De novo transcriptome assembly and SNP study with redundant PE reads

[mcastro]

Hi all,

I just started to work with NGS data and I have lots of questions/problems about this so I decided to ask for some advice.

I want to analyze SNP and indels of an organism without reference genome. My idea is to assemble de novo the transcriptome with Velvet/oases, use it with bowtie/tophat to map the reads and call the SNP with samtools. Do you think it's a valid protocol for my purpose?

My main doubt is about data preprocessing. I work with Illumina PE reads. After quality filtering and trimming I have two files of 24M 73bp reads each and with a 75% duplication level. I thought best option was to remove duplicated reads of each file but I'm having problems with this. I used some tools as fastx-collapser or other scripts available in the web but are not prepared to deal with PE data. Any suggestion to remove duplicated reads in PE or it's not necessary?

Thanks a lot for your help

[severin]

R bioconductor

You could do what you are asking using R and bioconductor ('ShortRead') package using indexes to keep track of the paired ends. You will need enough memory to store the reads.

Though you are also likely to run into the problem of differentiating the difference between a SNP and a recently duplicated gene in the same family and sequencing error, but that is true about all transcriptome assemblies.