DoChaP: the domain change presenter.

Alternative splicing results in multiple transcripts of the same gene, possibly encoding for different protein isoforms with different domains. Whereas it is possible to manually determine the effect of alternative splicing on the domain composition for a single event, the process requires the tedious integration of several data sources; it is error prone and not feasible for genome-wide characterization of domains affected by differential splicing. To fulfill the need for an automated solution, we developed the Domain Change Presenter (DoChaP, https://dochap.bgu.ac.il/), a web server for the visualization of exon-domain associations. DoChaP visualizes all transcripts of a given gene, the encoded proteins and their domains, and enables a comparison between the transcripts and between their protein products. The colors and organization make the structural effect of alternative splicing events on protein structures easily identified. To enable the study of the conservation of exons structure, alternative splicing, and the effect of alternative splicing on protein domains, DoChaP also provides a two-species comparison of exon-domain associations. DoChaP thus provides a unique and easy-to-use visualization of the exon-domain association and conservation, and will facilitate the study of the structural effects of alternative splicing in health and disease.

INTRODUCTION

Alternative splicing creates multiple transcripts from a single gene, with the transcripts possibly coding for different protein isoforms, and it therefore increases the functional complexity of the eukaryotic proteome (1–4). Proteins, in turn, are composed of structural domains, each of which may have an independent function (5). Protein domains can be encoded entirely from a single exon or from several consecutive exons. In both cases, alternative splicing can cause either the loss of the domain or a change in the length and composition of the RNA sequence that codes for the amino acids sequence that folds into the domain. Hence, the functionality of the domain or even of the entire protein may be affected by alternative splicing. Indeed, it has previously been reported that alternative splicing changes protein functionality in tissue-specific splicing (6), protein interactions (7), cancer (8,9), resistance to biological therapeutics (10), and the immune response to viral infections (11). The use of high-throughput RNA sequencing (RNA-seq) data to study genome wide alternative and differential splicing is becoming more and more common (12). However, determining the functional implications of alternative splicing at the protein level involves a tedious manual process for each case of alternative splicing.

Manual determination of the effect of alternative splicing on protein domains in a single gene includes the following steps:

Identify the transcripts of the gene (from RefSeq and/or Ensembl).

Find the protein isoforms that are encoded by the identified transcripts (from RefSeq and/or Ensembl).

Predict the domains in each isoform separately (e.g. by Pfam or SMART) or look for the domains of the isoform in databases (e.g., NCBI’s CDD).

Manually compare the type, position, and length of the protein domains found between each pair of isoforms.

For each domain whose properties have changed, translate the positions of the domain in the two protein isoforms into positions in the transcripts, translate positions in the transcripts into genomic positions, and examine which exon(s) of the transcripts that encode the pair of isoforms are in this position and how this(these) exon(s) has(have) changed between the transcripts.

For each domain that is missing in one of the isoforms, translate the position in the protein that carries the domain into the position in the transcript, translate the position in the transcript into the genomic position, and examine which exon is in this position and how this exon has changed between the transcripts.

Furthermore, at present, experimental biologists who are studying specific cases of different functionalities of alternative protein isoforms do not have a tool for exploring the different isoforms of the protein alongside their domains. Currently available domain prediction tools [e.g., SMART (13), Pfam (14), NCBI’s Conserved Domains Database (CDD; 15)], and TIGRFAMs (16)] predict protein domains either from the amino acid sequence of known proteins or from user supplied query sequences. Such tools provide the domains' annotation and visualization for the protein of interest, and some even specify the location of the introns (13), but they only present a single protein isoform (encoded by a single transcript) at a time, regardless of the other known isoforms and their domain compositions. Several database tools have thus attempted to address the problem of studying the effect of alternative splicing on protein domains, e.g. ExDom (17), ProSAS (18) and ASPicDB (19). However, ExDom (17) and ProSAS (18) were built around a decade ago on much smaller source databases, and at present they are down and no longer maintained. ASPicDB (19) is limited to the human genome and does not show the transcript to isoform association or the exon to domain association. Therefore, the gap between the quantity of existing information about the effect of alternative splicing on protein domains and the ability of researchers to visualize this effect remains to be bridged.

