Cyanea Labs

Modular bioinformatics in Rust.
Sequences, alignments, structures, stats — native, WASM, Python.

Docs • Architecture • Build Guide • Bindings • Usage Guide • Issues • Discussions

---

Cyanea Labs is a Cargo workspace of 13 crates covering the core primitives of computational biology — sequence analysis, alignment, genomic intervals, statistics, machine learning, cheminformatics, structural biology, and phylogenetics. Everything compiles to native, WebAssembly, and Python (via PyO3), with an Elixir NIF bridge for the Cyanea platform.

3,000+ tests. Zero unsafe. No heavyweight C/C++ dependencies in the core path.

Quick Start

# Check everything compiles
cargo check --workspace

# Run all tests (exclude cyanea-py unless Python env is configured)
cargo test --workspace --exclude cyanea-py

# Test a single crate
cargo test -p cyanea-seq

# Run benchmarks
cargo bench -p cyanea-align

From Rust

use cyanea_seq::DnaSequence;
use cyanea_align::needleman_wunsch;
use cyanea_stats::describe;

// Sequence analysis
let seq = DnaSequence::new("ATCGATCG").unwrap();
assert_eq!(seq.gc_content(), 0.5);

// Global alignment
let result = needleman_wunsch(b"ACGT", b"ACGTT", 1, -1, -2).unwrap();
println!("{}", result.cigar); // 4=1I

// Descriptive statistics
let stats = describe(&[1.0, 2.0, 3.0, 4.0, 5.0]).unwrap();
println!("mean={}, std={:.2}", stats.mean, stats.std_dev);

From Python

import cyanea

# Align two sequences
result = cyanea.smith_waterman("ACGTACGT", "CGTAC", 2, -1, -2)
print(result["score"], result["cigar"])

# Parse a VCF file
variants = cyanea.read_vcf("samples.vcf")

# PCA on a distance matrix
coords = cyanea.pca([[0,1,2],[1,0,3],[2,3,0]], n_components=2)

From JavaScript / TypeScript

import { align, seq, stats } from "@cyanea/bio";

const result = align.smithWaterman("ACGTACGT", "CGTAC", 2, -1, -2);
const gc = seq.gcContent("ATCGATCG");
const desc = stats.describe([1, 2, 3, 4, 5]);

Crates

Crate	What it does	Tests
cyanea-core	Error types, traits, SHA-256, zstd/gzip, mmap, log-space probability, rank/select bitvectors, wavelet matrix, Fenwick tree	58
cyanea-seq	DNA/RNA/protein sequences, FASTA/FASTQ, k-mers, 2-bit encoding, suffix array, FM-index, BWT, FMD-index, MinHash, pattern matching (KMP, Boyer-Moore, Myers bit-parallel), PSSM/motif scanning, ORF finder, codon tables, sequence masking, RNA secondary structure, protein properties, read simulation, de Bruijn graphs, assembly QC	474
cyanea-align	Needleman-Wunsch, Smith-Waterman, semi-global, banded, MSA, seed-and-extend, minimizers, WFA, POA, LCSk++, pair HMM, profile HMM, X-drop/Z-drop, spliced alignment, CIGAR utilities, substitution matrices (BLOSUM/PAM), SIMD (NEON/SSE4.1/AVX2), GPU dispatch	321
cyanea-omics	Genomic coordinates, interval sets/trees, genome arithmetic, expression matrices, sparse matrices, variants, gene annotations, coordinate liftover, AnnData/h5ad/zarr, variant annotation/VEP, CNV/CBS, methylation, spatial transcriptomics, single-cell (HVG, normalize, Leiden/Louvain, diffusion map, DPT, PAGA, markers, Harmony/ComBat/MNN)	434
cyanea-io	CSV, VCF, BED, BEDPE, GFF3, GTF, SAM, BAM, CRAM, BCF, Parquet, BLAST, BLAST XML, MAF, GenBank, bigWig, Stockholm, Clustal, Phylip, EMBL, PIR, ABI, bedGraph, GFA, indexed BAM/VCF, BAM ops, VCF ops, variant calling, fetch clients	357
cyanea-stats	Descriptive stats, correlation, hypothesis tests (t, chi-squared, Mann-Whitney, Fisher, KS), distributions, PCA, effect sizes, Bayesian conjugate priors, combinatorics, population genetics (Fst, Tajima's D, LD), differential expression, enrichment (GSEA, ORA), ordination (PCoA, NMDS), multivariate tests (PERMANOVA, ANOSIM), survival analysis, ecological diversity	384
cyanea-ml	K-means, DBSCAN, hierarchical clustering, pairwise distances, KNN, PCA, t-SNE, UMAP, random forest, GBDT, feature selection, HMM, classification metrics, cross-validation	269
cyanea-chem	SMILES/SDF V2000/V3000, SMARTS, molecular fingerprints (Morgan, MACCS), substructure search, stereochemistry, canonical SMILES, 200+ descriptors, drug-likeness (Lipinski, QED, PAINS), scaffolds (Murcko, MCS), 3D conformers (ETKDG), force fields (UFF, MMFF94), Gasteiger charges, chemical reactions (SMIRKS), standardization	200
cyanea-struct	PDB/mmCIF parsing, 3D geometry, DSSP secondary structure, Kabsch superposition, contact maps, Ramachandran analysis	76
cyanea-phylo	Newick/NEXUS parsing, distance matrices, UPGMA/NJ, Fitch/Sankoff parsimony, ML likelihood (GTR+G), bootstrap, tree search (NNI/SPR/TBR), model selection (AIC/BIC), protein models (LG/WAG/JTT), Bayesian MCMC, species tree (ASTRAL), UniFrac, simulation, consensus, dating, drawing	225
cyanea-gpu	Backend trait with CPU, CUDA, Metal, and WebGPU implementations, GPU buffer management, k-mer counting, Smith-Waterman, MinHash, benchmarks	62
cyanea-wasm	WebAssembly bindings via wasm-bindgen (seq, io, align, stats, ml, chem, struct_bio, phylo, omics, core)	223
cyanea-py	Python bindings via PyO3 (seq, align, stats, ml, chem, struct_bio, phylo, io, omics, sc) with optional NumPy support	—

