When and why should you trim adapters from sequencing reads?

Why

Every library fragment is flanked by adapters. If the insert is longer than the read, the read never reaches the adapter. If the insert is shorter, the read runs past it into adapter sequence.

Illumina's guidance says adapter sequence can appear at the 3' end of reads when read length exceeds insert size, and lists the sequences to trim for each kit, such as AGATCGGAAGAGC at the start of TruSeq adapters and CTGTCTCTTATACACATCT for many tagmentation kits. Short inserts are common in small RNA libraries, degraded samples and cell-free DNA, so trimming matters most there. Tools such as cutadapt find and remove the adapter, can trim low-quality 3' ends, and can drop reads that become too short.

For two-color instruments such as NextSeq, cutadapt notes that a read falling off the end of its fragment shows a run of high-quality but wrong G calls, which ordinary quality trimming misses, and offers a dedicated option. Trimming is not always needed. Many modern aligners soft-clip unmatched ends, and Illumina's FASTQ generation can trim adapters itself, so check what was already done before trimming twice.

How to work it through

1. Check the library kit and expected insert size against the read length.
2. Run a QC report and look at adapter content along the read.
3. Look up the adapter sequence for the kit from the vendor's list.
4. Trim with a tool such as cutadapt, using paired mode for paired reads.
5. Discard reads shorter than a minimum length that can still map uniquely.
6. Rerun QC to confirm adapter content has gone and record the settings.

Where the answer stops holding

Aggressive quality trimming can shorten reads enough to make them map less uniquely. Trimming with the wrong adapter sequence removes nothing useful. Some pipelines, such as those relying on fixed read lengths, can be affected by variable read lengths after trimming.