Limits...
VTBuilder: a tool for the assembly of multi isoform transcriptomes.

Archer J, Whiteley G, Casewell NR, Harrison RA, Wagstaff SC - BMC Bioinformatics (2014)

Bottom Line: From the simulated reads, VTBuilder constructed 55 transcripts, 50 of which had a greater than 99% sequence similarity to 48 of the SSTs.Unlike other approaches, VTBuilder strives to maintain the relationships between co-evolving sites within the constructed transcripts, and thus increases transcript utility for a wide range of research areas ranging from transcriptomics to phylogenetics and including the monitoring of drug resistant parasite populations.Additionally, improving the quality of transcripts assembled from read data will have an impact on future studies that query these data.

View Article: PubMed Central - PubMed

Affiliation: Department of Parasitology, Liverpool School of Tropical Medicine, Pembroke Place, Liverpool, L3 5QA2, UK. john.archer.jpa@gmail.com.

ABSTRACT

Background: Within many research areas, such as transcriptomics, the millions of short DNA fragments (reads) produced by current sequencing platforms need to be assembled into transcript sequences before they can be utilized. Despite recent advances in assembly software, creating such transcripts from read data harboring isoform variation remains challenging. This is because current approaches fail to identify all variants present or they create chimeric transcripts within which relationships between co-evolving sites and other evolutionary factors are disrupted. We present VTBuilder, a tool for constructing non-chimeric transcripts from read data that has been sequenced from sources containing isoform complexity.

Results: We validated VTBuilder using reads simulated from 54 Sanger sequenced transcripts (SSTs) expressed in the venom gland of the saw scaled viper, Echis ocellatus. The SSTs were selected to represent genes from major co-expressed toxin groups known to harbor isoform variants. From the simulated reads, VTBuilder constructed 55 transcripts, 50 of which had a greater than 99% sequence similarity to 48 of the SSTs. In contrast, using the popular assembler tool Trinity (r2013-02-25), only 14 transcripts were constructed with a similar level of sequence identity to just 11 SSTs. Furthermore VTBuilder produced transcripts with a similar length distribution to the SSTs while those produced by Trinity were considerably shorter. To demonstrate that our approach can be scaled to real world data we assembled the venom gland transcriptome of the African puff adder Bitis arietans using paired-end reads sequenced on Illumina's MiSeq platform. VTBuilder constructed 1481 transcripts from 5 million reads and, following annotation, all major toxin genes were recovered demonstrating reconstruction of complex underlying sequence and isoform diversity.

Conclusion: Unlike other approaches, VTBuilder strives to maintain the relationships between co-evolving sites within the constructed transcripts, and thus increases transcript utility for a wide range of research areas ranging from transcriptomics to phylogenetics and including the monitoring of drug resistant parasite populations. Additionally, improving the quality of transcripts assembled from read data will have an impact on future studies that query these data. VTBuilder has been implemented in java and is available, under the GPL GPU V0.3 license, from http:// http://www.lstmed.ac.uk/vtbuilder .

Show MeSH
Transcript reconstruction on simulated reads. (A) Lengths of all transcripts constructed by VTBuilder and Trinity compared to those of the SSTs. The top and bottom of the boxes represent the 25th and 75th percentiles respectively, while the top and bottom whiskers represent the third quartile +1.5 times the inter quartile range (IQR) and the first quartile - 1.5 times the IQR respectively. Outliers beyond these points are represented as black circles. (B) Lengths of transcripts constructed by VTBuilder and Trinity that had a sequence similarity of 90% or greater to the SSTs. (C) Network showing the relationship between the VTBuilder transcripts and the SSTs. Grey nodes represent the VTBuilder transcripts. Colored nodes represent the protein families to which the individual SSTs belong (see key). Node size is proportional to sequence length. Edges represent a 90% or greater sequence similarity. (D) Same as (C) but using Trinity to construct the transcripts.
© Copyright Policy - open-access
Related In: Results  -  Collection

