ecoPrimals — Sovereign Scientific Computing Platform: Capability & Parity Assessment

Domain-by-domain comparison against proprietary tools across 8 scientific domains

From: ecoPrimal — human + synthetic intelligence
Organization: ecoPrimals Date: March 17, 2026 License: All code AGPL-3.0-or-later; all documentation CC-BY-SA-4.0 Repositories: Springs at github.com/syntheticChemistry · Primals at github.com/ecoPrimals · Products at github.com/sporeGarden


What This Is

ecoPrimals is a sovereign scientific computing ecosystem built in pure Rust with GPU acceleration via WebGPU (WGSL shaders). It replaces Python/R/Fortran/Java tool chains across life science, pharmacology, physics, and data provenance domains. Every claim below has a validation binary that proves it — clone the repo, run the binary, verify the output.

This document provides honest parity assessments: what we match, what we exceed, what proprietary tools still do better, and where to find everything.


1. Bioinformatics Pipeline (vs Galaxy / QIIME2 / mothur)

What It Replaces

Commercial/Open ToolAnnual CostWhat It Does
GalaxyFree (hosted) / $50K+ (local)Web-based bioinformatics workflow
QIIME2Free16S/ITS amplicon analysis (Python)
mothurFree16S OTU-based pipeline (C++)
DADA2 (R)FreeAmplicon sequence variant denoising
phyloseq (R)FreeMicrobiome statistical analysis
vegan (R)FreeCommunity ecology analysis

ecoPrimals Replacement: wetSpring

Repository: github.com/syntheticChemistry/wetSpring Status: V127 — 1,443+ tests, 306 validation binaries, 376 experiments, 5,707+ checks

CapabilityParity LevelNotes
FASTQ parsing + quality filteringFull paritySovereign parser, no needletail dependency
Read merging (paired-end)Full parityOverlap detection, quality-aware consensus
DereplicationFull parityHash-based, GPU-accelerated
DADA2 denoisingFull parityError model + denoising validated against R DADA2
Chimera detectionFull parityde novo + reference-based
Taxonomy classificationFull parityNaïve Bayes, k-mer, spectral matching
UniFrac (weighted/unweighted)Full parityGPU-accelerated, validated against phyloseq
Diversity indices (Shannon, Simpson, Chao1, Pielou)Full parityValidated to 1e-12 against textbook definitions
Bray-Curtis dissimilarityFull parityGPU BrayCurtisF64 kernel
PCoA ordinationFull parityGPU BatchedEighGpu eigendecomposition
Rarefaction curvesFull parityMonte Carlo with bootstrap CI
Smith-Waterman alignmentFull parityAffine gap penalties
HMM phylogeneticsFull parityForward algorithm, GPU batch
Newick tree parsing / Robinson-FouldsFull parityFull phylogenetic tree comparison
ODE ecological models (Lotka-Volterra, QS, etc.)Exceeds5 biological ODE systems with GPU shader generation
Anderson localization for community structureNo equivalentNovel physics framework — no proprietary tool does this
GPU acceleration (all operations)Exceeds150+ GPU primitives, zero local WGSL
Spectral cosine matchingExceeds1,077× speedup over CPU

What Proprietary/Open Tools Still Do Better

CapabilityGapPath to Parity
GUI workflow builderNo GUI — CLI + validation binaries onlypetalTongue provides visualization; Galaxy-style builder not planned
Plugin ecosystemNo third-party plugin systemIPC capability discovery enables composition
Training documentationNo tutorials, no workshopsK-Nome methodology document exists; formal curriculum pending
Multi-user web interfaceSingle-user, local executionbiomeOS IPC enables multi-client; web UI not planned
Established communityOne developer, public repos3.2M lines of Rust, 107K+ tests, all validation executable
Cloud deploymentLocal/LAN onlyNUCLEUS bonding model supports distributed deployment

