Phraya

Probabilistic Heuristic for Running Awesome Yield-Agnostic Alignments

Unified SNP calling for bacterial genomics. Accepts reads, assemblies, or any mixture. Zero binary dependencies. Native Rust implementation of alignment algorithms optimized for 2026 hardware (x86 AVX2/AVX-512, ARM64/Apple Silicon, optional GPU acceleration).

Status

Pre-development. PRD draft complete. Implementation not started.

Quick Start

# Not yet implemented
phraya snp --manifest samples.tsv --reference ref.fasta --strategy balanced

Features (Planned)

• Input agnostic: Illumina reads (PE/SE), assemblies, long reads (ONT/PacBio), or hybrid
• Zero dependencies: No BWA, MUMmer, samtools, or external aligners required
• Speed-sensitivity continuum: exact, balanced, fast, sketch strategies in one binary
• Platform-native performance: SIMD (AVX2/AVX-512/NEON), multi-core (Rayon), GPU (CUDA opt-in)
• Structured evidence: Full transparency for downstream filtering

Architecture

Workspace with 6 crates:

• phraya-core: Evidence types, scoring model, error types
• phraya-align: Alignment algorithms (FM-index, WFA, Smith-Waterman, SIMD kernels)
• phraya-index: Reference indexing (BWT, FMD-index, k-mer sketches)
• phraya-io: I/O (FASTQ/FASTA/VCF/BAM, .phraya native format)
• phraya-filter: Post-alignment filtering, presets
• phraya-cli: Binary CLI

Building

cargo build --release

Optional GPU acceleration:

cargo build --release --features cuda

License

MIT OR Apache-2.0

Contributing

PRD and architecture review feedback welcome. Implementation contributions once Phase 1 starts.