Adapter trimming: how can permanent adapter trimming, or hard clipping, be achieved when following GATK Best Practices and commencing data preprocessing with uBAM files?
Dear GATK Community,
I understand there has already been a post regarding adapter trimming previously (Do I need to perform fastqc and adapters trimming before gatk pipeline?). However, many tools only appear to accept FASTQ files as input for adapter trimming rather than other formats, such as uBAM.
Therefore, I was hoping for some help with the following questions:
1) How can permanent adapter trimming be approached if one is following GATK Best Practices with uBAM as the initial file format?
I understand one can hard clip adapter sequences with SamToFastq prior to alignment. However, following alignment, all hard clips are transformed to soft clips with MergeBamAlignment and therefore, the adapter sequences would still be present in downstream data preprocessing and variant discovery.
2) If it is not possible to generate a clean, mapped BAM with hard-clipped adapters ((How to) Map and clean up short read sequence data efficiently), what effect would adapter sequences in the merged BAM have on downstream data preprocessing and variant discovery steps, such as those steps mentioned in Data pre-processing for variant discovery and Somatic short variant discovery (SNVs + Indels)?
Thank you for your time and help.
Kind regards.
-
This was one of my earlier questions back in the old forum and my short takeaway message from the response was that it depends to your personal taste of data integrity. I used to mark adapter sequences with base quality of 2 to keep them away from any potential interference however my later practices told me to completely remove all adapters and make sure that fastqc graph is a flatline.
If you wish to continue using MarkIlluminaAdapters and SamToFastq practice make sure that you remove or mark adapters and lower their base quality scores and merge the uBAM generated using the modified fastq files not the original unmodified uBAM before marking stage.
In general adapter cleanup is a necessary step no matter how you handle your data so don't hesitate to modify the best practices flows according to your needs when you see something does not fit your needs.
-
Thank you for your reply and advice SkyWarrior. I have read information in multiple resources and legacy forum posts and understand there is not a straightforward answer. Could I ask what it was in your later practices that led you to completely remove the adapters rather than using a base quality of 2?
-
Sure. When our data was completely from a single source it was a doable option. However after we acquired our own sequencer and started accumulating data from multiple resources with bunch of different practices to produce fastq files I decided to change my behavior and completely started removing all traces of adapters from the data.
-
Thank you for the additional information SkyWarrior.
I have previously questioned why permanent hard clipping is not a supported option within GATK workflows as there are many external recommendations and resources for quality trimming and adapter removal. I understand there is the tool ClipReads but hard clipping is not supported.
As it may help others in the future, I though I would collate reasoning for this based on my current understanding:
1) With quality trimming, this is discouraged as GATK tools are quality-aware and hard clipping based on quality scores could affect BQSR (CollectBaseDistributionByCycle (Picard)), part of GATK Best Practices (Data pre-processing for variant discovery), and other quality control assessments. Therefore, quality trimming is not recommended.
2) With adapter trimming, only soft clipping is available post-MergeBamAlignment (Picard). In current and legacy documentation, it is described how base quality scores marked by XT can be set to 2 to prevent adapter sequences from contributing to alignment (awaiting confirmation - see separate post), how BQSR will ignore the associated bases and that they will not contribute towards HaplotypeCaller variant calls. Looking at current default settings (GATK version 4.2.0.0) for BQSR, HaplotypeCaller and also Mutect2, this continues to be the case as long as the base qualities for adapters are maintained at 2 following MergeBamAlignment (Picard) (awaiting confirmation - see separate post).
A couple of cautions:
1) If one is analysing their data with FastQC, the FastQC Report will not necessarily reflect the above clipping settings and considerations. Therefore, it will most likely show the presence of adapters in the Adapter Content section even though these are accounted for.
2) As SkyWarrior has mentioned, a new uBAM could be generated from FASTQ files with adapters hard clipped or base quality scores permanently changed. However, during the original uBAM -> FASTQ -> new uBAM process, certain meta information and tags will be lost and not easily added later to the new uBAM (e.g. RX tags for molecular indexes). Therefore, this must be considered depending on planned downstream analyses.
I hope this is helpful and I look forward to hearing any further thoughts from the GATK Community.
Please sign in to leave a comment.
4 comments