To close this gap and thereby to provide researchers with an intuitive visualization of alternative splicing and information on protein domains, we built the Domain Change Presenter (DoChaP) web server. DoChaP provides a user-friendly, simple and intuitive gene-centric visualization of all the transcripts of a gene and the domains that they encode. DoChaP covers five species and presents the genes in their genomic context, alongside their transcripts and protein domains, thereby highlighting potential connections between exons and the protein domains that they encode.

METHODS

Data sources

DoChaP integrates information for transcripts and protein domains from several sources (Table 1) into a single SQLite database. Transcripts of RefSeq (20) and Ensembl (21), gene coding regions (CDSs) and exon annotations were extracted from GFF (general feature format) files, downloaded directly from the ftp sites of NCBI’s Genome (https://ftp.ncbi.nlm.nih.gov/genomes/refseq/) and Ensembl (ftp://ftp.ensembl.org/pub/current_gff3/). Protein information, connections of transcripts to protein isoforms, and domain annotations, descriptions and external identifiers were taken from two sources: (I) RefSeq: GenPept flat files (.gpff), downloaded from NCBI’s RefSeq ftp site (ftp://ftp.ncbi.nih.gov/refseq/{species}/mRNA_Prot/); and (II) Ensembl: BioMart data mining tool (22) (https://m.ensembl.org/biomart/martview/), using xml queries (templates can be found in the GitHub repository).

Table 1.

DoChaP data sources.

Source	Download path	File/table names	Obtained information
NCBI’s Genomes	https://ftp.ncbi.nlm.nih.gov/genomes/refseq/	Gff file	RefSeq's transcripts, genes CDS and exons annotations
NCBI’s RefSeq	ftp://ftp.ncbi.nih.gov/refseq/{species}/mRNA_Prot/	All GPFF files per species	RefSeq's proteins and protein domains data
Ensembl	ftp://ftp.ensembl.org/pub/current_gff3/	Gff3 file	Ensembl's transcripts, genes CDS and exons annotations
Ensembl	https://m.ensembl.org/biomart/martview/	Xml query templates are to be found in the GitHub	Ensembl's protein domains data
Ensembl	https://m.ensembl.org/biomart/martview/	Xml query templates are to be found in the GitHub	Ensembl-compara, Orthology information
InterPro	https://www.ebi.ac.uk/interpro/entry/InterPro/#table	Entries_table	Connection between domains accession from different sources
NCBI’s Gene	ftp://ftp.ncbi.nih.gov/gene/DATA/	Gene2ensembl	Transition from RefSeq to Ensembl identifiers

RefSeq and Ensembl use domain predictions from multiple sources, of which DoChaP presents NCBI’s CDD (15), Pfam (14), SMART (13), TIGRFAMs (16) and InterPro (23). Only domains that are associated with an entry of type ‘domain’ or ‘repeat’ (not ‘family’) from InterPro (23) are presented. Also excluded are InterPro domains that are not associated with any of the other domain sources of DoChaP. For DoChaP, the connection between domain accessions from different sources was inferred from InterPro entries that were downloaded directly from InterPro (23) (https://www.ebi.ac.uk/interpro/entry/InterPro/#table). The association between RefSeq and Ensembl accession numbers was taken from the Gene2ensembl table downloaded from NCBI’s Gene ftp site (ftp://ftp.ncbi.nih.gov/gene/DATA/). Orthology data for each pair of species in the DoChaP database was downloaded from Ensembl Compara via the BioMart data mining tool, using xml queries (xml templates can be found in the GitHub repository). For each species, DoChaP uses the latest genome build from RefSeq and Ensembl. In two species, rat and frog, the latest assembly from RefSeq was higher than the one from Ensembl. For those species, only the data from the source with higher version (RefSeq) is used. More detailed description of the data collection method can be found in the Supplementary note.