Where to Find / Rebuild

git clone [email protected]:syntheticChemistry/wetSpring.git
cd wetSpring/barracuda
cargo test --workspace                    # 1,443+ tests
cargo run --release --bin validate_diversity  # Diversity index validation
cargo run --release --bin validate_dada2_full # DADA2 pipeline validation

Key modules: barracuda/src/bio/ (47 CPU + 47 GPU modules), barracuda/src/io/ (FASTQ, mzML, mzXML, JCAMP-DX parsers) Dependency: barracuda (pure math primal, path dependency to ../../barraCuda/crates/barracuda)


2. Pharmacometric Modeling (vs NONMEM / Monolix / WinNonlin)

What It Replaces

ToolAnnual CostWhat It Does
NONMEM~$2,000/yrPopulation PK parameter estimation (FOCE)
Monolix~$1,500/yrPopulation PK parameter estimation (SAEM)
WinNonlin (Phoenix)~$3,000/yrNon-compartmental analysis (NCA)
CRO population PK$50K–200K/programContract research organization modeling

ecoPrimals Replacement: healthSpring

Repository: github.com/syntheticChemistry/healthSpring Status: V35 — 613 tests, 73 experiments, 113/113 cross-validation checks, 6 WGSL shaders

CapabilityParity LevelNotes
Hill dose-response (4-parameter)Full parityValidated for JAK inhibitors (Gonzales IC50 data)
One-compartment PK (IV bolus, oral Bateman)Full parityAUC trapezoidal, steady-state accumulation
Two-compartment PK (biexponential)Full parityDistribution/elimination phase separation
mAb PK cross-species (allometric scaling)Full parityBW^0.75 CL, BW^1.0 Vd
Population PK Monte Carlo (1,000+ patients)Full parityLognormal IIV, CL-AUC correlation
PBPK (5-tissue physiological)Full parityMass conservation, hepatic clearance, tissue Kp
Michaelis-Menten nonlinear PKFull parityCapacity-limited elimination (phenytoin)
FOCE estimationNear parity30% theta recovery on synthetic data
SAEM estimationNear parity50% theta recovery on synthetic data
NCA (λz, AUC∞, MRT, CL, Vss)Full parityAll 5 standard NCA metrics
NLME diagnostics (CWRES, VPC, GOF)Full parityCWRES ~N(0,1), 50-simulation VPC
GPU population Monte CarloExceeds207 M/s throughput (RTX 4070), 100K patients

What Proprietary Tools Still Do Better

CapabilityGapPath to Parity
FOCE/SAEM on real clinical dataValidated on synthetic onlyNeed MIMIC-IV (PhysioNet credentialed access)
FDA submission formattingNo CTD/eCTD outputInfrastructure exists; formatting layer needed
Covariate model building (stepwise)Manual covariate selection onlyAutomated stepwise selection planned
Regulatory track recordNo FDA submissions yetDeterministic reproducibility is an advantage for auditors
Interactive model explorationCLI onlypetalTongue dashboard provides visualization
Existing training / certificationNo formal training programK-Nome methodology available

Where to Find / Rebuild

git clone [email protected]:syntheticChemistry/healthSpring.git
cd healthSpring
cargo test --workspace                     # 613 tests
cargo run --release --bin exp001_hill      # Hill dose-response (Gonzales IC50)
cargo run --release --bin exp004_mab       # Cross-species PK (lokivetmab → human)
cargo run --release --bin exp075_nlme      # NONMEM/Monolix/WinNonlin replacement

Key modules: ecoPrimal/src/pkpd/ (compartmental, population, NLME), ecoPrimal/src/microbiome/ (gut Anderson), ecoPrimal/src/biosignal/ (ECG, PPG, EDA) Rust vs Python: 84× aggregate speedup across 14 benchmark cases (Exp084)


