Chimeric reads in Circular viral DNA

[ParthavJailwala]

Hi All,

I am a newbie in analysis of NGS data, so here is an issue I found while using Newbler (gsMapper) to map reads from a viral sequencing experiment. In the Newbler output reports, I found a high number of reads reported as chimeric, in only one MID group (one section of the genome). When I aligned some of these chimeric reads to the reference genome, it did show that part of the read aligned to the end of the genome and the other part to the start of the genome. We now understand that this occurs because the DNA is circular.

My question is, is there any tool to detect and split these chimeric reads from 454? Is there any Newbler option to specify a circular reference DNA, so that the algorithm automatically splits such reads and includes them in the mapping process?

I will appreciate any thoughts, ideas on this.

Thanks

[flxlex]

Newbler, as far as I know, can't be told to treat a reference as circular.

There is no ready-to-go program that I know of that will do what you want. I can only recommend what my approach would be: ask for the 454PairAlign.txt output file (adding the -pair flag to the command line). This file will list all pairwise alignments (potentially a big file). Extracting the chimeric reads should tell you the split points (chimeric reads should show two alignments to the reference). An alternative could be the tab-separated 454PairAlign.txt file (adding the -pairt flag), perhaps easier to parse...

[ParthavJailwala]

Hi flexlex,

I appreciate your response. I will try parsing the 454PairAlign.txt file. However, extracting the split reads manually will still not allow these reads to be included in the mapping/alignment step; newbler just ignores chimeric reads from being included in the mapping process; hence the SNP calling also ignores chimeric reads.
I am hoping the authors of Newbler software provide the option of specifying a circular DNA and also automatically split chimeric reads to include the correctly aligned split-read into the mapping process and the downstream SNP calling step.

Again, thanks for your input.

[nickloman]

I'm a bit surprised, I thought gsMapper would break your reads into two and align them at different positions, but I could well be wrong.

A pragmatic solution might be to make a copy of your reference sequence and change the start position to somewhere in the middle. Then perform a mapping run against the original and a mapping run against the new sequence. You could compare these side-by-side to get the full picture.

Another idea might be to take 500 bases from the end of the sequence and prepend it to the start of the sequence. That should mean more reads can map.

Alternatively, try a different mapping programme. I would suggest LASTZ for 454.

http://www.bx.psu.edu/miller_lab/dist/README.lastz-1.01.50/README.lastz-1.01.50.html#syntax

[ParthavJailwala]

Hi Nickloman,

Thanks for your tips on this. I will definitely try out LASTZ.

Meanwhile, I like your idea of moving the start for the reference sequence. In fact, it should work well as in our targeted resequencing data, only specific regions in the start and the end of the viral genome have been sequenced, leaving a large chunk in the middle where no reads align. So moving the start to somewhere in this unsequenced region should not create a situation where one read has to be split into two.
I will try out a few different ways and post a feedback.

Thanks