Database content

The DoChaP database currently includes information for five species: human (Homo sapiens, hg38), mouse (Mus musculus, mm39), rat (Rattus norvegicus, rn7), zebrafish (Danio rerio, danRer11) and frog (Xenopus tropicalis, xenTro10). The database content is detailed in Table 2 (relevant to March 2021), and the database schema is detailed in Supplementary Figure S1.

Table 2.

DoChaP database content for five species (as of February 2021)

Species	Gene IDs	Transcript- isoform pairs	Unique exons	Protein domain types
Homo sapiens (hg38)	20 272	177 531	425 709	11 167
Mus musculus (mm39)	23 161	157 623	390 326	10 676
Rattus norvegicus (rn7) a	22 063	74 270	275 677	7206
Danio rerio (danRer11)	29 633	71 205	361 335	9831
Xenopus tropicalis (xenTro10) a	21 683	44 877	248 980	7759
Total	116 812	525 506	1 702 027	46 639

aOnly includes RefSeq data because Ensembl current version is older than RefSeq current version.

Implementation

The DoChaP database was built using Python 3. The Ftplib package (https://docs.python.org/3/library/ftplib.html) was used for downloading from ftp sites. The Bio.SeqIO module of the Biopython (24) package was used for parsing the GFF and GPFF; the pandas (25) package was used for data parsing; and the SQLite3 package (https://docs.python.org/3/library/sqlite3.html) was used for building the database. The DoChaP web server is implemented with an Express.js on the Node.js environment and is distributed via XAMPP. The website is controlled by AngularJS and uses an HTML5 canvas for visualization.

RESULTS

Features and input

The DoChaP web server currently includes two visualization options for the exon–domain relationship. The first option is the ‘single species transcript comparison’ which provides genomic, transcript, and domain composition visualization of all the known RefSeq and Ensembl transcripts of a gene of interest in a specified species. The second visualization option is the ‘species comparison’ which provides the same information for a gene of interest in a selected species and its orthologous gene in one of the other available species.

The search term for DoChaP can be in any of the following formats: gene, transcript or protein identifiers (gene symbol; NCBI’s Gene ID or RefSeq transcript or protein ID; Ensembl gene, transcript or protein ID). In the single species transcript comparison, the species of interest must be specified. In the species comparison feature, the user is required to select a gene and species of interest; DoChaP will then search for all the orthologs of the gene in the other species and allow the user to select which species and ortholog to present. Orthologous genes are taken from Ensembl Compara. As some genes with identical gene symbols from different species are not defined as orthologs in Ensembl Compara (e.g. A2ml1 in human and mouse), comparison of such genes is also enabled in DoChaP.

Graphical visualization

DoChaP provides an intuitive, informative and straight-forward way of exploring the association of exons and domains. In the results page, all the transcripts of the gene of interest are shown with the following elements:

Genomic visualization

The genomic visualization provides the genomic context of the shown transcript, including the loci, strand direction, introns, exons and coding region. Exons are represented in different colors that are consistent across all the transcripts of the gene and through all the visualizations. Information on the start and end positions of the exon and its ordinal number in the transcript may be obtained by mouse hovering on each exon. Each transcript can be manually hidden from the view by clicking the ‘hide transcript’ button (another click will show it). The user can adjust the genomic range shown so as to focus on a specific set of exons. For genes that are coded on the reverse strand, the genomic visualization will be reversed (horizontally flipped), so that, regardless of the coding strand, the location of first exon is the left-most and that of the last exon is the right-most.

Transcript visualization

For each transcript in the genomic visualization, the exon structure of the coding region of the mature mRNA (without the untranslated regions) is provided. Exons are represented in different colors, corresponding to the relevant colors in the genomic visualization and remaining consistent across transcripts. The length and ordinal number of the exon in the transcript are available by mouse hovering over the exon. Transcript RefSeq and/or Ensembl IDs are shown above each transcript. The user can adjust the genomic range displayed so as to focus on a specific set of exons and domains.

