Trim off variable length 'N' strings at the end of the read

[illinu]

Hi,

I need to remove all 'N' strings in a fastq file. I have paired end files and there are N strings at the end of some reads of variable length (both the reads and the N strings are of variable length). I can't find any tool to do this. Trimmomatic will remove bases based on their quality score. Fastx_trimmer will keep 'x' first bases.

Anyone has a script for this or knows of a tool? It is important that the tool deals with paired files and keeps the pairs 'alive' after trimming in both files.

ps: I tried to install nesoni but I am uncapable to do this in the server without root permisions and an older python version.

Thanx
Illinu

[dpryan]

Aren't the quality scores of the Ns very low? Trimming by quality will normally remove stretches of them (at least unless they're in the middle, which happens).

[Brian Bushnell]

As dpryan said, quality-trimming should suffice; Ns should have a quality of zero, so you can just set the quality-trim threshold at 1. For example, with BBTools:

reformat.sh in1=read1.fq in2=read2.fq out1=trimmed1.fq out2=trimmed2.fq qtrim=rl trimq=1

That will keep the pairs together. That program will also automatically convert the quality of Ns to zero, if they happen to be non-zero.

[SES]

You can trim poly-Ns with PRINSEQ. There are (at least) three options to control the trimming, one to specify the minimum length of Ns at the 3-prime end, another option to specify the maximum N percentage to allow, and one option to specify the max number of Ns to allow.

-trim_ns_right <integer>
            Trim poly-N tail with a minimum length of trim_ns_right at the
            3'-end.

-ns_max_p <integer>
            Filter sequence with more than ns_max_p percentage of Ns.

-ns_max_n <integer>
            Filter sequence with more than ns_max_n Ns.

[illinu]

SES, thank you for this tip I think it's the approach I was looking for.
To answer the previous posts, I checked precisely this and surprisingly the scores are high. I don't understand why but I was expecting them to be null if nothing.

[illinu]

You can trim poly-Ns with PRINSEQ. There are (at least) three options to control the trimming, one to specify the minimum length of Ns at the 3-prime end, another option to specify the maximum N percentage to allow, and one option to specify the max number of Ns to allow.

-trim_ns_right <integer>
            Trim poly-N tail with a minimum length of trim_ns_right at the
            3'-end.

-ns_max_p <integer>
            Filter sequence with more than ns_max_p percentage of Ns.

-ns_max_n <integer>
            Filter sequence with more than ns_max_n Ns.

SES, I am thinking now... this will not handle paired files right?

[Brian Bushnell]

To answer the previous posts, I checked precisely this and surprisingly the scores are high. I don't understand why but I was expecting them to be null if nothing.

But like I said, BBTools will automatically change the quality of N bases to 0, because it makes no sense for them to have any other quality. So they will be trimmed anyway.

[SES]

SES, I am thinking now... this will not handle paired files right?

Yes, I believe recent versions of PRINSEQ will handle paired-end files correctly. If you run into issues, you could use Pairfq to re-pair your reads and separate the singletons after trimming.

[illinu]

I tried the bbmap option and it worked beautifully! Only 5 minutes wallclock time. The programs needs no installation it runs with java.