meru

meru is a small Crystal CLI for exploratory k-mer profiling in the terminal.

This is a simple project created with agentic AI coding. It displays Smudgeplot-like heatmaps in the terminal.

Build

make build release=1

Or:

shards build --release

Usage

bin/meru reads.fq.gz -k 21 -o sample

Tiny example:

bin/meru spec/fixtures/tiny.fastq -k 3 -o tiny

Inputs can be FASTQ (.fq, .fastq, optionally .gz) or FASTA (.fa, .fasta, optionally .gz), detected by filename extension.

--min-depth and --max-depth define the histogram depth range. Pair extraction uses max(2, --min-depth) and --max-depth by default, so singleton error k-mers stay out of the smudge workflow unless you opt in explicitly. --pair-min-depth and --pair-max-depth override only the pair-extraction range. When a smudgeplot is shown, meru expands the effective pair max depth to max_depth * 2 on the y-axis because the smudgeplot uses total coverage (depth_a + depth_b), not single-k-mer depth. The histogram uses log10(count + 1) scaling by default. Use --linear-hist if you want raw linear counts instead.

Current limitation: meru stores k-mer counts in a standard in-memory Hash(UInt64, UInt32). That keeps the implementation simple, but very large reference FASTA inputs such as whole human assemblies can exceed the hash capacity or available memory. In practice, meru is currently a better fit for read datasets and smaller references than for hg38-scale full-reference counting.

Main outputs:

• sample.kmer_hist.tsv
• sample.smudge.tsv
• sample.signals.tsv
• sample.summary.txt

Examples

The examples/ directory contains a reproducible synthetic workflow built around fictional tanuki genomes:

cd examples
rake tanuki MERU=../bin/meru

More detail is in examples/README.md.