trimming or filtering first?

[Seraphya]

Does the order of trimming reads first and then filterings or first filtering reads and then trimming make a diffrence?

[dpryan]

To copy and paste my comment on Biostars:

Do you mean, "Does the order of adaptor and base quality trimming matter?" Otherwise, which sort of trimming and filtering do you have in mind?

[gege_55]

I did trimming of bad quality bases first. I think you can retain more of your reads like this, because after that step more of your reads will pass the filtering step.
Maybe you could try both ways with a subset of reads and see which one yields more reads after filtering.

[Seraphya]

The subset test is a good idea.
I am actually wondering what the purpose of filtering is if you are trimming reads down and discard them when there are not enough bases left.

[ctseto]

Not sure myself. Generally, I've done

Filtering of contaminant genomes (YMMV), then Trimmomatic with the appropriate adaptors. For Trimmomatic, I'm not quite sure if they trim adaptors first then do quality trim, or vice versa.

That said...what did you mean by filtering?

[dpryan]

The subset test is a good idea.
I am actually wondering what the purpose of filtering is if you are trimming reads down and discard them when there are not enough bases left.

I don't bother with filtering after trimming. The quality dip is almost always just near the ends of the reads, so there's likely no benefit unless you're doing a homopolymer filter or something like that. Even then, unless you have a lot of reads that would otherwise be filtered out, the change in alignment time is likely about as long as the amount of time needed for the filtering.

[gege_55]

It depends on what you want to do with your reads. For mappings I wouldn't filter too strictly. For assemblies you want the best quality reads possible.
In the case of an assembly I remove reads with ambiguous bases first. Assemblers don't handle them well. Then you trimm and then you filter. I try to estimate how many reads I need in the end for a decent assembly. There are some numbers here in the forum for a few species. Then I iterate the filter criteria with a subset in a way to approximately reach that number.
The more reads you have to begin with the more you can filter out resulting in higher quality of the remaining.