDS, which should be consistent with all the information derived from DAVID and from the BLAST searches previously performed. In a second round of annotation, a fourth reviewer compared the results of the three independent analyses and determined a final ParaDB Annotation term for each CDS. Finally, all information regarding the ParaDB Annotation, obtained for each Pb18 CDS, was transferred to the orthologues present in any of the Paracoccidioides isolates, using the list of orthologous genes (available at http://paracoccidioides.com/paracoccidioides-orthologous/) as a guide.

Next, all CDSs present in the genome of Pb01, which did not contain an orthologue in Pb18, were submitted to the same analysis procedure described in the previous paragraph. The same process was successively repeated with the remaining CDSs from Pb03, Pb300 and PbCnh, generating thorough and consistent annotations for all Paracoccidioides genomes.

In a subsequent annotation step, all CDSs that remained identified as hypothetical proteins were submitted to additional BLAST analyses, using a less stringent cutoff (E-value < e^-5), essentially as described above. These CDSs were incorporated into the ParaDB database (see below) with a flag (E-value < e^-5) highlighting the lower stringency criterion used to determine their respective functional annotation.

Availability of the generated data

The information derived from the reannotation process described above has been compiled in a relational database named ParaDB, aimed at centralizing up-to-date genomic annotations for the major representatives of the five species/subgroups that compose the genus Paracoccidioides. ParaDB is available at the URL http://paracoccidioides.com and provides users with tabular archives describing each CDS and ncRNA sequences, including their final ParaDB Annotation consensus description, along with information derived from all databases evaluated during the present study. Additional information regarding these elements in all five Paracoccidioides isolates can also be downloaded (as nucleotide, or amino acid sequences) through multi-FASTA files, carrying strings that show, for each element, their respective locus_tag ID, protein_product ID and consensus_description (ParaDB Annotation). Updated General Feature Format (GFF3) files for each genome are also available through ParaDB, to assist researchers interested in performing large-scale OMICs analyses with the Paracoccidioides spp. genomes. These GFF3 files have been built upon the original GFF3 files available in RefSeq (Pb18 and Pb01) and GenBank (Pb03, Pb300 and PbCnh), by replacing the original GenBank/RefSeq annotations with the respective ParaDB Annotation, with the help of MS Excel. Information regarding the ncRNAs obtained from Ensembl were introduced in the GFF3 files using the BED-to-GFF Converter, as described above.

Computational structure of ParaDB

The ParaDB environment is based on the Database Management System (DBMS) MySQL and was developed in Docker [33]. Management configuration of the DB was made using Rancher [43], a robust Docker systems management tool, widely used in data center environments and other complex computing ecosystems. The Rancher cluster created to manage ParaDB resources is hosted in a cloud computing environment at CloudatCost [44]. Currently, the environment provides a total of 200 GB of disk space, in solid state drives (SSDs), and 20 GB of random-access memory (RAM). These resources are distributed in 10 virtual CPUs (vCPUs), in a Kubernetes cluster framework [45] (see S2 Fig for details). The ParaDB user interface has been developed in PHP language, using Wordpress [46] and tools to assist in keyword/BLAST searches were implemented with the help of Wordpress plugins and widgets [47].

A Docker service, carrying the Docker virtualization containers necessary to run ParaDB can be downloaded from https://cloud.docker.com/u/paradb/, allowing users to perform fully functional local installations of ParaDB, in different computational environments, given the platform-agnostic nature of Docker systems (see below).

Identification of orthologous genes among the main Paracoccidioides isolates

To guarantee consistent genomic annotation of protein-coding genes among Paracoccidioides isolates, orthologous genes shared by two or more isolates were initially identified with the aid of the software OrthoFinder [34] and the Paracoccidioides spp. pan-genome derived from this analysis, containing all protein-coding sequences within the group, is shown in S3 Fig (a complete description of this pan-genome can be found in the ParaDB website, at http://paracoccidioides.com/paracoccidioides-orthologous/). Overall, the five isolates display a protein-coding pan-genome composed of 8365 groups of orthologous genes (OG) and share a core genetic pool that consists of 6396 OGs (~75%), reinforcing the close phylogenetic relatedness among members of this group of human pathogenic fungi.