Protein composition visualization

Protein domains are shown on a black line below the mRNA exon structure, and are represented as ellipses. The colors and positions of the domains are consistent with the encoding exon/s. Different colors within the same domain indicate that more than one exon codes for the domain (in-domain junction). The domain name is shown below the filled ellipse representing the domain. By mouse hovering over the domain, additional details such as the full name, position on the protein, length in amino acids and link to the external domain source are displayed. For simplicity, overlapping domains are grouped and represented as an ellipse with a double frame. Mouse hovering will show how many domains are represented within this shape and a click on the shape will expand the visualization to show all the grouped overlapping domains. Protein RefSeq and/or Ensembl IDs are shown above each protein isoform visualization.

Color guide to the exons

A key to the exon colors is presented in table form at the bottom of the results page. Each color assigned to an exon in that page is shown, and the start and end positions as well as the IDs of all the transcripts in which the exon is included are detailed in the table.

The ‘change display option’ allows the user to choose the type of visualization that is shown: all views, genomic alone, transcript alone, protein composition alone, or transcript and protein composition together. The default view displays all views. The ‘hide predicted records’ option enables removal of all RefSeq's predicted transcripts and proteins (prefix XM_ and XP_), and only curated RefSeq (prefix NM_ and NP) and Ensembl (prefix ENS) records will be displayed in the results. In the species comparison tab, all transcripts of the relevant gene from both species are shown in similar format, side by side. Each result page can be exported to a PDF by using the ‘Save as PDF’ button.

Example and biological impact

To demonstrate how the visualization of DoChaP can promote the understanding of the functional effect of alternative splicing, we use the tumor suppressor breast cancer gene 1 (BRCA1) as a model gene. BRCA1 is one of the most commonly mutated genes in breast and ovarian cancer (26). In Figure 1, we present the DoChaP view of human BRCA1 (3 representative transcripts out of 32). BRCA1 is encoded on the reverse DNA strand and consists of 24 exons (Figure 1A). Alternative splice variants of BRCA1, mainly focused in the first non-coding exon, and the coding exons 3 and 10 (E3, E10, respectively; Figure 1B and C), have been previously reported and widely studied in different contexts (27–29). The frequency of the alternative transcripts of BRCA1 changes between tissues (29), and might play a role in the tumorigenesis of breast cancer (30). For example, exon number 3 is skipped in BRCA1 transcript NM_007297 (E3, Figure 1A left and zoom in Figure 1B). Since this exon encodes part of a RING finger motif (Figure 1A, right panel), the resulting protein product lacks the RING finger domain (‘RING-HC_BRCA1’), and this, in turn, might affect the cell cycle regulatory function of this protein isoform (29). The largest exon of the BRCA1 gene is exon E10 (3426 bp; Figure 1A left and zoom in Figure 1C). Strangely, this exon is sometimes referred to as exon 11, because of an upstream Alu element that was inserted into BRCA1 clone (historically referred to mistakenly as exon 4) (26). Two of the presented transcripts (NM_007297 and NM_007300) include the full-length E10 (3426 bp, Figure 1C). However, in transcript NM_007298, an alternative 5' splice site is used and only the first 117 bp of this exon remain in the mature mRNA (E10117, Figure 1C). The use of the alternative 5' splice site of exon 10 which leads to the expression of E10117 causes the loss of specific protein binding sites [such as RB, p53, MYC, RAD50, TUBG (γ-tubulin) and angiopoietin-1] and the loss of a nuclear localization signal, and therefore affects the functionality of the translated protein (27). At least some of those changes may be due to the loss of the ‘BRCT_assoc’ domain, which is a serine-rich domain associated with the BRCA1 C-terminus, and is encoded from the full version of exon 10 (Figure 1A, right panel) and is seen only in the protein products of the transcripts that include the full-length exon 10 (Figure 1A). In the case of BRCA1, DoChaP provides a clear and intuitive visualization of the complex pieces of information that have been collected over many years in several different studies and of all the known domains that exist (or not) according to the annotation of all transcripts.

Figure 1.