3. Drug Repurposing (vs Every Cure MATRIX / ROBOKOP)

What It Replaces

ToolCostWhat It Does
Every Cure MATRIX$48.3M ARPA-H grantDrug-disease scoring (4K drugs × 18K diseases)
ROBOKOPNIH-fundedKnowledge graph for drug-disease relationships
DrugBank$20K+/yr (commercial)Drug target database
repoDBFreeDrug-disease benchmark dataset

ecoPrimals Replacement: wetSpring Track 3 + neuralSpring nS-06 + groundSpring

CapabilityParity LevelNotes
Pathway-based drug scoring (PI3K/AKT/mTOR)Full parityFajgenbaum 2019 reproduced (Exp157, 8/8)
MATRIX 4-stage pipelineFull parityPharmacophenomics pipeline (Exp158, 9/9)
NMF drug-disease factorizationFull parityYang 2020 + repoDB (Exp159–160, 16/16)
TransE knowledge graph embeddingFull parityROBOKOP-style (Exp161, 7/7)
Anderson geometry-aware drug scoringNo equivalentNovel — adds tissue penetration physics to MATRIX
Tissue lattice + barrier promotionNo equivalent3D Anderson Hamiltonian for skin/gut geometry
Cross-species PK translationPartial parityValidated canine→human; not yet feline→human
GPU compound screeningExceeds207 M/s Hill sweep on RTX 4070

What Every Cure / ROBOKOP Still Do Better

CapabilityGapPath to Parity
Scale: 4K drugs × 18K diseasesWe validate on 6 drugs × 6 diseasesData pipeline, not algorithm — ChEMBL + NCATS Translator
Curated disease pathway profilesUse published pathways onlyNCATS Translator API (same source as Every Cure)
Clinical outcome integrationPublished parameters onlyMIMIC-IV, FAERS via openFDA
Institutional backing / regulatory relationshipsOne developerMSU Drug Discovery + ADDRC collaboration
Pre-built ontology mappings (MONDO, DO)Manual mappingsMONDO/DO are open; integration work needed

Where to Find / Rebuild

# wetSpring — Track 3 drug repurposing
cd wetSpring/barracuda
cargo run --release --bin validate_fajgenbaum_pathway     # Exp157
cargo run --release --bin validate_matrix_pharmacophenomics # Exp158
cargo run --release --bin validate_nmf_drug_repurposing    # Exp159
cargo run --release --bin validate_repodb_nmf              # Exp160
cargo run --release --bin validate_knowledge_graph_embedding # Exp161

# neuralSpring — immunological Anderson + MATRIX scoring
cd neuralSpring
cargo run --release --bin validate_immunological_anderson           # 20/20
cargo run --release --bin validate_immunological_anderson_extended  # 28/28

# groundSpring — tissue Anderson drug scoring
cd groundSpring
cargo run --release --bin validate_tissue_anderson  # Exp033–034

4. Analytical Chemistry (vs MassHunter / Chromeleon / MZmine)

What It Replaces

ToolAnnual CostWhat It Does
MassHunter (Agilent)~$10K+/yrLC-MS data acquisition + analysis
Chromeleon (Thermo)~$5K+/yrChromatography data system
MZmineFreeLC-MS feature extraction
FindPFASFreePFAS mass spectrometry screening

ecoPrimals Replacement: wetSpring Track 2

CapabilityParity LevelNotes
mzML parsingFull paritySovereign XML parser (no quick-xml)
mzXML parsingFull paritySovereign parser
JCAMP-DX parsingFull paritySpectroscopy format
EIC extractionFull paritym/z window extraction from raw data
Peak detection (signal processing)Full parityCWT-based, validated against Python baseline
Spectral cosine matchingExceedsGPU kernel, 1,077× speedup
KMD groupingFull parityKendrick mass defect for homologous series
PFAS screening (mass defect + RT)Full parityFindPFAS algorithm reproduced
Retention index (Kovats, Lee)Full parityBoth linear and polynomial calibration