Surprisingly, very few OGs could be found in association with only one isolate. In fact, no exclusive OGs were found in the genomes of Pb3 and Pb300, while Pb18 carries only 2 exclusive OGs of this type (OG0000046 and OG000738) and PbCnh displays only one (OG0006453). Even Pb01, which represents P. lutzii, the most distantly related species within the Paracoccidioides genus, contained only 2 exclusive OGs carrying protein-coding genes (OG0000018 and OG0006445). Not surprisingly, most of these OGs carry genes that typically display structural variations even among closely-related species, since they may perform similar biochemical functions, but interact with alternative substrates: OG0000018 and OG0000046 contain a series of Ser-Thr Protein Kinases, while OG0006445 and OG0006453 contain a series of plasma membrane ATP-binding cassette (ABC) transporters (OG0007385 contain genes that remained identified as hypothetical proteins). Currently, it is not possible to establish whether these genes contribute any kind of adaptive/biological specificities for the different Paracoccidioides isolates.

GenBank/RefSeq [20] also contained information regarding a series of non-coding RNAs from the five Paracoccidioides spp. isolates, including tRNAs and rRNA genes (18S, 28S and 5S rRNAs) (see Methods). All isolates displayed a similar number of tRNA genes, capable of providing all amino acids required for protein synthesis (see OG0008368 to OG0008478, at http://paracoccidioides.com/paracoccidioides-orthologous/). A similar situation was observed with the 5S rRNA genes (OG0008365), but the 18S and 28S rRNAs were present in significantly different copy numbers across the genomes of each isolate and could not be found in the genome of isolate PbCnh, probably reflecting problems with the currently available genomic assemblies (see OG0008366 and OG0008367, at http://paracoccidioides.com/paracoccidioides-orthologous/). A total of 32 additional ncRNA genes, responsible for transcribing small RNAs (sRNAs), small nuclear RNAs (snRNAs, including the spliceosome RNAs), small nucleolar RNAs (snoRNAs), the Telomerase RNA Component (tercRNA) and the Signal Recognition Particle RNA have also been mapped in the genomes of all five isolates, using as reference, a list of ncRNAs identified in the genomes of Pb18 and Pb01, available from the Ensembl Fungi database [21] ftp site. All Paracoccidioides spp. isolates share a common set of such elements, whose orthologues were easily identified in all genomes, with the aid of the short read Bwa-MEM aligner, as described in Methods (see OG0008479 to OG0008510, at http://paracoccidioides.com/paracoccidioides-orthologous/).

Functional reannotation of CDSs in the main Paracoccidioides isolates

To prevent using different nomenclature while annotating corresponding genes across the Paracoccidioides genomes, the reannotation process of protein-coding genes was carried out as described in Methods. Thus, genome reannotation of Paracoccidioides spp. was initially performed with Pb18 and the annotation data obtained for all genes in this isolate were propagated to the corresponding genes present in the remaining Paracoccidioides spp. genomes, using the list of orthologous genes, described above, as a guide. Next, genes present in the genome of Pb01, which did not contain an orthologue in Pb18, were submitted to the same procedure and the same process was successively repeated with the remaining genes from Pb03, Pb300 and PbCnh, generating consistent annotations for all Paracoccidioides spp. genomes. As a result, specific functions and/or structural descriptions could be assigned to 6003 out of 8390 protein coding genes mapped in the genome of Pb18, reducing the proportion of genes described in association with hypothetical proteins to 28,5% (2386 genes) (Fig 1). In a second reannotation step, all CDSs that remained identified as hypothetical proteins were submitted to a new BLAST analysis, using a less stringent cutoff (E-value <e^-5), allowing functional identification of additional 241 CDSs in Pb18, further reducing the proportion of hypothetical proteins in this isolate to 25.5% (2145 genes) (Fig 1). Reannotation of the remaining Paracoccidioides spp. genomes provided similar results with the other isolates, reducing the number of CDSs associated with hypothetical proteins to ~29–25% (with E-values < e^-10 and < e^-5, respectively), thus increasing the proportion of genes with functional/structural identification to ~71–75% (with E-values < e^-10 and < e^-5, respectively) in all cases, significantly improving the annotations of Paracoccidioides genomes, in comparison to the annotations currently available in any public biological data repository (see Fig 1).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Quantitative assessment of hypothetical and non-hypothetical proteins annotated in the Paracoccidioides spp. genomes.

