qbonenfant
commented
11 months ago Hi @Rohit-Satyam, I am the main developper of Porechop_ABI, @rchikhi and @chadisaad are not working on this project (afaik). If you want to contact someone involved in the project, and currently working in the Bonsai team, i invite you to check the corresponding author of our paper.
Short answer: just like the original Porechop, Porechop_abi should not trimm your dataset if no adapters are present. If -abi find something, you may want to align it against the reference genome of the sequenced organism.
But as always, it is not that simple.
Long Answer:
RabbitQCplus:
From what I read from the supplementary materials of the RabbitQCplus paper, and line 185 in main.cpp on their master branch, I think the trimming part is implemented for NGS datasets only.
fprintf(stderr, "WARNING : the TGS module does not support trimming and will not produce output files, so ignore the -o and -O parameter.\n");
This may explain why you do not see any adapter output. Even if they did, the current adapter strategy they use may not be applicable for long read anyway. They use a simple k-mer frequency based seed-and-extend strategy, which may not work very well on noisy reads such as ONT.
Our ABI core module uses a similar method, but with more complex (thus slower) algorithms, data structures and post processing ( 01*0 seeds, mixed assembly methods, frequency-based termination detection, consensus over multiple run, bi-directionnal debruijn graphs...). Since RabbitQCplus focus seems to be on speed and efficiency, this may not be of interest for the authors.
On a side note, I am not convinced fastp really can trimm ONT adapters either, and seeing the number of opened issues referring to adapters, i have doubts about it being reliable for this task.
Excessive Trimming Despite all of our efforts, there is one case where Porechop_ABI may want to trimm sequence with no adapters: over represented sequences. This may append if your dataset contains a lot of highly expressed RNA (ex: ribsomal RNA, constituve proteins such as actine, tubuline, or plant storage proteins like zein_alpha, etc...), or very short genomes (virus, plasmids) that can be fully covered by a single read.
What should you do ? In the case of Porechop_ABI, i would suggest you align the inferred adapter against the reference genome of the sequenced organism, or the NR database of the NCBI. It may help you to check the quality of your adapter. A good example of a (to be confirmed) great case of over-represented sequence interferring with the adapter end detection can be seen in my reply on issue #9.
Is there a reliable method for trimming ? I have been out of the academic world for almost a year now, and may be wrong about this: To this day, I am not sure there is a 100% reliable method to detect the presence of adapter sequence in ONT dataset, at least not until ONT will pubicly release all adapter sequences. Despite their effort to integrate trimming in their basecalling pipeline, ONT let the trimming part to an RNN, which is just a blackbox for end users. Dedicated QC and post processing tools are still required to ensure good data quality.
One of the tool i saw the most in ONT read processing pipelines is Porechop, on which we based our tool. Despite its quirk, the fact it was never published, and is even unsupported since 2018, Porechop is still in use today.
I won't make an exhaustive list of ONT trimming tools, but i would suggest picking one that match the goal you fixed for your pipeline (data quality, speed, ...), as long as you are transparent on what pre-processing you performed in your publications. In the end, this is what really matters.
Rohit-Satyam
Hi @rchikhi @chadisaad
I am planning to use Porechop_abi in building an automated pipeline. However, I have a concern to share. Does Porechop_abi performs unnecessary trimming even in absence of adapters. I am asking this because I ran RabitQCPlus on my fastq files and it reported no adapter but then when I run porechop_abi using
-abiargument, it finds some adapters ab-initio and removes a lot of reads. Now, I was thinking porechop_abi likefastpwhich performs trimming only if adapter presence is observed. I am still new with handling of MinIon data and I don't know a reliable method to detect the presence of adapters.