fake paired end data from titanium shotgun reads?
Hello All -
Has anyone tried to fake paired end data from GS-FLX/Titanium shotgun reads?
Briefly, we haven't implemented paired end sequencing yet and don't have reliable reference genomes for some of our isolates (bacteria with genomes <6 Mb). Metrics indicate that we're getting good coverage (30X+) but our de novo assemblies contain many more contigs than they should. A key reason for this is gsAssembler - and Newbler's annoying habit of splitting sequencing reads when it encounters multicopy elements (e.g transposons, rRNA operons, even genes with repetitive domains). We also see a lot of artificially short (<50 bp) contigs. By manually sifting through the '454ContigGraphs' data and/or running the 'split' reads through a 3rd party assembler, it's possible to properly order and join adjacent contigs (most of which are separated by a gap of 0 bp), but our current process is laborious. My hope is that, by faking paired end data, Newbler will stop splitting reads, or at least produce fewer contigs that have 0 bp gaps between them.
To convert shotgun reads to fake pairs, which of the following might be the best approach?
1) Use all Assembled/PartiallyAssembled reads - or just those that are 'split' (e.g. ReadStatus file indicates that 5' and 3' ends of the read are in different contigs)
2) Use the entire read - or split the reads in half? For the former, I'm thinking of adding the paired end 'A primer' sequence to the 5' end of the reads AND generating a reverse complemented set of the same reads and adding the paired end 'B primer' sequence to the 5' end of those reads. For the latter, I'd follow a similar approach, but use the first 200 bp of the read with the A primer and the last 200 bp with the B primer - and perhaps even bridge the two with the circularization primer sequence (to make the sequence look like a proper paired end read).
Any comments, concerns, experience with such an approach? Do scripts already exist?
Can anyone provide paired end primer sequences?
Cheers -
David
Has anyone tried to fake paired end data from GS-FLX/Titanium shotgun reads?
Briefly, we haven't implemented paired end sequencing yet and don't have reliable reference genomes for some of our isolates (bacteria with genomes <6 Mb). Metrics indicate that we're getting good coverage (30X+) but our de novo assemblies contain many more contigs than they should. A key reason for this is gsAssembler - and Newbler's annoying habit of splitting sequencing reads when it encounters multicopy elements (e.g transposons, rRNA operons, even genes with repetitive domains). We also see a lot of artificially short (<50 bp) contigs. By manually sifting through the '454ContigGraphs' data and/or running the 'split' reads through a 3rd party assembler, it's possible to properly order and join adjacent contigs (most of which are separated by a gap of 0 bp), but our current process is laborious. My hope is that, by faking paired end data, Newbler will stop splitting reads, or at least produce fewer contigs that have 0 bp gaps between them.
To convert shotgun reads to fake pairs, which of the following might be the best approach?
1) Use all Assembled/PartiallyAssembled reads - or just those that are 'split' (e.g. ReadStatus file indicates that 5' and 3' ends of the read are in different contigs)
2) Use the entire read - or split the reads in half? For the former, I'm thinking of adding the paired end 'A primer' sequence to the 5' end of the reads AND generating a reverse complemented set of the same reads and adding the paired end 'B primer' sequence to the 5' end of those reads. For the latter, I'd follow a similar approach, but use the first 200 bp of the read with the A primer and the last 200 bp with the B primer - and perhaps even bridge the two with the circularization primer sequence (to make the sequence look like a proper paired end read).
Any comments, concerns, experience with such an approach? Do scripts already exist?
Can anyone provide paired end primer sequences?
Cheers -
David