Version Compatibility

Reference examples tested with: samtools 1.19+, pysam 0.22+

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

• Python: pip show <package> then help(module.function) to check signatures
• 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.

Alignment Amplicon Clipping

"Trim primer-derived bases from amplicon BAM" -> Soft- or hard-clip the 5' primer footprint after alignment using a primer BED, then repair fixmate/MD/NM tags.

• CLI: samtools ampliconclip -b primers.bed input.bam -o clipped.bam (since samtools 1.11)
• Alternative: iVar trim, BAMClipper, fgbio ClipBam

Why Primer Trimming After Alignment

Amplicon panels (SARS-CoV-2 ARTIC, hereditary cancer panels, ctDNA hot-spot panels, fusion panels, 16S rRNA) use designed PCR primers for enrichment. Primer-derived bases at read 5' ends do NOT reflect biological sequence -- they reflect the primer template. Without trimming:

• False reference confirmation at primer footprint positions.
• Variant allele frequency suppressed at variants under primers (the primer sequence cannot capture the variant base).
• Strand bias artifacts (primers are typically one-strand).

Standard amplicon BAMs should NEVER be processed by samtools markdup -- by design every read at a primer location is a "duplicate" by coordinate. See duplicate-handling for the assay-aware decision.

Tool Selection

Soft-Clip vs Hard-Clip

Goal: Decide whether trimmed bases are kept in the BAM (reversible) or discarded (irreversible).

Approach: Soft-clip is the safe default; hard-clip only when archiving and disk is constrained.

Soft-clip is the recommended default. Hard-clip is irreversible -- once applied, the trimmed bases cannot be recovered for re-analysis with different primer coordinates.

Basic ampliconclip Workflow

Goal: Trim primers from a coordinate-sorted, indexed amplicon BAM and produce a downstream-ready BAM.

Approach: Run ampliconclip with primer BED, optionally --both-ends and --strand, then re-fixmate (CIGAR changed) and re-calmd (MD/NM tags invalidated by clip).

# 1. Soft-clip primers (default; reversible)
samtools ampliconclip --both-ends --strand --soft-clip \
    -b primers.bed input.bam -o clipped.bam

# 2. Re-pair tags (CIGARs changed -- mate info needs refresh)
samtools sort -n clipped.bam | \
    samtools fixmate -m - - | \
    samtools sort -o sorted.bam -

# 3. Repair MD/NM tags (invalidated by clip; required by mpileup BAQ and IGV)
samtools calmd -b sorted.bam reference.fa > clipped_final.bam
samtools index clipped_final.bam

Strand-Aware Clipping

--strand clips primer bases only on the strand the primer is designed for. Without --strand, both strands are clipped at the primer site, removing valid biological sequence on the off-strand.

Both-End Clipping

--both-ends allows clipping at both 5' and 3' positions of the read (some primers can appear at either end after alignment). Necessary for amplicon designs where reads can read through the entire amplicon.

Primer BED Format

# tab-separated, 0-based half-open like all BED
chr1   100   125   primer_1_F    +
chr1   500   525   primer_1_R    -
chr1   600   625   primer_2_F    +
chr1   1000  1025  primer_2_R    -

Tools that consume the BED: column 1-3 (region), column 6 (strand) is required for --strand. ARTIC primer schemes ship pre-built BEDs (e.g., primer.bed from artic-network/primer-schemes).

SARS-CoV-2 ARTIC Comparison

Modern viral consensus pipelines tend to use ampliconclip then samtools consensus --config illumina --ambig for IUPAC heterozygote handling. See reference-operations for consensus generation.

After Clipping: Required Re-Processing

Clipping invalidates several tags and CIGAR-derived fields:

A clipped BAM that bypasses fixmate + calmd causes silent failures in bcftools mpileup BAQ (which depends on MD), IGV mismatch coloring, and any tool using NM for filtering.

Why Not Markdup

Amplicon reads at primer locations are by design coordinate-degenerate -- every read mapped to the same amplicon shares the same start/end coordinates because they all come from the same primer pair. samtools markdup would mark essentially every read as a duplicate and erase the dataset. For amplicon panels:

• WITHOUT UMIs: skip dedup entirely; rely on coverage uniformity from amplicon design.
• WITH UMIs (deep panels): use fgbio GroupReadsByUmi -> CallMolecularConsensusReads instead of markdup. See duplicate-handling.

Common Errors

Quick Reference

Related Skills

• duplicate-handling - Why amplicon BAMs should not be markdup'd; UMI-aware alternatives
• alignment-filtering - Post-clip filtering for amplicon variant calling
• alignment-sorting - Re-sort after fixmate
• pileup-generation - mpileup flags for amplicon (-aa -A -d 600000 -B)
• reference-operations - Consensus generation from amplicon BAMs (samtools consensus)
• read-qc/quality-reports - Pre-alignment adapter/quality trimming