KamScanRust

KamScanRust Performs fast parallel statistical tests over long k-mer, contig or other features count matrices, streaming differential analysis between two conditions (Welch T-test, Wilcoxon rank-sum, or ANCOVA) — multi-threaded. Adapted from https://github.com/Transipedia/KamScan.

---

🛠️ Installation

Prerequisites

• Rust 1.70 or later (Install Rust)
• Git (optional)

Steps

# Clone the repository
git clone https://github.com/Transipedia/KamScanRust.git
cd KamScanRust

# Build and install
cargo build --release

---

Usage

Basic Command

kamscan [OPTIONS] --design <DESIGN>

Options

• -t, --tsv Input counts file, space- or tab-separated. Use '-' or omit to read from stdin

• -d, --design Design file: sample_name condition [covariate]

• -o, --output Output file [default: stdout]

• -p, --pseudo Pseudocount added before log2 transformation [default: 1.0]

• -m, --min-count <MIN_COUNT> Skip features where ALL relevant samples have count <= this value [default: 0.0]

• -s, --max-pvalue <MAX_PVALUE> Only report features with p-value (or adj p-value when --bh) <= this threshold [default: 0.05]

• --test Statistical test to use [default: ttest] Possible values: - ttest: Welch's two-sample t-test on log2-transformed counts (default) - wilcoxon: Wilcoxon rank-sum test (Mann-Whitney U) on log2-transformed counts - ancova: ANCOVA: remove linear covariate effect before testing group difference. Requires a numeric third column in the design file

• --no-header <HEADER_FILE> Counts file has no header row; instead, read column names (one per line) from HEADER_FILE. The first name is the feature-ID column; remaining names are sample names matched against the design file in order

• -j, --threads Number of worker threads [default: logical CPU count]

• --chunk-factor <CHUNK_FACTOR> Chunk size multiplier: each chunk holds chunk_factor × 64 KiB of raw input text. Increase for files with very long lines. [default: 16 → 1 MiB chunks]

• --norm-counts <NORM_COUNTS> Optional file of pre-computed column sums for CPM normalization. Format: sample_name TAB count_sum (one line per sample, any order)

• --norm-scale <NORM_SCALE> Normalization factor for the CPM normalization [default: 1000000]

• --bh Apply Benjamini-Hochberg FDR correction (two-pass run). Incompatible with reading the counts file from stdin

• --output-mode <OUTPUT_MODE> What to output for significant features. 'stats': means, logFC, test statistic, p-value (default). 'raw': the original input line verbatim (useful for large count matrices)

     Possible values:
     - stats: Output statistics columns: means, logFC, test stat, p-value (default)
     - raw:   Output the original input line verbatim for each significant feature
  

• -h, --help Print help (see a summary with '-h')