Two Fastq files from a same patient with 2 technologies : When combine ?

[aangs94]

Hi everyone,

I have datas from a patient sequenced two years ago and now I have datas from the SAME patient re-sequenced last month with a different technology.
Both are exomes.

Maybe some people had the same case, but if I want to do a combine analysis (SNP calling), when do I have to combine the two files ?
1/ concatenate the FASTQ files and then proceed to analysis of the unique file
OR
2/ combine each BAM files (cleaned or not) for examples or any further files?

If somebody had already done this kind of analysis, or maybe any publication with such examples...It will help me a lot.

Thanks very much to the NGS community.

[swbarnes2]

I don't think combining the fastqs is a good idea. Two different technologies will have there own error profiles, I don't think analysis software is designed to work with that. For instance, Illumina is not prone to homopolymer errors, but 454 is. So if you combined the two together, you might think you have a mixed indel in a homopolymer region, where examination of the Illumina data alone would show you that the 454 was causing that artifact. I think the right thing to do is to use whatever SNP calling program is best for each technology, then compare your results after that point. So don't merge fasts or bams, but vcfs.

[aangs94]

re

an

I don't think combining the fastqs is a good idea. Two different technologies will have there own error profiles, I don't think analysis software is designed to work with that. For instance, Illumina is not prone to homopolymer errors, but 454 is. So if you combined the two together, you might think you have a mixed indel in a homopolymer region, where examination of the Illumina data alone would show you that the 454 was causing that artifact. I think the right thing to do is to use whatever SNP calling program is best for each technology, then compare your results after that point. So don't merge fasts or bams, but vcfs.

Thanks very much swbarnes2 for your clear answer.

I think you are right. If I can ask you one more thing :
what if the same sample is sequenced two times with the same technology but in separate dates (for example one year difference)?, is it ok if I merge the fastq files or the bam files (like a biological or technical replicate)? I imagine it could improve the depth or coverage?

[swbarnes2]

Assess the overall quality of the two datasets. Looks at % of base >=Q30, something like that. Do they look like they are in the ballpark of being equally good quality?

Then I'd align them separately, and look at the flagstat figures, or something like that. Do the alignments look like they are about the same quality?

If that's the case, merge the .bams, and call SNPs from that. That'll be better than calling SNPs on them apart.

[aangs94]

Thanks very much swbarnes2, very efficient!!

Ok, so if the two fastq files are equivalent in term of quality, I start the analysis pipeline until the flagstat figure. Then merge the .bams.
Do you think that merging cleaned or unclead bams may have an impact at this step ?

Tks again