Every option, with its default. Required options have no default and must be given.

Run kmate <command> --help for the same information at the terminal.

This page describes kMate 0.1.5 (same options as 0.1.4). 0.1.3 lacks build-index --tmp-dir. In 0.1.2, run still accepted --kmer-weight and --emit-af-se and spelled --smooth-windows as --no-local-only, and decompose had no --gfa. kmate --help is authoritative for the version you have installed.

command	what it does
run	estimate founder-haplotype frequencies and project to allele frequencies
build-index	build the per-bubble k-mer index from a panel VCF
build-kmer-pa	build the founder-haplotype × k-mer matrix
build-var-pa	build the founder-haplotype × variant matrices
build-kmer-db	count a pool's k-mers once for reuse
decompose	multi-allelic → biallelic VCF
transfer-id	copy INFO/ID onto a genotyped VCF
filter-pa	re-filter an existing kmer_pa matrix
selftest	run the bundled fixture end to end

---

option	default	meaning
--kmer-pa-prefix	required	prefix of the k-mer matrices; _<CHROM>.kmer_pa.npz and _<CHROM>.meta.npz are appended
--var-pa-prefix	—	prefix of the variant matrices; _<CHROM>.{var_pa,var_called,meta}.npz appended. Use instead of the three options below
--var-pa	—	explicit path to var_pa.npz
--var-called	—	explicit path to var_called.npz. Guessed from --var-pa if omitted
--var-meta	—	explicit path to meta.npz
--reads	required	FASTQ(s) for one pool
--kmer-db	—	pre-counted Jellyfish DB from build-kmer-db; skips re-scanning the reads
--sample	required	sample name written into the output
--out	required	output TSV
--chroms	Chr1 Chr2 Chr3 Chr4 Chr5	chromosomes to process
--h-only	off	estimate the founder-haplotype mixture only, skip the allele-frequency projection

option	default	meaning
--unit	chrom	where the mixture is fitted: chrom, bp, ld, tsv. See Haplotypes and windows
--normalize	per_founder	M-step normalisation. per_founder divides each haplotype's update by its own k-mer content; global is the legacy multinomial form, under which k-mer-poor haplotypes collapse toward zero
--haploblock-eps	0.0	merge haplotypes whose k-mer presence differs by at most this fraction of the unit's k-mers. 0 merges only exactly identical ones
--max-kmer-cov-mult	5.0	repeat guard: drop k-mers seen more than this multiple of the estimated coverage (never below 1 read). 0 disables. See Running kMate
--threads	4	threads for k-mer counting
--hash-size	3G	Jellyfish hash size. Lower it on memory-capped jobs

Used when --unit bp, ld or tsv; ignored for --unit chrom.

option	default	meaning
--window-bp	10000	window width for --unit bp
--ld-r2	0.1	r² cutoff for --unit ld
--ld-window	100	variants considered per LD block
--ld-blocks	—	precomputed LD blocks
--blocks-tsv	—	explicit block partition, required by --unit tsv
--min-kmers-per-block	200	a window with fewer observed k-mers gets no local fit
--smooth-windows	off	tie neighbouring windows together: enables the anchor and smoothing below, and fills low-support windows with the chromosome-wide mixture instead of NaN. Off by default, so each window uses only the information its own k-mers carry
--global-anchor-weight	0.3	anchor strength toward the chromosome-wide mixture. Applies only with --smooth-windows
--hmm-smooth-passes	5	cross-window smoothing passes. Applies only with --smooth-windows
--hmm-smooth-alpha	0.5	smoothing strength; smaller smooths more
--hmm-smooth-recomb-rate	4e-08	recombination rate used by the smoother

option	meaning
--block-mode	alias for --unit: global→chrom, window→bp. --unit wins if both are given
--local-only / --no-local-only	old spelling; --no-local-only is --smooth-windows

---

option	default	meaning
--vcf	required	the graph VCF (multi-allelic, before decompose)
--ref	required	reference FASTA, indexed
--out	required	output prefix; writes <prefix>_<CHROM>_kmers.tsv.gz
-k, --kmer-size	31	k-mer length
--haploid	off	accept haploid genotypes. Required for a haploid panel VCF
--cap-biallelic	16	k-mers kept per allele at biallelic bubbles
--cap-multiallelic	32	k-mers kept per allele at multi-allelic bubbles
--overhang-cap	12	k-mers kept per bubble side
--no-caps	off	disable the caps above
--no-add-reference	off	omit the synthetic all-reference path
--emit-overhang	off	also write overhang k-mers
--jellyfish-threads	4	threads for the reference k-mer count
--jellyfish-hash	100000000	hash size for it
--keep-tempfiles	off	keep intermediates for debugging
--tmp-dir	next to --out	where the temporary files go (~2× genome size per assembly set; ~10 GB for 135 Arabidopsis assemblies). Use node-local disk: kMate warns when it is a network filesystem

---

option	default	meaning
--kmers	required	kmers.tsv.gz from build-index
--vcf	required	the panel VCF (decomposed, biallelic)
--ref	required	reference FASTA. IUPAC codes are normalised to N internally
--chrom	required	chromosome to build
--out	required	output prefix
--treat-missing-as-n	off	mask a missing genotype with N over the record, dropping the overlapping k-mers. Without it, missing is treated as reference
--filter-production	off	apply the column filter below at build time
--min-ac	2	keep k-mers carried by at least this many haplotypes. Use 1 on small panels: on 8 founders, 2 discarded ~46% of k-mers
--invariant-margin	1	drop k-mers carried by more than F - margin haplotypes
--max-bubbles	—	stop after this many bubbles (testing)

---

option	default	meaning
--vcf	required	the same panel VCF. Must be biallelic and haploid
--out	required	output prefix; writes .var_pa.npz, .var_called.npz, .meta.npz
--chrom	—	restrict to one chromosome

---

option	default	meaning
--reads	required	FASTQ(s) for one pool
--out	required	output Jellyfish DB
--threads	4	counting threads
--hash-size	3G	Jellyfish hash size

---

Runs two bundled PanGenie scripts; see Building a panel. Give either --gfa or both --annotated-vcf and --biallelic-catalog.

option	default	meaning
--genotyped-vcf	required	the VCF to decompose; with --gfa, the Minigraph-Cactus VCF of that graph
--gfa	—	the graph's gzipped GFA; annotates --genotyped-vcf against it
--annotated-vcf	—	annotated multi-allelic VCF carrying INFO/ID (instead of --gfa)
--biallelic-catalog	—	biallelic catalog with the same symbolic IDs (instead of --gfa)
--out	required	output VCF, sorted; indexed when it ends in .gz
--annotate-vcf-script	bundled	use another copy of annotate_vcf.py
--convert-to-biallelic	bundled	use another copy of convert-to-biallelic.py
--haploidize	off	also reduce genotypes to one allele
--het	missing	with --haploidize: missing for inbred founders, split for outbred phased ones
--threads	4	threads for bcftools
--keep-temp	off	keep intermediates
--citation	—	print attribution and exit

---

option	default	meaning
--cactus	required	annotated VCF carrying INFO/ID
--pg	required	genotyped VCF from the same graph, lacking the IDs
--out	required	output VCF

---

Applies the build-kmer-pa column filter to a matrix that was built unfiltered.

option	default	meaning
--in-prefix	required	existing kmer_pa prefix
--out-prefix	required	filtered output prefix
--min-ac	2	as in build-kmer-pa
--invariant-margin	1	as in build-kmer-pa