The graph indicates the percentage of hypothetical and non-hypothetical proteins present in the ParaDB annotation, when compared to the official annotations available through GenBank/RefSeq, showing that the proportion of CDSs annotated solely as hypothetical proteins was reduced from ~60–90% to ~25–28%, in all strains. Figure also shows the absolute number of CDSs with functional classification, identified at high stringency (E-value < e^-10) and low stringency (E-value < e^-5) BLAST analyses, as well as CDSs that remained identified as hypothetical proteins, in the genomes of each Paracoccidioides isolate.

https://doi.org/10.1371/journal.pntd.0007576.g001

Organization of the ParaDB database

Results from the functional reannotations shown in Fig 1 were compiled in ParaDB, a relational database developed to centralize standardized functional genomic annotation from all five reference isolates of the genus Paracoccidioides. ParaDB (available at http://paracoccidioides.com) presents a simple and intuitive interface, through which such information is made available (S4 Fig). Initial access to the annotation data can be made by the “Databases” button at the center of the webpage, or through a specific pull-down menu, available at the upper right corner of the main ParaDB webpage (S4A Fig). The “Full Database” option directs users to specific annotation data for each of the Paracoccidioides isolates under study (S4A Fig). In the “Full Database” mode (S4B Fig), users have initial access to a table that displays each Paracoccidioides spp. CDS (identified by numeric codes that correspond to their original GenBank/RefSeq IDs) and their respective ParaDB consensus functional/structural designation. Genes responsible for transcribing ncRNAs, which were not present in the GenBank/Refseq CDS files can also be accessed from these table and carry the suffix NC (non-coding) in the gene IDs assigned to them during our reannotation effort (see Methods). Information regarding data derived from all databases employed in the comparative analyses described above can be accessed by clicking on the (+) symbol, available in each of the CDS cells. Alternatively, such information can be accessed by clicking the “Columns” button on the upper right corner of the table and selecting the desired databases (S4B Fig). In either case, links are provided to direct users to the orthologous genes found in the fungal-specific databases (SGD, CGD and AspGD), where a variety of additional information can be found.

Users may also download the entire annotation files using the “Downloads” button, available from the pull-down menu (S4 Fig). The download site also allows users to access multi-FASTA files (containing either nucleotide or amino acid sequences) for each Paracoccidioides strains and/or their respective General Feature (GFF3) format files, which may be of great assistance for a series of downstream analyses, such as the evaluation of data derived from large-scale gene expression experiments involving microarray hybridization, or RNA-seq, for example.

Local installation of ParaDB using Docker

ParaDB can be accessed and browsed directly on the web, at the URL http://paracoccidioides.com, through a variety of platforms, including personal computers of any kind, as well as mobile devices, such as cell phones and tablets, operating under either android or iOS operating systems. However, it is also possible for users to download and install a fully functional version of the database in their own personal computers, avoiding problems derived from low internet trafficking, host server instability, or communication restrictions, due to the presence of firewalls in local servers, for example. To accomplish that, ParaDB was developed in Docker [33], allowing configuration of the entire ParaDB environment in a Docker service, which has been submitted to both GitHub (https://github.com/paracoccidioidesdb) and Open Science Framework (https://osf.io/3sq97/) repositories, ensuring long-term data availability to researchers. This service can be downloaded and used to perform local installations of the database in alternative computing ecosystems, allowing users to conduct their data mining and analyses in a personal and stable working environment. Once installed in a local host, ParaDB will operate from two single containers, called ParaDB-Web and ParaDB-BLAST. Both containers can be consistently interchanged and deployed across different platforms, regardless of hardware and/or operating system (OS) specificities. This implementation is designed to ensure continued and full availability of ParaDB, independently of the original installation in our servers, allowing users to maintain mirrors of the entire database in their local environments. Additionally, the container concept allows the ParaDB infrastructure to be easily scalable, ensuring that hardware resources are provisioned whenever the computational environment reaches its limitations.

Hardware requirements for a local installation of ParaDB are reduced, and a personal computer (or notebook), running on Linux (recommended), containing 2 GB of RAM and 5 GB of available disk space can be used as host for installing a local version of ParaDB. The installation process is extremely simple and only requires previous installations of Docker [33] and Docker Compose in the host machine [48]. Once these components are available, only two steps are required to start a ParaDB environment in the host. In the first step, a Docker-Compose file is downloaded from the GitHub servers to the host machine. In the second step, the Docker Compose is executed, downloading and installing the images/containers of the standard ParaDB modules, starting the service. When Linux is the host machine's operating system, the following commands must be run on the terminal:

$ git clone https://github.com/ParacoccidioidesDB/paradb.git

$ cd paradb

$ docker-compose up -d

During the deployment process, some ports and disk volumes will be automatically configured on the host machine. Details about the ports and volumes created are available on the ParaDB website. In a standard implementation, ParaDB will use the local host address (IP: 0.0.0.0; HOST: http://localhost) as the default address for internal links. A video demonstrating the entire ParaDB implementation process in a local Linux environment is available at the URL http://paracoccidioides.com/local-install/.

Finally, it is worth mentioning that the ParaDB infrastructure can also be used by independent researchers to develop genome annotation projects involving other closely related fungi and its source code is freely available at: https://cloud.docker.com/u/paradb/.