paired end reads cutadapt

Question

[anna]

I have several paired end reads files. After performing the FastQC analysis i found out that some pairs have one or more overrepresented sequences. Should i trimm these sequences from both of the two files (1_1 and 1_2)? or just in the only one that have them overrepresented (1_2)? This is an example:

1. File 1_1.fq: No Overrepresented sequences
2. File 1_2.fq AAGCAGTGGTATCAACGCAGAGTACTTTTTTTTTTTTTTTTTTTTTTTTT TTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTT

Then I'm considering the command line as follows:

cutadapt -b AAGCAGTGGTATCAACGCAGAGTACTTTTTTTTTTTTTTTTTTTTTTTTT \
         -b TTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTT \
         -B AAGCAGTGGTATCAACGCAGAGTACTTTTTTTTTTTTTTTTTTTTTTTTT \
         -B TTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTT \
         -o out_1.fastq \
         -p out_2.fq \
         1_1.fastq 1_2.fq

[GenoMax]

Please use the formatting bar (especially the code option) to present your post better. It is difficult to see what you are presenting.

Thank you!

[Ram]

I formatted OP's code this time. I think we should add splitting up long one-liners to the (upcoming) formatting manual.

[ATpoint]

How many reads are affected (%)?

Answer 1

[Kevin Blighe]

There have been a few threads on this topic already:

• How to remove poly T in RNA-sequencing data
• Statistics About Poly-A Tails In Rnaseq Reads
• https://www.researchgate.net/post/Is_it_necessary_to_remove_over_represented_sequences_in_the_NGS_data
• Fastqc Over-Represented Sequences Are Adapters ?

In conclusion, I would just remove the standard adapters that are known to CutAdapt from the sequences, and also filter / trim reads based on length and quality, and then proceed to alignment. My feeling is that the main thing that is affected by trimming an filtering reads is the quality metrics like percent alignment. Most 'junk' reads, including poly A and T, will not align anyway.

Kevin

[anna]

Thanks for your answer. I will consider removing the overrepresented sequences and compare the "clean" data against the raw data, which already have a very good quality of the reads.