Norovirus Capture Sequencing Simulator. Generates realistic paired-end FASTQ from simulated hybrid capture enrichment of Norovirus VP1, calibrated from empirical data in Bhamidipati et al. (2025) Sci Rep 15:20526.

Human Norovirus genotype is defined by the VP1 region (ORF2, ~1,700 bp). Hybrid capture enrichment followed by paired-end sequencing on NovaSeq is the current gold standard for lineage surveillance. This simulator produces controlled FASTQ datasets — with known genotypes, Ct values, and ground truth coverage — so that full analysis pipelines can be benchmarked before deployment.

Empirical calibration source:

Bhamidipati et al. (2025) "Complete genomic characterization of RSV and HuNoV using probe-based capture enrichment." Sci Rep 15:20526 (PMC12216758)

The simulation pipeline has six stages that mirror the wet-lab workflow:

1. Ct-to-viral-fraction model. Each sample's Ct value is converted to a viral fraction using vf = 0.03 * 2^((25 - ct) / 3.32). The exponential relationship follows standard qPCR theory — each 3.32 Ct difference corresponds to a 10-fold change in template concentration (2^3.32 ≈ 10). The calibration parameters (base fraction 0.03 at Ct 25) were tuned to produce coverage outputs consistent with the empirical results in Table 1 of Bhamidipati et al. (2025); they are not directly reported in the paper. A Ct of 20 gives ~30% viral reads; a Ct of 35 gives ~0.03%. The total pre-capture library (default 500,000 fragments) is split into n_viral and n_background based on this fraction.

2. Fragment generation. Viral fragments are sampled from the reference genome using a truncated normal length distribution (mean=380 bp, sd=80 bp, range 200-600 bp) matching heat fragmentation at 94 C for 10 minutes. Each fragment records its genomic coordinates, strand, GC content, and whether it overlaps VP1. Background fragments come from human DNA (GC ~0.41), gut microbiome (GC ~0.52), or wastewater environmental sources (GC ~0.48), mixed according to sample type (stool: 80/20 human/microbiome; wastewater: 10/30/60 human/microbiome/environmental).

3. Capture enrichment. Hybrid capture is modelled per-fragment. Viral fragments that overlap VP1 are retained with probability P = exp(-6.0 * (gc - 0.47)^2), a Gaussian centered on the optimal probe GC of 0.47. Fragments outside VP1 and all background fragments pass through at the configured off-target rate (default 59.2%, from 1 - 0.408 on-target rate in Table 1 of Bhamidipati et al.). This produces on-target and off-target fragment pools.

4. PCR duplicate model. Each captured fragment is assigned a copy count drawn from a geometric distribution with p = 1 - dup_rate. At the default 40% duplicate rate, most fragments appear once, but some are amplified 2-5x, matching the 12-20 PCR cycles used in post-capture amplification.

5. Read simulation. The duplicated fragment pool is written as a PBSIM3 TSV (sequence, start, end per fragment) and passed to art_modern, which generates paired-end 2x150 bp reads with Illumina NovaSeq-calibrated base quality scores and error profiles. The output is deinterleaved into R1/R2 FASTQ files and gzip-compressed.

6. Truth and QC. Ground truth VP1 coverage is computed directly from fragment coordinates: mean depth and the fraction of VP1 bases at >= 20x. Each sample receives a completeness call — complete (>= 90% of VP1 at

= 20x), low_coverage (>= 90% breadth but < 20x mean), or incomplete (< 90% breadth) — following the criteria in the paper.

• Python >= 3.11
• art_modern (external binary, see below)
• Python dependencies: click, biopython, numpy, scipy

git clone <repo>
cd nocasim
uv venv .venv && source .venv/bin/activate
uv pip install -e ".[dev]"

Download the .deb from the art_modern releases page, then install it:

sudo dpkg -i art-modern_*.deb
sudo apt-get install -f -y

Verify the install:

art_modern --version

Pre-built images are available from Google Artifact Registry:

docker pull us-docker.pkg.dev/general-theiagen/theiagen/nocasim:v0.3.0

Run a single sample:

docker run --rm -v $(pwd)/results:/output \
  us-docker.pkg.dev/general-theiagen/theiagen/nocasim:v0.3.0 single \
    --reference /opt/nocasim/data/references/GII.4.fasta \
    --ct 28.0 --outdir /output --art-modern art_modern

Run batch mode with a custom sample sheet:

docker run --rm \
  -v $(pwd)/samples.tsv:/data/samples.tsv \
  -v $(pwd)/results:/output \
  us-docker.pkg.dev/general-theiagen/theiagen/nocasim:v0.3.0 simulate \
    --sample-sheet /data/samples.tsv \
    --references /opt/nocasim/data/references/ \
    --art-modern art_modern \
    --outdir /output

To build locally instead:

docker build -t nocasim .

nocasim single \
  --reference data/references/GII.4.fasta \
  --ct 28.0 \
  --outdir results/ \
  --art-modern art_modern

nocasim simulate \
  --sample-sheet samples.tsv \
  --references data/references/ \
  --art-modern art_modern \
  --outdir results/