License 1 - License 2
getmorefigures.php?uid=PMC4260244&req=5

Fig3: Transcript reconstruction on simulated reads. (A) Lengths of all transcripts constructed by VTBuilder and Trinity compared to those of the SSTs. The top and bottom of the boxes represent the 25th and 75th percentiles respectively, while the top and bottom whiskers represent the third quartile +1.5 times the inter quartile range (IQR) and the first quartile - 1.5 times the IQR respectively. Outliers beyond these points are represented as black circles. (B) Lengths of transcripts constructed by VTBuilder and Trinity that had a sequence similarity of 90% or greater to the SSTs. (C) Network showing the relationship between the VTBuilder transcripts and the SSTs. Grey nodes represent the VTBuilder transcripts. Colored nodes represent the protein families to which the individual SSTs belong (see key). Node size is proportional to sequence length. Edges represent a 90% or greater sequence similarity. (D) Same as (C) but using Trinity to construct the transcripts.

Mentions: In brief, 50,000 reads of length 250 bases were copied from the 54 SSTs at random locations. For each read, its pair was copied randomly from a window 500 bases wide anchored on the last base of the read itself. Read coverage across each SST was normalised by length resulting in an upper bound of 1930 reads covering the longest SST and a lower bound of 480 covering the shortest. This is equivalent to an upper per site coverage of 209 and a lower per site cover of 190, typical of the coverage observed in an NGS dataset. Note 50,000 reads is far less than would be expected within an NGS dataset but here the reads are covering far fewer transcripts (54 SSTs) than the thousands of transcripts typically found within a transcriptome. This read/transcript ratio was selected to represent approximately 7 M reads covering a transcriptome of around 7500 genes. VTBuilder, running default parameters (min. read ln. 150; min. transcript ln. 250; min isoform sim. 96%) and on a desktop with 16 cores, 32 gigabytes of RAM and Biolinux 7 (Ubuntu 12.04) [57], was then used to construct transcripts from the simulated paired end reads (see user guide). VTBuilder constructed 55 transcripts of comparable length distribution (ranging from 500 to 2298 bp) to the input SSTs (Figure 3A). Using the same simulated paired end data as input, Trinity (using default parameters) resulted in the construction of many more (112) transcripts that ranged in length from 217 to 2104 bp (Figure 3A).Figure 3


VTBuilder: a tool for the assembly of multi isoform transcriptomes.

Archer J, Whiteley G, Casewell NR, Harrison RA, Wagstaff SC - BMC Bioinformatics (2014)

Transcript reconstruction on simulated reads. (A) Lengths of all transcripts constructed by VTBuilder and Trinity compared to those of the SSTs. The top and bottom of the boxes represent the 25th and 75th percentiles respectively, while the top and bottom whiskers represent the third quartile +1.5 times the inter quartile range (IQR) and the first quartile - 1.5 times the IQR respectively. Outliers beyond these points are represented as black circles. (B) Lengths of transcripts constructed by VTBuilder and Trinity that had a sequence similarity of 90% or greater to the SSTs. (C) Network showing the relationship between the VTBuilder transcripts and the SSTs. Grey nodes represent the VTBuilder transcripts. Colored nodes represent the protein families to which the individual SSTs belong (see key). Node size is proportional to sequence length. Edges represent a 90% or greater sequence similarity. (D) Same as (C) but using Trinity to construct the transcripts.
© Copyright Policy - open-access
Related In: Results  -  Collection

