SPLOOCE

Abstract

Understanding alternative splicing is crucial to elucidate the mechanisms behind several biological phenomena, including diseases. The huge amount of expressed sequences available nowadays represents an opportunity and a challenge to catalog and display alternative splicing events (ASEs). Although several groups have faced this challenge with relative success, we still lack a computational tool that uses a simple and straightforward method to retrieve, name and present ASEs. Here we present SPLOOCE, a portal for the analysis of human splicing variants. SPLOOCE uses a method based on regular expressions for retrieval of ASEs. We propose a simple syntax that is able to capture the complexity of ASEs.

Introduction

Alternative splicing events (ASEs) are present in almost all multi-exonic human genes^1,2 and are believed to be one of the most significant components behind the complexity of multi-cellular organisms.^1,3,4 Furthermore, ASEs are clearly involved in the etiology of a wide variety of diseases, including cancer,^5-7 ischemia⁸ and other common human disorders.⁹ Recently, several studies have shown that constitutive and alternative splicing are regulated by a complex network of cellular elements, which include a set of trans-acting factors and cis-acting sequences found in the primary RNAs.^10-18

The complex regulation of splicing and the high frequency of ASEs explain the appearance of Complex Alternative Splicing Events (CASEs), which are composed by a regulated combination of two or more single ASEs in transcripts from the same gene, or even in the same transcript. The most striking example of CASE is the Dscam in Drosophila, a gene containing a cluster of 48 mutually exclusive exons that, in principle, can generate thousands of splicing variants.¹⁹

In humans, some ASEs and CASEs occurring in oncogenes and tumor suppressor genes have already been associated to cancer.^6,20-23 For example, the gene NTRK1 (nerve growth factor) has a sequence variant, TrkAIII, which is common in certain tumors and lacks three exons that affect a regulatory immunoglobulin-like domain.²⁴ Further CASE examples include the gene CD44, which is a known marker of malignancy and invasiveness and has about ten ASEs that can occur in different combinations in its region coding for the extra-cellular portion of the protein.^20,21,23

In spite of the efforts of other groups,^25-27 a simple and efficient nomenclature to take into account all the variability generated by alternative splicing, especially for CASEs, is still missing. Here, we present a web portal, SPLOOCE, which uses a method based on regular expressions with an associated syntax. SPLOOCE provides a series of tools that allow users to profile splicing variants and analyze their functional impacts.

Results and Discussion

For the sake of space and clarity, we opted to describe the method used in this report as supplementary material, although an overview is present in Figure 1. In this section we present an implementation of the method in a computational tool, SPLOOCE and illustrate the use of SPLOOCE discussing a few examples.

Figure 1. General strategy to process and analyze CASEs in a set of expressed sequences.

Implementation

To make the method described in the supplementary material available to the community in an easy way, a web tool called SPLOOCE was implemented. SPLOOCE is available at http://www.bioinformatics-brazil.org/splooce. Data sources include RefSeqs, mRNAs, ESTs and data from NGS.

To help the users, SPLOOCE provides in the query box a quick reference table explaining the syntax and showing some illustrative examples. All ASEs and CASEs identified by SPLOOCE can also be displayed for a specific gene by simply typing the gene name between quotes in the query box (Fig. 2). SPLOOCE also provides advanced options for querying. Filters for chromosome, strand, gene name and sequence type are provided. Users can also evaluate the specificity of ASEs and CASEs expression regarding both tissue and pathology. A score for expression specificity is provided, which is a simple Χ² distribution analysis done among the expressed sequences supporting the corresponding variant. The analysis is based on the annotation provided by eVOC³³ for ESTs and manual curation for NGS sequences.

Figure 2. (A) SPLOOCE query form also showing a tab for advanced parameters. (B) Example of the table of results for a query. (C) Some results that can be found in the section “Details” provided by SPLOOCE. In this example, a double skipping (syntax: -e-s-s-e-) that codifies a protein domain is shown.

Results provided by SPLOOCE can be downloaded in a GFF file format. By default, results are shown in a table containing chromosome, genomic position, gene name and a pictorial view of the respective ASE or CASE, followed by the amount of their respective supporting sequences. SPLOOCE also provides a link to the UCSC Genome Browser (with tracks), and a local link for additional information.

When a Reference Sequence (RefSeq) is involved in an ASE, it is used as a template for creating a new mRNA sequence containing the specified event. For each of these new sequences, SPLOOCE predicts its open reading frame (ORF), which is then translated to a protein. Moreover, aiming to infer additional biological significance, SPLOOCE analyzes the protein domains using Pfam data and HMMER 3.0 program.³⁴ All these data are shown graphically (Fig. 2C).

