Thanks for your quick answer ;)
• 0 views
•
link
Ngs: Filtering Out Reads That Have A Hit...
I am going to be working with a lot of Illumina next generation sequencing data, but I will be doing something I suspect is rather unusual: Within the data, there will be a vast majority of sequences that will map fairly well to my reference genome. However, those are not the sequences I am interested in: I want to retrieve all the reads that do not map well to my reference genome.
Update: To clarify, within my dataset I have a vast majority of the reads that will match to my reference genome. However, a small portion of the dataset does (should) not match at all (i.e. I also exclude SNPs, short indels, etc.). I want to reduce my dataset to those reads that do not correspond to the conserved parts of the genome in order to do de novo assembly on them, without the overhead of assembling the whole genome from scratch. End of Update.
- Is this something commonly done?
- Are the usual tools capable of doing that for me or will I need to implement something myself?
- Any suggestions on the best method(s) to proceed?
• 4,395 views
•
link
4 answers
No problem, glad I could help.
• 0 views
•
link
Normally for most of the aligners there is some option where you can output the unaligned reads into a different file. Otherwise look for the reads that were flagged "4" as Leonor suggested. In case of paired end reads "5", "13" and some other flags may apply too. (See http://picard.sourceforge.net/explain-flags.html) to select the read pairs or read pair that is unaligned.
• 0 views
•
link
Thanks: I'll look for that option to output the unaligned reads!
• 0 views
•
link
I would think that looking at the differences from the reference genome is what most of us do. It is the more usual thing to do !
• 0 views
•
link
Sorry, but what do you mean? You'd be looking at all the reads that did not map to your reference usually?
• 0 views
•
link
Like Arun, I'm not sure to understand what you mean either... I think most people look for mapped reads that contain small differences. That is not what I am doing. I need to remove from my dataset all the reads that align to the reference (including SNPs, etc.) before I continue with my downstream analysis.
• 0 views
•
link
With bwa, you can parse by the NM tag. NM:A:0 means that the read mapped perfectly. NM:A:3, for instnace, would mean three mismatches.
• 0 views
•
link
Log in to answer this question.
Just to be sure, when you bitwise AND with "4", you'll get those reads that are unmapped. If you are interested in reads that
do not map WELL to your reference, they could also mean reads that areNOT uniquely mapped. Usually, softwares indicate this with an optional flag. In case of BWA, it would beXT:A:and it would beXT:A:Uin case of uniquely mapped and you would be interested in all other values for that. In case of tophat/bowtie, it would beNH:i:optional flag, whereNH:i:1denotes a read with exactly 1 hit. So you would be interested in all other numbers forNH:i:.Thanks for the complement of information. I'll clarify in the question what I'm looking for!