GPTomics/bio-read-qc-adapter-trimming

Version Compatibility

Reference examples tested with: Cutadapt 4.4+, Trimmomatic 0.39+, fastp 0.23+, FastQC 0.12+

Before using code patterns, verify installed versions match. If versions differ:

• CLI: <tool> --version then <tool> --help to confirm flags

If code throws ImportError, AttributeError, or TypeError, introspect the installed package and adapt the example to match the actual API rather than retrying.

Adapter Trimming -- adapter content IS the insert-size distribution

Remove adapter sequence that the polymerase read INTO once it ran off the end of a short insert, using Cutadapt (precise, the correctness reference) or Trimmomatic (palindrome mode for paired read-through).

"Trim adapters from my reads" -> Detect and remove 3' adapter introduced by read-through, then length-filter the survivors.

• CLI: cutadapt -a AGATCGGAAGAGC -A AGATCGGAAGAGC -m 20 -o R1.fq -p R2.fq in_R1.fq in_R2.fq
• All-in-one alternative: fastp (PE overlap analysis needs no adapter sequence) -> read-qc/fastp-workflow

Scope: this skill OWNS adapter/primer removal. Quality and length filtering -> read-qc/quality-filtering. Single-pass trim+QC -> read-qc/fastp-workflow. Contaminant/PhiX k-mer removal -> read-qc/contamination-screening. OUT OF SCOPE: quality-score trimming as a standalone goal (usually unnecessary before soft-clipping aligners; see insight 2).

The Single Most Important Modern Insight

1. Adapter appears only when the insert is shorter than the read, so adapter content is a direct readout of the insert-size distribution -- and adapter trimming is 3'-only for standard Illumina. The library is [P5]-[insert]-[P7]; a read primes at the insert boundary and reads 5'->3' into the insert, running into the 3'/P7-side adapter only if it runs out of insert. Short-insert libraries (small-RNA ~22 nt, cfDNA ~167 bp, FFPE, degraded RNA, ancient DNA) are read-through-dominated; long-insert WGS may show almost none. The FastQC adapter-content curve climbing toward the 3' end IS that insert-size signal.

2. Adapter trimming is the one near-universal preprocessing step; quality trimming usually is not. Local aligners (BWA-MEM, STAR, Bowtie2 local, HISAT2) SOFT-CLIP low-quality tails, so quality trimming is redundant or harmful for alignment-based DNA/RNA (MacManes 2014, Williams 2016; GATK discourages it before BQSR). But aligners do NOT reliably remove ADAPTER -- adapter is foreign sequence with genuine base quality, so the aligner may try to align it and anchor a wrong placement. Trim adapter; leave quality trimming to the cases that need it (assembly, k-mer/pseudo-alignment, small-RNA, amplicon, no-BQSR variant calling).

3. Small-RNA inverts the logic: the adapter is on EVERY read, so DISCARD reads with no adapter. A ~22 nt miRNA insert is far shorter than a 50-75 nt read, so read-through is universal; a read with no detectable adapter is an adapter dimer, a too-long contaminant, or junk. Use --discard-untrimmed plus a tight length gate (-m 18 -M 30). This is the OPPOSITE of genomic DNA, where the no-adapter reads are the good full-length inserts.

Two-color trap: on NextSeq/NovaSeq, a high-quality poly-G tail is NOT adapter and is not removed by adapter trimming -- it needs a chemistry-aware poly-G trim (cutadapt --nextseq-trim=20, or fastp's auto poly-G). See read-qc/quality-reports and read-qc/fastp-workflow.

Verified Adapter Sequences

| Kit | Read | Sequence | |-----|------|----------| | Illumina TruSeq | R1 3' | AGATCGGAAGAGCACACGTCTGAACTCCAGTCA | | Illumina TruSeq | R2 3' | AGATCGGAAGAGCGTCGTGTAGGGAAAGAGTGT | | TruSeq (shared stem -- catches both) | -- | AGATCGGAAGAGC | | Nextera / Tn5 | transposase | CTGTCTCTTATACACATCT | | TruSeq small-RNA | 3' | TGGAATTCTCGGGTGCCAAGG |

The R1 3' adapter is the reverse complement of the R2-side region; trimming the shared 13 bp stem AGATCGGAAGAGC on both mates catches TruSeq read-through.

