Eric Coissac 36c7247acb feat: add k-mer sharing estimation and Poisson-binomial null model
Add a new `estimate --index` command to compare observed k-mer sharing against
a random-sequence model using a Poisson-binomial distribution. Introduce the
`row_presence_histogram` API to support this analysis across various matrix
types. Update documentation and dependencies accordingly.

- Add `src/obikstats/src/null_model.rs` implementing `SharingNullModel` and
  `SharingRow` for computing cumulative probabilities and false discovery rates
  based on a Poisson-binomial distribution.
- Add `src/obikstats/src/sharing.rs` defining `GenomeSharing` and `SharingSpectrum`
  to compute k-mer sharing spectra using parallel map-reduce.
- Add `src/obikmer/src/cmd/estimate/sharing.rs` with `run` and `write_csv` to
  execute sharing estimation and output CSV results containing observed, expected,
  and deviation metrics.
- Modify `src/obikmer/src/cmd/estimate/mod.rs` to support the new `--index`
  subcommand, adding `EstimateArgs` fields for index path, presence threshold,
  and output file.
- Add `src/obicompactvec/src/row_histogram.rs` implementing `row_presence_histogram`
  for `PersistentIntMatrix`, `PersistentBitMatrix`, and sparse matrices to count
  rows present in exactly $j$ columns.
- Update `src/obikindex/src/layer/content_layer.rs` and `src/obikindex/src/layer/typed_layer.rs`
  to expose the new `row_presence_histogram` method, dispatching based on layer type.
- Add tests in `src/obicompactvec/src/tests/bitmatrix.rs`, `intmatrix.rs`, and
  `sparse.rs` to verify `count_geq` and `row_presence_histogram` against naive
  implementations for various matrix structures.
- Update `UserDocMD/theory/kmer_indexing/kmers.md` to document the Poisson-binomial
  approach, $D_r$ bounds, and FDR calculations, replacing previous binomial approximations.
- Update `UserDocMD/usage/estimate.md` to describe the two uses of `estimate`:
  parameter estimation and index comparison via `--index`.
- Add `wiki` target to `Makefile` for regenerating documentation from user docs.
- Bump `obikmer` package version from 1.4.5 to 1.4.6 in `Cargo.toml` and `Cargo.lock`.
2026-11-19 17:50:29 +01:00
2026-08-17 09:28:53 +02:00

obikmer

obikmer is a Rust toolkit for indexing, querying, and comparing DNA sequences represented as sets of k-mers. It targets individual genome datasets (tens of Gbases) with maximum efficiency in computation, memory, and disk usage.

Key principles

Compact k-mer encoding. Each k-mer is stored in a u64 at 2 bits/base. k is odd, k ∈ [11, 31], fixed at runtime. The canonical form min(kmer, revcomp(kmer)) halves the effective space by collapsing both strands.

Superkmer-based partitioning. Sequences are decomposed into superkmers — maximal runs of k-mers sharing the same minimizer. Superkmers route naturally to partitions via the minimizer hash, enabling partition-parallel indexing and querying with no cross-partition communication.

Layered MPHF index. Each partition holds a stack of layers. Each layer is a minimal perfect hash function (MPHF) over the k-mers of one input genome, paired with a per-genome presence/count matrix. Queries scatter k-mers to their partition, probe each layer in order, and aggregate results.

Approximate indexing (Findere). With -z Z, the index stores k-mers of size s = k − z + 1 instead of k. At query time, results are produced at size s, then a per-genome sliding window of size z aggregates z consecutive s-mer hits into one confirmed k-mer answer. This reduces the false-positive rate from 1/2^b per s-mer to 1/2^(b·z) per k-mer, at the cost of z−1 unconfirmed positions at each sequence break. The aggregation window spans the full query sequence, not individual superkmers, to avoid false negatives at superkmer boundaries.

Multi-genome. A single index can hold any number of genomes. Each k-mer slot carries a per-genome count or presence vector. Distance matrices, NJ/UPGMA trees, and classification are derived from these vectors without rebuilding the index.

