Alignment Amplicon Clipping

Trim PCR primers from aligned reads in amplicon-panel BAMs.

Overview

Amplicon panels (SARS-CoV-2 ARTIC, hereditary cancer panels, ctDNA hot-spot panels, fusion panels, 16S rRNA) use designed PCR primers for enrichment. The primer footprint at read 5' ends does not represent biological sequence; without trimming, downstream variant calls suppress true variants under primers and falsely confirm reference at primer locations.

This skill covers post-alignment primer clipping with samtools ampliconclip, the required tag-repair steps (fixmate, calmd), and the relationship to duplicate marking (do not run markdup on amplicon BAMs without UMIs).

Prerequisites

samtools 1.11 or later (ampliconclip introduced in 1.11)
Coordinate-sorted, indexed amplicon BAM
Primer BED file with strand information in column 6 (for --strand mode)

conda install -c bioconda samtools
# Verify ampliconclip is available
samtools ampliconclip --help

Quick Start

Tell your AI agent what you want to do: - "Trim ARTIC SARS-CoV-2 primers from my amplicon BAM" - "Soft-clip primers using my primer BED" - "Why is markdup marking everything as duplicate on my amplicon panel?" - "Repair the MD/NM tags after primer clipping" - "Hard-clip primers for archival storage"

Example Prompts

SARS-CoV-2 / ARTIC

"I have a SARS-CoV-2 ARTIC BAM aligned with bwa-mem. Soft-clip primers using artic_v3 primer BED, then prepare for consensus calling."

"Compare samtools ampliconclip vs iVar trim for the ARTIC SARS-CoV-2 pipeline."

Cancer Panels

"ctDNA hot-spot panel with TruSeq UMIs. Trim primers, then run UMI consensus rather than markdup."

"Hereditary cancer panel from a Roche AVENIO design. Why is my variant caller missing variants at primer-overlapping positions?"

General Amplicon Workflows

"My amplicon BAM ran through markdup and now flagstat shows 99% duplicates. What went wrong?"

"After ampliconclip, my downstream mpileup is reporting wrong reference bases. What tag did clipping invalidate?"

What the Agent Will Do

Inspect the BAM to confirm it is amplicon-style (read coordinates clustered at amplicon start/end positions; check samtools view -H for @PG indicating an amplicon panel kit).
Locate or generate the primer BED file matching the kit.
Run samtools ampliconclip --strand --soft-clip -b primers.bed (or --both-ends instead of --strand for read-through amplicons).
Re-fixmate and re-calmd to repair invalidated tags.
Confirm with a region inspection (samtools view -h clipped.bam chr1:start-end) that primer footprint has soft-clip CIGAR ops.

Decision Points

Soft-clip vs hard-clip: default to soft-clip (reversible). Hard-clip only when archiving and the trimmed bases will never be needed.
--strand on or off: on for primer-strand-specific trimming (default for amplicon panels with single-stranded primers); off only if both strands carry primer-derived sequence.
--both-ends on or off: on when reads can read through the entire amplicon and primers can appear at either 5' or 3'. Standard for short-amplicon panels (<300 bp). Note: --both-ends overrides --strand (both ends are clipped regardless of strand), so use one or the other.
Re-calmd: required if downstream uses BAQ in bcftools mpileup, IGV mismatch visualization, or any tool that reads NM/MD.

Tips

Do NOT run samtools markdup on amplicon BAMs without UMIs -- amplicon reads share start coordinates by design and markdup will erase the dataset. See duplicate-handling for the assay-aware decision matrix.
For UMI amplicon panels (Twist UMI, IDT xGen UMI, Roche AVENIO): use fgbio GroupReadsByUmi -> CallMolecularConsensusReads instead of markdup.
For ARTIC SARS-CoV-2 specifically, modern pipelines often pair ampliconclip with samtools consensus --config hiseq --ambig (Illumina preset) for IUPAC heterozygote handling.
Strand bias at every amplicon end usually means --strand was forgotten.
After clipping, mpileup with default BAQ may be wrong (MD tag stale); always re-calmd.
For long-read amplicon (ONT viral, PacBio HiFi 16S): tool support varies; samtools ampliconclip works but verify CIGAR semantics with the long-read team's recommended pipeline.

duplicate-handling - Why amplicon BAMs need ampliconclip, not markdup
alignment-filtering - Post-clip filtering for variant calling
alignment-sorting - Re-sort after fixmate
pileup-generation - mpileup flags for amplicon
reference-operations - Consensus generation from amplicon BAMs
read-qc/quality-reports - Pre-alignment QC