Version history

First release of nf-core/deepmutscan, a pipeline for deep mutational scanning (DMS) experiments. It takes paired-end reads from the input and output libraries of a selection experiment to variant counts, sequencing error correction, library QC and fitness estimates. All results are summarised in one self-contained HTML report.

Code name: This pipeline was largely developed in Barcelona, so releases will be named after FC Barcelona players. v1.0.0 is dedicated to one of the club’s most famous team captains and centre-back defenders, representing a solid foundation for the pipeline.

Many thanks to members of the nf-core community, in particular @mashehu, @mirpedrol for their contributions, and @piplus2 and @maxulysse for their thorough release reviews.

v1.0.0 - Puyol - [2026-11-14]

Added

  • Samplesheet input of paired-end FASTQ files per library, annotated with sample, type (input, output, wildtype) and replicate, validated at start-up (one sample per run, unique sample/type/replicate, matching input and output replicate numbers, one wildtype library for --error_correction wildtype), together with --reading_frame against the reference
  • Input FASTQ checks before any reads are processed: quick checks of the first read of every file at start-up (empty, not gzip or not FASTQ, same file or same reads twice, mates with different names, file shared between libraries), then fq lint on the complete files (truncated or corrupted gzip, malformed records, unequal read numbers, mates out of order); all faulty libraries are reported together
  • Raw read QC with FastQC and MultiQC
  • Read-pair merging into consensus reads with vsearch --fastq_mergepairs
  • Alignment of the merged reads to a gene-sized reference ORF with BWA-MEM (--fasta, --reading_frame)
  • Filtering of unmapped, secondary, supplementary, low mapping-quality (MAPQ < 30) and indel-containing alignments with samtools, then coordinate sorting and indexing; wildtype reads are retained for error correction
  • Light-weight variant counter built on the pysam and polars libraries in Python, replacing GATK AnalyzeSaturationMutagenesis with a column-compatible output, including a minimum base quality (--base_qual, default Q40) and read-edge exclusion window (--min_flank, default 2 bp)
  • Annotation and filtering of variant counts against the programmed mutagenesis library (--mutagenesis_type nnk, nns, nnh, nnn, nnk_nns, nnk_nns_nnh or custom with --custom_codon_library)
  • Single-nucleotide variant sequencing-error correction (--error_correction):
    • false_doubles (default): error rates estimated from read-linked false double mutant codons, by maximum likelihood (--false_doubles_method mle, default) or empirical Bayes (eb), within a configurable codon window (--false_doubles_codon_window)
    • wildtype: error profile from additional deep sequencing of the unmutated template (type: wildtype samplesheet rows)
    • none
  • Interactive, run-level HTML error-correction report
  • DMS library QC per library: count and count-per-coverage heatmaps, sorted count distributions with LogDiff (variants without counts drawn at a floor derived from the data and counted separately; LogDiff not reported when more than 10% of variants have no counts), sliding-window coverage and count profiles (--sliding_window_size, --aimed_cov), and sequencing-depth rarefaction on a geometric depth grid (--run_seqdepth)
  • Optional fitness estimation (--fitness): merged count tables, experimental design, synonymous wildtype proxy selection, default log-ratio fitness with replicate rescaling and summary statistics, replicate correlation plots and fitness heatmap
  • Optional fitness estimation with DiMSum (--dimsum) and mutscan edgeR / limma (--mutscan)
  • Minimum number of reads a variant needs in every input replicate to receive a fitness estimate (--min_counts, default 10), applied in the same way by the default estimator, DiMSum and mutscan; variant counting, the sequencing-error correction and the library QC always use all reads
  • Pseudocount for variants that drop out of an output replicate (--output_pseudocount, default 1; 0 gives dropouts no fitness value)
  • Optional interactive 3D variant effect inspection tool built from a user-supplied wildtype structure (--pdb)
  • Self-contained, all-in-one run report (deepmutscan_report.html) embedding QC, error-correction, fitness and MultiQC results, with links to Nextflow’s execution report and timeline
  • Docker and Singularity images (amd64 and arm64) and conda lock files for all modules, built from each module’s environment.yml; DiMSum and mutscan are amd64 only for now
  • test profile on a 50,000 read-pair subsample of a GID1A nicking-mutagenesis GluePCA experiment, with three nf-test pipeline tests (all options; without fitness, error correction, SeqDepth and structure; stub run), and test_full profile on the complete dataset (ENA PRJEB110196); test data and the PDB structure are hosted on nf-core/test-datasets

Dependencies

Dependency Version
fastqc 0.13.0
fq 0.12.0
multiqc 1.35
bwa 0.7.19
samtools 1.21, 1.22.1
vsearch 2.30.0
python 3.12
pysam 0.24.0
polars 1.33.1 (lts-cpu)
pyarrow 24.0.0
biopython 1.87
numpy 1.26.4, 2.5.1
pandas 2.2.1
r-base 4.4.2 (DiMSum), 4.5.1
biostrings 2.74.0, 2.78.0
r-tidyverse 2.0.0
r-ggplot2 3.5.1, 4.0.2
r-zoo 1.8_15
r-dimsum 1.4
mutscan 1.0.0