Rhys White

Research software + resources

Methods that expose their assumptions.

Focused software and public genomic resources for analytical steps where exact inputs, transformations, selection rules and uncertainty need to remain inspectable.

BacSelect
Selector v1.0.0 · reference panels validated

BacSelect

Bacterial genome diversity, at the scale you need.

Selects compact, reproducible panels of complete bacterial genomes from a defined public universe, using identity-blind, deterministic, diversity-seeking selection.

  • 55,306-genome frozen selector-v1 validation foundation
  • 13,765 species groups with species-abundance control
  • Nested validated reference panels from N=10 to N=500
  • Selection based on genome architecture rather than organism identity or clinical relevance
  • Versioned inputs, provenance and checksums for reproducible reuse
WebsiteScientific repositoryMethodUse in a paper

Validated foundation: selector v1.0.0 and its reference panels are validated and frozen. Dated monthly releases are versioned separately.

Current monthly release2026.09
Open current BacSelect →
Released · v0.1.1

NEXCISION

Removes coordinate-labelled rows from transposed NEXUS matrices without silently changing unrelated content.

  • Dependency-free Python
  • Deterministic JSON provenance
  • SHA-256 checksums
  • Independent oracle, synthetic, malformed-input and fault-injection validation
  • 105/105 scalability benchmark runs completed
Reachupdated monthly with boast

Automated metrics activate after the first scheduled snapshot.

Publication attention
CodeDocsPreprintValidation

Install / discover: PyPI · Bioconda · bio.tools
Archive: Zenodo

Released · v0.1.0

BRANCHSNV

Separates two branch-level properties that are often conflated: strict clade exclusivity and a substitution reconstructed on the focal phylogenetic edge.

  • Dependency-free Python
  • Rooted-tree branch analysis
  • Strict fixed-exclusive marker classification
  • Equal-cost Sankoff parsimony
  • All optimal focal-edge state pairs retained
  • Deterministic reports and SHA-256 input provenance
Reachupdated monthly with boast

Automated metrics activate after the first scheduled snapshot.

CodeDocsPreprint soonValidation

Install / discover: PyPI · Bioconda · bio.tools
Archive: Zenodo

Validation

Independent checks, adversarial cases, reproducible benchmarks.

The software repositories preserve the evidence needed to test the implementation: independent expected results, deliberately faulted alternatives, boundary cases, checksums and performance runs.