License 1 - License 2
Show All Figures
getmorefigures.php?uid=PMC4260244&req=5

Fig3: Transcript reconstruction on simulated reads. (A) Lengths of all transcripts constructed by VTBuilder and Trinity compared to those of the SSTs. The top and bottom of the boxes represent the 25th and 75th percentiles respectively, while the top and bottom whiskers represent the third quartile +1.5 times the inter quartile range (IQR) and the first quartile - 1.5 times the IQR respectively. Outliers beyond these points are represented as black circles. (B) Lengths of transcripts constructed by VTBuilder and Trinity that had a sequence similarity of 90% or greater to the SSTs. (C) Network showing the relationship between the VTBuilder transcripts and the SSTs. Grey nodes represent the VTBuilder transcripts. Colored nodes represent the protein families to which the individual SSTs belong (see key). Node size is proportional to sequence length. Edges represent a 90% or greater sequence similarity. (D) Same as (C) but using Trinity to construct the transcripts.
Mentions: In brief, 50,000 reads of length 250 bases were copied from the 54 SSTs at random locations. For each read, its pair was copied randomly from a window 500 bases wide anchored on the last base of the read itself. Read coverage across each SST was normalised by length resulting in an upper bound of 1930 reads covering the longest SST and a lower bound of 480 covering the shortest. This is equivalent to an upper per site coverage of 209 and a lower per site cover of 190, typical of the coverage observed in an NGS dataset. Note 50,000 reads is far less than would be expected within an NGS dataset but here the reads are covering far fewer transcripts (54 SSTs) than the thousands of transcripts typically found within a transcriptome. This read/transcript ratio was selected to represent approximately 7 M reads covering a transcriptome of around 7500 genes. VTBuilder, running default parameters (min. read ln. 150; min. transcript ln. 250; min isoform sim. 96%) and on a desktop with 16 cores, 32 gigabytes of RAM and Biolinux 7 (Ubuntu 12.04) [57], was then used to construct transcripts from the simulated paired end reads (see user guide). VTBuilder constructed 55 transcripts of comparable length distribution (ranging from 500 to 2298 bp) to the input SSTs (Figure 3A). Using the same simulated paired end data as input, Trinity (using default parameters) resulted in the construction of many more (112) transcripts that ranged in length from 217 to 2104 bp (Figure 3A).Figure 3

Bottom Line: From the simulated reads, VTBuilder constructed 55 transcripts, 50 of which had a greater than 99% sequence similarity to 48 of the SSTs.Unlike other approaches, VTBuilder strives to maintain the relationships between co-evolving sites within the constructed transcripts, and thus increases transcript utility for a wide range of research areas ranging from transcriptomics to phylogenetics and including the monitoring of drug resistant parasite populations.Additionally, improving the quality of transcripts assembled from read data will have an impact on future studies that query these data.

View Article: PubMed Central - PubMed

Affiliation: Department of Parasitology, Liverpool School of Tropical Medicine, Pembroke Place, Liverpool, L3 5QA2, UK. john.archer.jpa@gmail.com.

ABSTRACT

Background: Within many research areas, such as transcriptomics, the millions of short DNA fragments (reads) produced by current sequencing platforms need to be assembled into transcript sequences before they can be utilized. Despite recent advances in assembly software, creating such transcripts from read data harboring isoform variation remains challenging. This is because current approaches fail to identify all variants present or they create chimeric transcripts within which relationships between co-evolving sites and other evolutionary factors are disrupted. We present VTBuilder, a tool for constructing non-chimeric transcripts from read data that has been sequenced from sources containing isoform complexity.

Results: We validated VTBuilder using reads simulated from 54 Sanger sequenced transcripts (SSTs) expressed in the venom gland of the saw scaled viper, Echis ocellatus. The SSTs were selected to represent genes from major co-expressed toxin groups known to harbor isoform variants. From the simulated reads, VTBuilder constructed 55 transcripts, 50 of which had a greater than 99% sequence similarity to 48 of the SSTs. In contrast, using the popular assembler tool Trinity (r2013-02-25), only 14 transcripts were constructed with a similar level of sequence identity to just 11 SSTs. Furthermore VTBuilder produced transcripts with a similar length distribution to the SSTs while those produced by Trinity were considerably shorter. To demonstrate that our approach can be scaled to real world data we assembled the venom gland transcriptome of the African puff adder Bitis arietans using paired-end reads sequenced on Illumina's MiSeq platform. VTBuilder constructed 1481 transcripts from 5 million reads and, following annotation, all major toxin genes were recovered demonstrating reconstruction of complex underlying sequence and isoform diversity.

Conclusion: Unlike other approaches, VTBuilder strives to maintain the relationships between co-evolving sites within the constructed transcripts, and thus increases transcript utility for a wide range of research areas ranging from transcriptomics to phylogenetics and including the monitoring of drug resistant parasite populations. Additionally, improving the quality of transcripts assembled from read data will have an impact on future studies that query these data. VTBuilder has been implemented in java and is available, under the GPL GPU V0.3 license, from http:// http://www.lstmed.ac.uk/vtbuilder .

Show MeSH