Sample output of DoChaP for the human breast cancer susceptibility gene BRCA1. (A) Left, genomic visualization of the transcripts in their genomic context (genomic region, genomic range and strand are on the top). The sliding scale and zoom buttons control the genomic region displayed. The genomic range shown is chr17:43 127,790–43 044 294 which is the genomic region coding for the displayed transcripts of the gene BRCA1. Right, mRNA and protein domain composition for each transcript. Different colors represent different exons and are consistent across all the visualizations of the same gene. Domains are shown as circular shapes and are colored according to the exons that encode for them. In BRCA1, the top transcript encodes for four domains. The BRCT domain exists in all three isoforms shown. The x-axis is the position in the protein and coding region, as shown on top. Sliding scale and zoom buttons control the transcript and protein region shown, and double click on an exon in a transcript in the genomic visualization will zoom in to the relevant region of the corresponding transcript in the mRNA and protein domain composition visualization. (B) In the second transcript, the third exon (E3) is skipped and therefore the RING protein domain is missing (‘RING-HC_BRCA1’). (C) The third protein isoform does not include the serine rich domain (‘BRCT_assoc’), encoded by exon 10 (E10), as its associated transcript has a shorter exon 10 (E10117) due to an alternative 5’ splice site event. For the sake of simplicity, only three representative transcripts of BRCA1 are shown.

Limitations

The public databases used as sources for DoChaP include redundancies and inconsistencies. To overcome this problem, we currently use the RefSeq – Ensembl conversion table gene2ensembl (from NCBI’s ftp, see Methods). In cases where matched RefSeq and Ensembl protein records have different lengths (of more than one amino acid), the records are separated into two transcripts in the visualization. For example, the protein isoform NP_001351656.1 (RefSeq) and ENSP00000498543.1 (Ensembl) of the gene AGRN have different lengths of 1944 and 1940 amino acids, respectively), and thus appear as two transcripts even though they are mapped to each other. In all other cases, matched records are merged into one transcript to avoid redundancies.

Different data sources may indicate slightly different start and end positions of the domains in the protein, probably due to differences in domain prediction methods. For example, in the protein isoform NP_000236.2 of the human gene Met, Pfam locates the Sema domain between amino acids 57–491 (pfam01403) and SMART locates the same Sema domain at 52–496 (smart00630). In such cases, we present the coordinates of all the overlapping domains and show the source from which the coordinates of each domain occurrence were taken. Overlapping domains are shown as a single ellipse with a double frame that can be expanded upon a click.

Finally, some very long transcripts and transcripts with many exons make the genomic visualization of an entire gene less efficient in terms of colors and relative exon sizes. In such cases, the user is advised to zoom in and explore the gene in several consecutive windows.

CONCLUSIONS AND FUTURE EXTENSIONS

Improved sequencing technologies are leading to increased interest in the connection between exons and domains and in the evolution of the regulatory mechanisms of alternative and differential splicing. Previously, to explore exon–domain associations, researchers had to search manually and then integrate and interpret textual data from several sources in a tedious and error-prone process. With the development of DoChaP (made possible by the increasing quantities of freely available data in the public domain), the research community has access to a fast, intuitive and easy-to-use visualization tool for exploring the exon–domain relationship.

DoChaP is an ongoing project, and we plan to add new features and visualization capabilities and to increase the database in updates. Specifically, we intend to add additional species and other types of functional features of proteins, such as signal peptides. In addition, we will integrate into DoChaP a method for comparison of transcripts that will allow the transcripts to be ordered according to their genomic similarity and domain organization and that will present similarity scores for transcripts of the same gene and between orthologs.

Given the increased interest in the effect of alternative splicing on functionality in health and disease, we believe that DoChaP will allow researchers from the biological, medical and computational biology fields to interpret the structural effect of splicing changes toward new functional findings.

DATA AVAILABILITY

The DoChaP web server is freely available at https://DoChaP.bgu.ac.il. The latest DoChaP database may be downloaded under ‘Downloads’ tab in the DoChaP website. The source code is available on GitHub (https://github.com/Tal-Shay-Group/DoChaP).

Click here for additional data file.