Hello,
I am trying to align my raw sequencing data using bowtie2. This is for ATAC-seq analysis. What are the best commands settings to use on bowtie for that? I know I want paired-end alignments and local sequencing, but I am not sure what other options are good for ATAC-seq, such not including discordant pairs, limitations to concordant pairs (e.g. orientation, distance etc.), or anything else! Any advice is appreciated!
Thanks,
•
written
by
mrs.whirly •
1 answer
Leaving everything at default is just fine expect that I find --very-sensitive and -X 2000 helpful. --very-sensitive simply triggers a more sensitive alignment mode and -X 2000 will flag reads with TLEN up to 2kb as concordant, still I doubt this will have a notable effect compared to the defaults which are always a good starting point.
Log in to answer this question.
-
How to use pre-alignments by bowtie2
written by Aki •
-
ATAC seq enhancer help
written by mrs.whirly •
-
Relaxed alignment using Bowtie
written by seqalign •
-
bowtie (1) 0% alignment on paired-end RNAseq data
written by gulamaltab •
-
BWA vs. bowtie
written by blur
-
Optimization of RNA-Seq data mapping with tophat2
written by Denis
-
Chip-seq mapping (using bowtie2 & tophat2) What's different ?
written by leenaehyeon •
-
99.9% are discordant alignments, please help
written by dixit.anirudha •
-
Comparing two ChIP-seq libraries
written by ritarebollo
-
Orientation in paired-end sequencing?
written by John Smith
How about leaving everything else as the default and then just taking a look at the results. As a rule of thumb, it's best to start with more-or-less default settings on tools and then tweak settings further only after having a look at the results.
This guideline also add
-k 10but gives no reason. I guess this may relative to some very short fragments?The author of this page you link (John Gaspar) developed a peak caller (Genrich) for ATAC-seq that can use multimappers, I guess therefore they allow for multimappers during alignment. I personally do not do that as it has not formally been shown that this is robust. I guess if you publish this you would need to robustly show that the way the peak caller uses the multimappers is indeed meaningful and does not create flawed results due to the inherent uncertainty of multimappers.
I second this. I have been trying to find a proper justification for using this parameter but I did not come to anything conclusive. They indeed use multimappers in their methods: https://github.com/jsh58/Genrich#multimap