Input formats

Command              Formats accepted
───────────────────  ──────────────────────────────────────────────────────────────
index, superkmer     FASTA (.fa .fasta), FASTQ (.fq .fastq), GenBank flat file
                     (.gb .gbk .gbff), all optionally gzip-compressed.
                     Directories expanded recursively. Streaming stdin via -.
query                FASTA, FASTQ, optionally gzip-compressed. Stdin via -.

Non-ACGT characters act as hard breaks between k-mer segments in all formats.

Commands

Command    Role
─────────  ────────────────────────────────────────────────────────────────────
index      Build a genome index from sequence files.
           Runs scatter → dereplicate → count → layered MPHF.
           Resumes automatically if interrupted.
merge      Merge multiple independently built indexes into one.
           Schedules partitions largest-first under a memory budget semaphore
           to avoid OOM on machines with many cores. The worst partition runs
           alone first to calibrate the expansion estimator; subsequent
           partitions run in parallel within the budget.
           --budget-fraction F  fraction of available RAM to use as budget
                                (default 0.5; reduce if OOM persists).
filter     Filter and compact an existing index: apply count thresholds,
           drop layers, rewrite as a single-layer index.
reindex    Convert evidence in-place across all layers:
           exact (evidence.bin) ↔ approximate (fingerprint.bin).
           Does not touch the MPHF or unitigs.
query      Query an index with FASTA/FASTQ sequences.
           Annotates each sequence with per-genome k-mer match counts
           and optional per-position coverage vectors (--detail).
           Parallel over sequence chunks.
phylo      Compute pairwise evolutionary-distance proxies between all
           indexed genomes (Bray-Curtis, Jaccard, etc.), optionally
           a Newick NJ/UPGMA tree, and optionally a central-position
           SNP/Sankoff calibration with exports for external
           phylogenetic tools (TNT, PhyG, IQ-TREE).
annotate   Add or update genome metadata (taxonomy, etc.) from a CSV
           file; or dump the current metadata as CSV.
estimate   Dry-run: resolve and print approximate-index parameters
           (z, evidence bits b, FP rates) given any two of (b, z, fp).
           Does not touch any index.
dump       Dump all indexed k-mers as CSV with per-genome counts or
           presence flags.
superkmer  Extract superkmers from a sequence file and write to stdout.
           Diagnostic / pipeline use.
unitig     Dump the unitig sequences stored in a built index. Debug use.
utils      Miscellaneous utilities.
           --new-label NEW=OLD      rename a genome label in-place.
           --bits-per-kmer          print MPHF / evidence / matrix size breakdown.
           --stats                  per-genome k-mer counts as CSV.
           --partition-stats        partition size distribution across one or more
                                    indexes (markdown report to stdout). Useful to
                                    diagnose minimizer imbalance before a large merge.
           --csv FILE               write per-(partition, source) raw data to FILE
                                    (used with --partition-stats).

Quick start

# Build an exact index for each genome independently
obikmer index --kmer-size 31 --label genome_a genome_a.fa --output index_a/
obikmer index --kmer-size 31 --label genome_b genome_b.fa --output index_b/

# Merge into a single multi-genome index
obikmer merge --output index/ index_a/ index_b/

# Convert to approximate index (z=5, 8-bit fingerprints)
obikmer reindex --approx -z 5 --evidence-bits 8 index/

# Query reads
obikmer query index/ reads.fq.gz > annotated.fa

# Pairwise distances
obikmer phylo index/ > distances.tsv

Parameter constraints

Parameter              Constraint
─────────────────────  ──────────────
k  (--kmer-size)       odd, 11 ≤ k ≤ 31
m  (--minimizer-size)  odd, 3 ≤ m ≤ k−1
z  (-z, --approx only) 1 ≤ z ≤ k−1

Documentation

Extended architecture and implementation notes are in DevDocMD/. Build with make doc (requires Python + MkDocs Material). User-facing documentation is in UserDocMD/, built with make doc-user.

S
Description
No description provided
Readme
382 MiB
0 Stars 1 Watchers 0 Forks
v1.4.7
Latest
2026-11-20 13:22:55 +00:00
Languages
HTML 50.3%
Rust 42.4%
Shell 2.2%
Python 2.1%
TeX 1.9%
Other 1%