Use --sample-type wastewater to simulate wastewater surveillance samples. The background composition shifts from stool (80% human / 20% microbiome) to wastewater (10% human / 30% microbiome / 60% environmental):

nocasim simulate \
  --sample-sheet samples.tsv \
  --references data/references/ \
  --art-modern art_modern \
  --sample-type wastewater \
  --outdir results/

Optionally provide a wastewater background FASTA for realistic environmental sequences instead of synthetic ones:

nocasim simulate \
  --sample-sheet samples.tsv \
  --references data/references/ \
  --art-modern art_modern \
  --sample-type wastewater \
  --wastewater-bg data/background/wastewater_metagenome.fasta \
  --outdir results/

Built-in presets simulate realistic wastewater genotype distributions:

nocasim simulate \
  --sample-sheet samples.tsv \
  --references data/references/ \
  --art-modern art_modern \
  --preset us-2024 \
  --outdir results/

For custom mixtures, use --mixture instead:

nocasim simulate \
  --sample-sheet samples.tsv \
  --references data/references/ \
  --art-modern art_modern \
  --mixture "GII.17:0.60,GII.4:0.25,GI.1:0.15" \
  --outdir results/

--mixture and --preset are mutually exclusive.

nocasim download-probes --outdir data/probes/

This downloads Supplementary File 2 from PMC12216758 and saves it as data/probes/hunov_probes.txt (BED-like probe interval file).

The sample sheet is a TSV file with three columns:

sample_id    genotype    ct_value
sample_001   GII.4       24.5
sample_002   GII.17      30.0
sample_003   GI.1        31.7

The genotype value must match the filename stem of a FASTA file in the --references directory. For example, GII.4 maps to data/references/GII.4.fasta.

The genotype column also accepts comma-separated mixture specs for simulating wastewater samples with multiple co-circulating lineages:

sample_id    genotype                              ct_value
ww_001       GII.17:0.75,GII.4:0.15,GI.1:0.10    24.5
ww_002       GII.4                                 28.0

Proportions must sum to 1.0. Samples with a bare genotype name are treated as single-lineage (equivalent to GII.4:1.0).

Alternatively, use --preset or --mixture CLI flags to apply a mixture to all samples at once (see Mixture presets below). Per-sample mixture syntax in the TSV overrides CLI flags.

Example sample sheets are included at data/samples.tsv (single-lineage) and data/samples_mixture.tsv (multi-lineage).

The repository includes 35 reference sequences (9 GI + 26 GII genotypes) from the CDC Calicivirus Typing Tool. One FASTA file per genotype in data/references/, where the filename stem matches the genotype string in the sample sheet:

data/references/
├── GI.1.fasta    (EU085529, 3081 bp)
├── GI.2.fasta    (AF435807, 2354 bp)
├── ...
├── GII.4.fasta   (X76716,   3881 bp)
├── GII.17.fasta  (KJ156329, 3723 bp)
├── ...
└── GII.27.fasta  (MK733205, 7308 bp)

References range from partial VP1 (~1,000 bp) to full genome (~7,700 bp). Full-genome references use the configured vp1_start/vp1_end coordinates to extract the target region (defaults: 5,100-6,800 bp). Shorter sequences are treated as the target region in their entirety.

results/
├── summary.tsv
└── sample_001/
    ├── sample_001_R1.fastq.gz
    ├── sample_001_R2.fastq.gz
    └── sample_001_manifest.json

summary.tsv contains one row per sample with columns: sample_id, genotype, ct_value, vp1_mean_depth, vp1_completeness_20x, completeness_call, lineage_detail. For multi-lineage samples, lineage_detail contains per-lineage stats in the format GII.17:67.1x/0.94;GII.4:13.4x/0.82 (depth/completeness per genotype).

sample_001_manifest.json contains per-sample ground truth metrics including achieved on-target rate, duplicate rate, mean VP1 depth, and completeness call. For multi-lineage samples, the manifest also includes mixture (input proportions), per_lineage (per-genotype coverage stats), and aggregate (weighted overall stats).

Completeness calls follow the paper's criteria:

• complete: >= 20x coverage across >= 90% of VP1
• low_coverage: >= 90% of VP1 breadth covered but mean depth < 20x
• incomplete: < 90% of VP1 covered

Background FASTA files (--human-bg, --microbiome-bg, --wastewater-bg) are optional. When omitted, synthetic sequences with realistic GC content profiles are generated as placeholders.

Fragment length distribution is a truncated normal (mean=380 bp, sd=80 bp, range 200-600 bp) matching heat fragmentation at 94 degrees C for 10 minutes as described in the paper. This differs from sonication (log-normal) — do not substitute a log-normal distribution.

pytest tests/ -v

• ORF1 / polymerase region (recombination makes it irrelevant for genotyping)
• Intra-lineage quasispecies diversity or SNV simulation
• Recombinant genomes (chimeric ORF1/ORF2 junctions)
• RNA extraction efficiency
• Reverse transcription efficiency