error correction on single-end reads --necessary?

Question

[st.ph.n]

Hello BioStars users,

I have a bunch of samples of SE 80bp Illumina reads. We've assembled a reference sequence for our organism, and we want to align these reads to the reference. I used Trimmomatic to trim the sequences and look for adapters, and an average of 99.55% of the reads were remaining for each sample. Is it necessary to perform error correction on these short reads for alignment? The aligner I plan to use is bowtie.

Thanks!

Answer 1

[Devon Ryan]

There's generally no need to perform error correction for the alignment phase (unlike assembly). Bowtie (or bowtie2) should typically have no problem aligning a read to the same place regardless of whether it's been error corrected or not.

[st.ph.n]

That's what I figured. We have a robust QC pipeline that includes error correction for paired end reads. This is the first time we did SE sequencing. I was thinking Trimmomatic would be enough, but wanted a second opinion.

[lh3]

If you intend to find variants, it is better NOT to correct errors at all. Error correctors may introduce new errors or correct true variants away. Also, you should not use bowtie1 for >=80bp reads these days. Use bowtie2 instead.