Tool Taxonomy

| Tool | Mechanism | When it wins | |------|-----------|--------------| | Cutadapt | Error-tolerant semiglobal alignment of a supplied adapter | PRECISION: small-RNA 3' adapter, amplicon/16S primers, anchored/linked adapters, demultiplexing. The correctness reference. | | Trimmomatic | ILLUMINACLIP simple + palindrome modes; ordered step pipeline | Legacy/reproducibility pipelines; palindrome PE read-through detection | | fastp | PE overlap analysis (no adapter sequence needed) + auto poly-G | DEFAULT general-purpose trim; one fast pass (route OUT -> fastp-workflow) | | Trim Galore | Cutadapt + FastQC wrapper with adapter auto-detect | Bisulfite/RRBS (--rrbs), Bismark pipelines | | BBDuk | k-mer match against an adapter/contaminant reference | Contaminant/PhiX removal in the same pass (route OUT -> contamination-screening) |

Decision Tree by Scenario

| Scenario | Use | Why | |----------|-----|-----| | General Illumina PE WGS/WES/RNA | fastp, or cutadapt with the TruSeq stem | Overlap analysis needs no sequence; cutadapt for explicit control | | Small-RNA / miRNA | cutadapt -a TGGAATTCTCGGGTGCCAAGG -m 18 -M 30 --discard-untrimmed | Adapter on every read; gate length and drop no-adapter reads | | Amplicon / 16S primers | cutadapt linked/anchored adapters | Primers are at fixed positions; needs precise placement | | PE read-through, no adapter sequence known | fastp overlap, or Trimmomatic palindrome | Both detect read-through from the R1/R2 overlap | | Bisulfite / RRBS | Trim Galore --rrbs | Handles MspI fill-in and Bismark conventions | | NextSeq/NovaSeq with poly-G tails | fastp (auto) or cutadapt --nextseq-trim | Poly-G is high-Q; quality trim alone misses it |

Default when uncertain: fastp for bulk PE, cutadapt with the TruSeq stem for explicit single-tool control.

Cutadapt

The algorithm is semiglobal (overlap) alignment, so a partial 3' adapter at the read end is detected. Two defaults drive behavior: -e (error rate, default 0.1) is computed against the LENGTH OF THE MATCHED REGION, not the whole adapter (an 8 bp match with 1 error is rate 0.125 and is rejected at the default); -O (minimum overlap, default 3) costs only ~0.07 bases lost per read by chance.

# Single-end 3' adapter
cutadapt -a AGATCGGAAGAGC -m 20 -o trimmed.fq.gz in.fq.gz

# Paired-end TruSeq (shared stem on both mates); both reads of a pair are discarded together
cutadapt -a AGATCGGAAGAGC -A AGATCGGAAGAGC -m 20:20 \
         -o R1.fq.gz -p R2.fq.gz in_R1.fq.gz in_R2.fq.gz

# Small-RNA: adapter on every read -> discard untrimmed, gate length
cutadapt -a TGGAATTCTCGGGTGCCAAGG -m 18 -M 30 --discard-untrimmed -j 8 \
         -o mirna.fq.gz raw.fq.gz

# Amplicon: linked 5'...3' primers (anchor with ^ to require the 5' primer)
cutadapt -g ^FWDPRIMER...REVPRIMER -o trimmed.fq.gz in.fq.gz

# 2-color poly-G aware (treats G as low quality so high-Q poly-G is trimmed)
cutadapt --nextseq-trim=20 -a AGATCGGAAGAGC -m 20 -o out.fq.gz in.fq.gz

# Higher error tolerance / longer required overlap when matches are missed / spurious
cutadapt -a ADAPTER -e 0.15 -O 5 -m 20 -o out.fq.gz in.fq.gz

Key flags: -a/-g/-b (3'/5'/anywhere, R1), -A/-G/-B (R2), -q (quality trim, BWA running-sum, runs BEFORE adapter removal), --pair-filter {any,both,first} (default any), --max-n, --action {trim,mask,lowercase,none}. When a filtering option discards reads in PE mode, both files MUST be processed together or they fall out of sync.

