RCDA: A highly sensitive and specific alternatively spliced transcript assembly tool featuring upstream consecutive exon structures

Abstract

When applied to complex transcript datasets, current tools for automated assembly of mRNA sequences require long run times and produce exponentially increasing numbers of splice variants. Here, we describe RCDA, a genome-based transcript assembly tool comprising RCluster, that recursively clusters transcripts, and DAssemble, that generates composite transcript sequences through path-finding using a directed acyclic graph. Each exon included in a final transcript is associated with an array of all upstream consecutive exon structures obtained from original transcripts. When a depth-first-search path reaches an exon, the path is retained only if it contains a structure from that exon’s array. RCDA assemblies, therefore, include only those transcripts with experimentally supported exon patterns. When applied to >23,000 transcripts from human chromosome 21, using biologically reasonable filters, RCDA execution time was approximately 4 hours. RCDA outperformed ECgene in reconstructing RefSeq transcripts and in limiting the total number of transcripts and transcripts per gene.

EasyCluster2: an improved tool for clustering and assembling long transcriptome reads

“…We also tested our software on a second human dataset including 21,599 transcripts from chromosome 21. This dataset was used to compare the performance of a recently released program named RCDA with the first version of EasyCluster [ 7 ]. RCDA predicted 379 clusters while EasyCluster produced 389 groupings in 90 min.…”

“…To demonstrate the reliability of EasyCluster2 we compared our software with two recently released programs to cluster ESTs, RCDA [ 7 ] based on EST to genome alignments and wcd [ 8 ], an improved similarity-based tool. In order to make unbiased the comparison, we generated simulated Roche 454 reads in which the relationship between reads and genomic locus of origin is perfectly known.…”

“…Since the advent of next generation sequencing technologies, however, no updated software to employ genome-based clustering of long transcriptome reads has been released. Indeed, as a novelty, we found the program [ 7 ], developed to group and assemble ESTs pre-aligned to a reference genome, and an improved wcd [ 8 , 9 ] release to cluster ESTs without genomic information. While wcd implements a new algorithm based on suffix arrays to handle huge amount of reads generated by high-throughput sequencers [ 8 ], RCDA has been conceived only for ESTs produced by the classical Sanger sequencing and, thus, never tested on long sequences produced by next generation technologies as those from Roche 454 [ 7 ].…”

“…In absence of large curated benchmarks in which the relationship between reads and genomic loci of origin is perfectly known, the reliability of EasyCluster2 has been assessed by simulated reads generated taking into account the Titanium Roche 454 chemistry. Same simulated datasets have been used to compare EasyCluster2 results to those obtained using other recent programs as RCDA [ 7 ] or clustering software not genome-based as wcd [ 8 , 9 ].…”