wetSpring — Life Science & Analytical Chemistry

16S metagenomics, LC-MS, PFAS screening — 5,707+ checks across 376 experiments, 63/63 papers reproduced in sovereign Rust

Domain

16S metagenomics, LC-MS feature extraction, PFAS screening, microbial ecology.

Repository: syntheticChemistry/wetSpring

The Science Story

wetSpring proves barraCuda can replace the Galaxy/QIIME2/Python bioinformatics stack with sovereign Rust. The complete 16S pipeline — FASTQ → quality → merge → dereplicate → DADA2 → chimera → taxonomy → diversity → UniFrac — runs in Rust with 1 runtime dependency (flate2 for gzip). The sovereign XML parser eliminates quick-xml; the sovereign FASTQ parser eliminates needletail.

63/63 papers reproduced across 4 research tracks: Waters c-di-GMP/QS, Liu comparative genomics, deep-sea metagenomics, Jones PFAS. 50/50 three-tier eligible papers have full CPU + GPU + metalForge validation.

Headline Results

  • 5,707+ checks across 376 experiments — all passing
  • 1,077× GPU speedup for spectral cosine matching
  • 30 sovereign bio modules, 1 runtime dependency
  • Public benchmark against 4 BioProjects (22 samples) — all match paper ground truth

Validation Phases

PhaseKey Result
1-2 (Galaxy→Rust)30 sovereign bio modules, 135/135 checks
3-7 (GPU pipeline)Complete 16S on GPU. 1,077× spectral cosine speedup
V53+ (Anderson QS)52/52 papers, Anderson localization applied to biology
V86 (Cross-spring)23/23 across 5 springs. -4,753 net lines (deep debt elimination)

Researchers Reproduced

ResearcherDepartmentDomain
Christopher WatersMMG, MSUQuorum sensing, c-di-GMP
Kevin LiuCMSE, MSUComparative genomics, phylogenetics
Jesse Cahill & Chuck SmallwoodBioscience, SandiaBiosurveillance
A. Daniel JonesBMB/Chemistry, MSUMass spectrometry, PFAS
Rika AndersonBiology, CarletonVent metagenomics, pangenomics

What the Constraint Revealed

Zero local WGSL — every GPU operation delegates to barraCuda via ToadStool. wetSpring consumes 79 barraCuda primitives without duplicating any math. The three-tier validation pattern (CPU → GPU → metalForge) was pioneered here and adopted across all springs.

wetSpring found and fixed the log_f64 bug in ToadStool (coefficients halved, causing 1e-3 instead of 1e-15 precision) during Shannon entropy validation — the spring improved the infrastructure it depends on. Also resolved 4 barraCuda gaps: ODE solver, Gillespie stochastic sim, HMM Viterbi, Smith-Waterman alignment.

Cross-Spring Connections

  • → ToadStool: log_f64 bug found and fixed; 79 primitives consumed; three-tier validated
  • → airSpring: kriging spatial interpolation; dynamic Anderson W(t) models soil moisture coupling
  • → hotSpring: Anderson localization applied to biology — shared spectral primitives
  • → neuralSpring: ESN/LSTM anomaly detection for sentinel microbes; NPU int8 quantization validated
  • → healthSpring: diversity indices, Anderson lattice → gut colonization resistance, 16S pipeline
  • → groundSpring: sequencing noise calibrates rarefaction; 86 named tolerances with provenance

Public Notebooks

Interactive Jupyter notebooks that visualize wetSpring’s frozen experiment data:

  • 01 — 16S Pipeline Validation: Galaxy bootstrap, Track 2 LC-MS, R/vegan diversity parity, NCBI real data
  • 02 — Python vs Rust vs GPU: Benchmark timings, speedup charts, energy consumption
  • 03 — Paper Reproductions: 63/63 papers across 5 researchers and 6 tracks
  • 04 — Cross-Spring Connections: 79 barraCuda primitives, constraint-driven discoveries
  • 05 — Soil Anderson Deep Dive: Track 4 domain exemplar, QS-pore geometry, chemotaxis

Clone the repository, cd notebooks/, jupyter lab. Or access via JupyterHub.

baseCamp Papers

Papers 01, 03, 04, 05, 06 — see baseCamp Science for full list.