Trimmomatic

ILLUMINACLIP:<adapters.fa>:<seedMismatches>:<palindromeClip>:<simpleClip>:<minAdapterLen>:<keepBothReads>

• seedMismatches: mismatches tolerated in the initial seed (commonly 2).
• palindromeClip (~30): log-odds threshold for the PE palindrome alignment; ~30 needs ~50 matched bases.
• simpleClip (~10): log-odds threshold for an adapter-vs-read match; ~10 needs ~16 bases.
• keepBothReads: DEFAULT False -- after palindrome detects read-through, R2 is redundant (reverse complement of R1) and is DROPPED; set True if a downstream tool needs both mates.

SIMPLE mode tests each adapter against each read. PALINDROME mode (PE-only) aligns R1+adapter against the reverse complement of R2+adapter, so it detects read-through even when only a few adapter bases remain or the adapter is entirely past the read end. Steps run in COMMAND-LINE ORDER; put ILLUMINACLIP first and MINLEN last so the length check reflects all prior trimming.

# Paired-end, palindrome-capable adapter file, MINLEN last
trimmomatic PE -phred33 -threads 8 \
    in_R1.fq.gz in_R2.fq.gz \
    R1_paired.fq.gz R1_unpaired.fq.gz R2_paired.fq.gz R2_unpaired.fq.gz \
    ILLUMINACLIP:TruSeq3-PE-2.fa:2:30:10:2:keepBothReads MINLEN:36

# Built-in adapter files ship with the install
ls $CONDA_PREFIX/share/trimmomatic-*/adapters/

PE mode emits FOUR files: paired (both mates survived) and unpaired/orphan (mate dropped). Feed the paired files to the aligner; the orphans stay synchronized out of the way.

Common Errors

| Symptom | Cause | Solution | |---------|-------|----------| | FastQC still shows adapter after trimming | Wrong adapter, too-low -e, or only partial stem used | Use the shared stem AGATCGGAAGAGC; raise -e to 0.15; BLAST the overrepresented sequence | | Reads truncated / many lose a few bp | -O too low -> random 3-mer matches | Raise -O (e.g. 5); the default loses ~0.07 bp/read by chance | | Aligner reports R1/R2 out of sync | Mates trimmed/filtered independently | Process pairs together (cutadapt -p; Trimmomatic paired outputs) | | Small-RNA yields huge "reads" | Forgot --discard-untrimmed / length gate | Add --discard-untrimmed -m 18 -M 30 | | 3' G-content rise persists after trimming | 2-color poly-G is high-quality, not adapter | cutadapt --nextseq-trim or fastp auto poly-G | | Half of R2 disappears in Trimmomatic | keepBothReads default False drops redundant R2 | Add keepBothReads (True) if both mates are needed downstream |

References

Martin M. 2011. Cutadapt removes adapter sequences from high-throughput sequencing reads. EMBnet.journal 17(1):10-12. Bolger AM, Lohse M, Usadel B. 2014. Trimmomatic: a flexible trimmer for Illumina sequence data. Bioinformatics 30(15):2114-2120. MacManes MD. 2014. On the optimal trimming of high-throughput mRNA sequence data. Frontiers in Genetics 5:13. Williams CR, Baccarella A, Parrish JZ, Kim CC. 2016. Trimming of sequence reads alters RNA-Seq gene expression estimates. BMC Bioinformatics 17:103. Chen S, Zhou Y, Chen Y, Gu J. 2018. fastp: an ultra-fast all-in-one FASTQ preprocessor. Bioinformatics 34(17):i884-i890.

Related Skills

read-qc/quality-reports - Read the adapter-content panel that triggers trimming read-qc/quality-filtering - Quality and length filtering after adapter removal read-qc/fastp-workflow - All-in-one adapter + quality trim with auto poly-G read-qc/contamination-screening - k-mer removal of PhiX/vector/contaminant sequence small-rna-seq/smrna-preprocessing - Full small-RNA adapter + length workflow read-alignment/bwa-alignment - Soft-clipping aligner that handles low-quality tails without trimming