Architecture

See docs/ARCHITECTURE.md for the full architecture guide with ASCII diagrams, feature flag details, data flow pipelines, and platform support matrix.

cyanea-core (foundation)
├── cyanea-seq          Sequences, indexing, k-mers
├── cyanea-align        Pairwise & multiple alignment
├── cyanea-omics        Genomic intervals, matrices, variants
├── cyanea-io           File format I/O
├── cyanea-stats        Statistics & distributions
├── cyanea-ml           Machine learning & clustering
├── cyanea-chem         Chemical structure & fingerprints
├── cyanea-struct       Protein structure & geometry
├── cyanea-phylo        Phylogenetics (optional: cyanea-ml)
├── cyanea-gpu          GPU compute backends
├── cyanea-wasm         → WebAssembly (@cyanea/bio on npm)
└── cyanea-py           → Python (PyO3 + maturin)

Every domain crate depends only on cyanea-core. The binding crates (wasm, py) aggregate the domain crates to expose a unified API. The Elixir NIF bridge lives in a separate repo and depends on these crates via path.

Feature Flags

All domain crates default to std. Opt into additional capabilities:

Flag	Scope	What it enables
parallel	align, ml, stats, chem, struct, phylo, io	Rayon parallelism
simd	align	SIMD-accelerated alignment
cuda	gpu, align	CUDA GPU backend (requires CUDA toolkit)
metal	gpu, align	Metal GPU backend (macOS)
wgpu	gpu	WebGPU backend (cross-platform)
blas	ml, stats	BLAS-backed PCA
wfa	align	Wavefront alignment
minhash	seq	MinHash / FracMinHash sketching
sam	io	SAM format support
bam	io	BAM format (implies sam)
cram	io	CRAM format (implies sam)
parquet	io	Apache Parquet columnar format
h5ad	omics	HDF5-backed AnnData I/O
zarr	omics	Zarr v3 I/O
single-cell	omics	Single-cell analysis pipeline (Leiden, HVG, pseudotime, integration)
serde	all	Serialization support
wasm	wasm	wasm-bindgen annotations
numpy	py	NumPy array interop

Building Bindings

See docs/BUILDING.md for the complete build guide covering all targets, feature combinations, and troubleshooting.

WebAssembly

cargo check -p cyanea-wasm --features wasm
wasm-pack build cyanea-wasm --features wasm

Published as @cyanea/bio on npm. TypeScript types and a Web Worker API are included in cyanea-wasm/ts/.

Python

PYO3_USE_ABI3_FORWARD_COMPATIBILITY=1 cargo check -p cyanea-py
cd cyanea-py && PYO3_USE_ABI3_FORWARD_COMPATIBILITY=1 maturin develop --release

Elixir NIF

cargo check -p cyanea-native   # Check only (linking requires BEAM)
cd ../cyanea && mix compile     # Full build via Rustler

Cross-Crate Pipelines

See docs/GUIDE.md for complete cross-crate workflow examples: FASTQ→alignment→variants, single-cell analysis, cheminformatics, phylogenetics, population genetics, and microbiome analysis.

Contributing

# Full CI check
cargo fmt --all -- --check
cargo clippy --workspace --exclude cyanea-py
cargo test --workspace --exclude cyanea-py

Using Claude Code? See CLAUDE.md for project context, architecture, and coding conventions.

License

Apache-2.0