Analysis

To illustrate the use of SPLOOCE, some basic questions about the frequency and mode of alternative splicing events were addressed. Table 1 shows the frequency of all types of ASEs in our data set. As expected, exon skipping (ES) is the most frequent type of alternative splicing. One interesting aspect that our method allowed us to explore is the distribution of dual-specific splicing (DSS) events. This type of event was found in 53 (0,27%) human genes (Table S2). In a less restrictive analysis, without considering the number of supporting sequences, the number of genes showing this type of event increased to 577. DSSs events were found to occur frequently in genes such as DIABLO, IRF3 and MAG, and they can occur together with other events as shown in Table 2.

Table 1. Frequency of the major types of ASEs.

Simple alternative splicing event	Genes	Total Events	Events per Gene
Exon skipping	10125 (51,77%)	38060	1,95
Alternative 3′ splice site	7490 (38,30%)	30172	1,54
Alternative 5′ splice site	7258 (37,11%)	27585	1,41
Intron retention	6565 (33,57%)	12632	0,65
Dual-specific splice site	53 (0,27%)	112	0,0057

Table 2. Frequency of DSS events coupled to other types of ASEs.

Syntax	Frequency (Genes)	Pattern (Simple)
d	577 (2.95%)	23 or 32
-d-	181 (0.93%)	0230 or 0320
-d-s-	85 (0.43%)	023030 or 032020
-Ed-	57 (0.29%)	01230 or 01320
-d-T	57 (0.29%)	023031 or 032021
-dE-	56 (0.28%)	02310 or 03210
f-d-T	26 (0.13%)	12023031 or 13032021
E-d-T	20 (0.10%)	1023031 or 1032021
-EdE-	17 (0.087%)	012310 or 013210
-d-t	11 (0.06%)	023021 or 032031
-d-S-	8 (0.04%)	023020 or 032030
-df-	5 (0,025%)	023120 or 032130
-tD-	3 (0.015%)	021320 or 031230
f-D-T	2 (0.01%)	12032021 or 13023031

Another interesting feature that our method allows us to evaluate is the combination of events occurring in the same mRNA molecule. Each pair-wise comparative sequence in our pipeline contains on average 1.6 events and 46.87% and 12.21% of these pairwise comparisons report more than one or two events, respectively.

To better understand what influences the frequency of ASEs and CASEs, some patterns were further explored. For example, the combination of two adjacent ES events is significantly more frequent among all sets of CASEs (Table 3). This excess is absent in situations when both events are not adjacent, like in the patterns –s-E-s- and –s-E-E-s- . Do these adjacent events tend to maintain the phase of an ORF? When adjacent, 60.78% of -E-E-s-s-E-E- maintains the ORF. This is significantly higher than what one would expect by chance based on all pairs of exons in the human genome that maintain an ORF (p < 0.001). This strongly suggests that adjacent ES events are under selection to maintain the ORF. The same pattern is not observed for other types of CASEs (data not shown).

Table 3. Number of genes affected by different types of ASEs and CASEs.

Syntax	Freq. Genes	Syntax	Freq. Genes	Syntax	Freq. Genes
-s-	14461 (73.04%)	-s-F-	3258 (16.66%)	-s-S-s-	597 (3.05%)
-f-	11220 (57.37%)	-s-f-	3052 (15.60%)	-t-E-t-	591 (3.02%)
-s-s-	10020 (51.23%)	-s-E-S-	2718 (13.90%)	-f-E-f-	549 (2.81%)
-s-s-s-	6844 (34.99%)	-s-S-S-	2627 (13.43%)	-f-E-F-	541 (2.77%)
-f-S-	5733 (29.31%)	-t-S-	2563 (13.10%)	-rR-	430 (2.20%)
-s-S-	5532 (28.28%)	-s-s-S-	2168 (11.08%)	-r-r-	272 (1.39%)
-s-T-	5302 (27.11%)	-s-E-E-s-	1627 (8.32%)	-r-R-	228 (1.17%)
-r-	4687 (23.96%)	-s-E-E-S-	1580 (8.08%)	-f-E-t-	226 (1.16%)
-s-t-	3934 (20.11%)	-t-T-	1150 (5.88%)	-rrr-	224 (1.15%)
-f-s-	3710 (18.97%)	-f-F-	1033 (5.28%)	-f-E-T-	218 (1.11%)
-f-T-	2866 (14.65%)	-t-t-	1022 (5.23%)	-s-E-S-E-s-	73 (0.37%)
-f-t-	2800 (14.32%)	-f-f-	1021 (5.22%)	-rRR-	42 (0.21%)
-s-E-s-	2768 (14.15%)	-rr-	928 (4.74%)	-rrR-	37 (0.19%)
-E-r-E-	2222 (11.36%)	-t-E-T-	629 (3.22%)	-rRr-	27 (0.14%)

