Efficient and accurate detection of splice junctions from RNAseq with Portcullis

Abstract

Next generation sequencing (NGS) technologies enable rapid and cheap genome-wide transcriptome analysis, providing vital information about gene structure, transcript expression and alternative splicing. Key to this is the the accurate identification of exon-exon junctions from RNA sequenced (RNA-seq) reads. A number of RNA-seq aligners capable of splitting reads across these splice junctions (SJs) have been developed, however, it has been shown that while they correctly identify most genuine SJs available in a given sample, they also often produce large numbers of incorrect SJs. Herein we describe the extent of this problem using popular RNA-seq mapping tools, and present a new method, called Portcullis, to rapidly filter false SJs junctions from spliced alignments produced by any RNA-seq mapper capable of creating SAM/BAM files. We show that Portcullis distinguishes between genuine and false positive junctions to a high-degree of accuracy across different species, samples, expression levels, error profiles and read lengths. Portcullis makes efficient use of memory and threading and, to our knowledge, is currently the only SJ prediction tool that reliably scales for use with large RNAseq datasets and large highly fragmented genomes, whilst delivering highly accurate SJs.

Time-resolved transcriptome of barley anthers and meiocytes reveals robust and largely stable gene expression changes at meiosis entry

“…The BaRTv1 transcripts were mapped to the Golden Promise reference assembly (GMAP version 2018-07-04; -n 0 --min-trimmed-coverage=0.80 --min-identity=0.90). Input files for the Mikado file included a splice junction file generated by Portcullis (Mapleson et al ., 2017; default parameters), open reading frame identification of the transcripts by TransDecoder (https://github.com/TransDecoder/TransDecoder, default parameters; Haas et al , 2013) and blast results using DIAMOND BLASTx (Buchfink et al , 2015; --evalue 1e-5) against the NCBI non-redundant protein database as evidence. Together with the stringtie.gtf, scallop.gtf, pacbio.gtf and bartv1.gtf those were combined and scored, generating the Barley Anther Transcriptome (BAnTr).…”

Leveraging multiple transcriptome assembly methods for improved gene structure annotation

“…During the Mikado serialise step, data from multiple sources are brought together inside a common database. Mikado by default analyses and integrates three types of data: openreading frames (ORFs) currently identified via Trans-Decoder, protein similarity derived through BLASTX or Diamond and high quality splice junctions identified using tools such as Portcullis [25] or Stampy [26]. The selection phase (Mikado pick) groups transcripts into loci and calculates for each transcript over fifty numerical and categorical metrics based on either external data (e.g.…”

An improved assembly and annotation of the allohexaploid wheat genome identifies complete families of agronomic genes and provides genomic evidence for chromosomal translocations

“…full length, canonical splicing, non-redundant) transcripts used to train an AUGUSTUS wheat-specific gene prediction model. AUGUSTUS was then used to generate a first draft of the genome annotation, using as input Mikado-filtered transcript models, reliable junctions identified with Portcullis (Mapleson et al, 2016b), and peptide alignments of proteins from five close wheat relatives (B. distachyon, maize, rice, S. bicolor and S. italica). This draft annotation was refined by correcting probable gene fusions, missing loci and alternative splice variants.…”