Forum:Trimming tools - your experience please?

Question

[paigemiller]

Hello, I just posted a comment about trimming in another thread. Now I would like to ask anyone who has experience (or any desire to share their two cents) to offer some feedback that might help me and other "newbies" dealing with NGS data and the different tools out there. I am relatively new to this and am now dealing with A LOT of data from 3 different projects all at once. I am finding that some tools are just not worth it - but maybe that is just ignorance on my part. I'm copying my comment here and I'm asking what do you think?

ORIGINAL comment added to thread "Trimmomatic: Trimming only custom adapter sequences":

Hello to anyone new to bioinformatics who is reading this to try and solve a problem trimming with trimmomatic. I wish I had read this post a week ago and saved myself many headaches. Please save yourself time and go use trim-galore (which uses cutadapt as part of the process) or just cutadapt (or another alternative like bbtrim, which is supposed to work well too). I did finally get trimmomatic to work (many hours and finally help from an IT person). It was a big waste because it runs VERY slowly (even on a high powered compute system) and it did not give consistent output. A lot of the adapters were not removed when I got the reads and ran them on fastqc afterwards.

I then ran everything using trim-galore (installed with miniconda 3 from the bioconda channel) and got VERY fast and very clean reads that were also trimmed properly for quality and already run through fastqc (you can choose that option easily). I also found out it is easy to add an additional unique adapter sequence to trim out at the same time, and I ran trim-galore on a different dataset (produced by a flavor of radseq) and was able to get rid of the regular illumina adapters and the unique adapter.

Answer 1

[GenoMax]

I am finding that some tools are just not worth it - but maybe that is just ignorance on my part.

That may be the case here. Sure trimmomatic does not have the easiest to understand command line options but the manual is fairly descriptive and does exist. There are plenty of threads describing all things that one can do/go wrong with trimmomatic command line on biostars. If you did not use multiple threads with it then it could have run very slowly, especially if you had a large dataset.

These things are part of learning and you will get better with time. It is normal to feel overwhelmed, if you are new to data analysis.

bbduk from BBTools suite (GUIDE), fastp (LINK) are performant options besides cutadapt that you have already discovered.

Answer 2

[Mensur Dlakic]

In my experience most tools do a good job here, so I don't think you will go wrong no matter what you pick. It is a matter of either going with heavyweights such as trimmomatic that may have lots of exotic options, or with simpler and potentially faster tools. Keep in mind that in most cases you will get reads that are already trimmed unless you are doing sequencing on your own.

Couple of tools that were not mentioned before:

• https://github.com/MikkelSchubert/adapterremoval
• https://github.com/cihga39871/Atria

Answer 3

[i.sudbery]

I do tend to find that getting Trimmomatic to remove adaptor sequences efficiently takes some tuning. In that sense, I prefer cutadapt to remove adaptors, which seems to be much more eager to remove sequences out of the box than trimmomatic without optimising alignment thresholds, and specifying the sequences is mouch more transparent.

However, I do mostly use Trimmomatic when my aim is to trim or remove reads with low quality base calls.