Material and Methods

Public data

The human genome reference sequence (NCBI36/hg18) was downloaded from UCSC Genome Bioinformatics portal (http://genome.ucsc.edu). RefSeq sequences were downloaded from the Reference Sequence database (release 25; http://www.ncbi.nlm.nih.gov/RefSeq/). A total of 203,649 mRNAs sequences were downloaded from UCSC Genome Bioinformatics portal (file mrna.fa, for Homo sapiens). ESTs sequences were downloaded from dbEST (http://www.ncbi.nlm.nih.gov/dbEST/). RNA-seq reads were downloaded from SRA database (http://www.ncbi.nlm.nih.gov/sra; IDs: SRX003935, SRX003934, SRX003933, SRX003932, SRX003931, SRX003930, SRX003929, SRX003928, SRX003927, SRX003926).

Sequence alignment

All RefSeqs, mRNAs and ESTs sequences were aligned to the human genome using the protocol described previously.²⁸ In brief, first all long sequences (RefSeq, mRNAs and ESTs) were mapped against the human genome using the BLAT alignment tool²⁹ and only the best alignment for each sequence was selected. Next, those transcripts showing alignment identity greater than 95% and a covering more than 90% of its sequence length were remapped using SIM4³⁰ and stored in a relational database. RNA-seq data were mapped to human genome using Tophat-based pipeline,³¹ and all mapped reads were submitted to Cufflinks.³² Default parameters were used in both algorithms.

Sequence clustering

All mapped sequences sharing the same genomic region were grouped together using a gene-oriented strategy as described previously.^4,28 First, all RefSeq sequences were annotated based on the corresponding “official gene name” from NCBI-Gene (http://www.ncbi.nlm.nih.gov/gene/). Second, all non-RefSeq transcripts (mRNAs, ESTs and RNA-seq) presenting multiple exons and sharing one or more exon-intron boundaries (splice junctions) with a RefSeq sequence were merged together. Third, the remaining non-RefSeq transcripts showing only one exon and overlapping greater than 30 nt with a RefSeq transcript were grouped together. All clustering information was stored in a relational database.

Alternative splicing

Analyses of ASEs and CASEs were done using regular expressions, as detailed in the supplementary material. In the analysis of ASEs, only cDNA clusters containing RefSeqs and/or mRNAs or more than 10 ESTs/RNA-seq were used. The redundancy of ASEs, such as exon skipping and intron retention, was eliminated by comparing all events of the same type from each gene. Afterwards, all events showing position overlap were clustered together and counted as one event. All other events, such as 3′/5′ alternative splice sites and dual-specific splice sites, were counted by verifying the position of the alternative splice site. The number of genes affected by specific CASEs was defined through the analysis of the full list of comparative matrices, not considering the number of supporting sequences.

Conclusion

Although previous methods and interfaces have been proposed^25-27 for the study of alternative splicing, all of them present limitations as discussed before.²⁷ SPLOOCE, described here, is an efficient and complementary alternative for the analysis of alternative splicing events due to its high flexibility in the querying patterns and variety of applications.

Like ASTALAVISTA,²⁷ the method used by SPLOOCE is based in a comparison of all transcripts for a given locus. SPLOOCE, however, uses a notation system based on regular expressions to provide a simple and straightforward syntax for splicing events. The design of the syntax was developed to provide a set of simple and intuitive characters, and is actually capable of representing any CASE pattern, including those that have rare DSS events. The proposed syntax was successfully implemented and it may become a standard way for representing ASEs and CASEs.

Glossary

Abbreviations:

ASE

alternative splicing event

CASE

complex alternative splicing event

RNA

ribonucleic acid

mRNA

messenger ribonucleic acid

UCSC

University of California, Santa Cruz

NCBI

National Center for Biotechnology Information

EST

expressed sequence tag

RNA-seq

whole transcriptome shotgun sequencing

SRA

sequence read archive

RefSeq

reference sequence

BLAT

BLAST-like alignment tool

cDNA

complementary deoxyribonucleic acid

NGS

next-generation sequencing

eVOC

expressed sequence annotation for humans

ORF

open reading frame

Pfam

protein families

ES

exon skipping

DSS

dual-specific splicing

Regex

regular expression

SIM4

Computer Program for Aligning a cDNA Sequence with a Genomic DNA Sequence

HMMER

Computer Program for Biosequence Analysis Using Profile Hidden Markov

Supplemental Material

Supplemental material may be found here: 
www.landesbioscience.com/journals/rna/article/22181