What Proprietary Tools Still Do Better

CapabilityGapPath to Parity
Instrument control / data acquisitionSoftware only — no instrument driversNot planned; focus is analysis
Vendor-specific raw formats (.d, .raw)mzML/mzXML only (open formats)Vendor conversion is standard practice
Method development wizardsCLI onlyNot planned
FDA 21 CFR Part 11 complianceProvenance chain exists; no formal auditBearDog signing + loamSpine certs map to Part 11
Integrated LIMSNo LIMS — provenance trio provides chain-of-custodyPaper 21 architecture covers this

Where to Find / Rebuild

cd wetSpring/barracuda
cargo run --release --bin validate_features     # EIC + peak detection
cargo run --release --bin validate_peaks         # Signal processing
cargo run --release --bin validate_pfas_decision_tree  # PFAS screening
cargo run --release --bin validate_massbank_gpu_scale  # GPU spectral matching

Key modules: barracuda/src/io/mzml/, barracuda/src/io/mzxml/, barracuda/src/bio/eic.rs, barracuda/src/bio/signal/


5. Biosignal Processing (vs LabChart / MATLAB / Python-MNE)

ecoPrimals Replacement: healthSpring Track 3

CapabilityParity LevelNotes
Pan-Tompkins QRS detectionFull parityValidated against MIT-BIH reference
HRV metrics (SDNN, RMSSD, pNN50)Full parityTime-domain from R-peak intervals
PPG SpO2 calibrationFull parityBeer-Lambert AC/DC ratio
EDA tonic/phasic decompositionPartialSovereign implementation; numpy convolution still faster for rolling average
Arrhythmia beat classificationFull parityTemplate matching: Normal/PVC/PAC/BBB
WFDB format parsingFull parityPhysioNet Format 212/16 + beat annotations
Multi-channel fusionFull parityECG + PPG + EDA → composite health assessment

6. Provenance & Data Integrity (vs LabArchives / Benchling / LIMS)

ecoPrimals Replacement: SCYBORG Provenance Trio + BearDog

CapabilityParity LevelNotes
Sample chain-of-custodyExceedsCryptographic DAG ( rhizoCrypt), not database records
Ed25519 digital signatures on resultsNo equivalentBearDog signs every computation result
ISO 17025/15189 traceability mappingArchitectural parityloamSpine certificates map to ISO requirements
Fraud detection (6 types)ExceedsGraph analysis: phantom sample, broken cold chain, etc.
Consent-gated access (medical)No equivalentPaper 22: DID-based consent certificates
Audit trailExceedsEvery DAG vertex is immutable, signed, and attributed

What LIMS / Benchling Still Do Better

CapabilityGapPath to Parity
Inventory managementNo physical inventory trackingfm-pipette (FIELDMOUSE) planned for wet lab integration
Barcode / QR scanningNo hardware integrationNestGate mobile scanning planned
Regulatory pre-validation (GxP)Architecture exists; no formal GxP auditBearDog + loamSpine architecture maps to GxP; audit needed
SaaS convenienceLocal deployment onlyNUCLEUS enables LAN/WAN; no cloud planned

7. GPU Scientific Computing (vs CUDA / Kokkos / MATLAB Parallel)

ecoPrimals Replacement: barraCuda + toadStool + coralReef

CapabilityParity LevelNotes
f64 precision GPU computeFull parityAll WGSL shaders use f64 (via DF64 emulation where needed)
Vendor-agnostic GPU targetingExceedsWebGPU: NVIDIA + AMD + Intel + Apple (CUDA is NVIDIA-only)
Sovereign shader compilerExceedscoralReef compiles WGSL without vendor toolchain
Hardware discoveryExceedstoadStool discovers GPU + CPU + NPU at runtime
Mixed hardware dispatchExceedsCPU + GPU + NPU routing by capability, not hardcoding
Neuromorphic (NPU) supportNo equivalentBrainChip AKD1000 via pure Rust driver
806+ validated WGSL shadersSpecializedBio/physics domain; not general-purpose

What CUDA / Kokkos Still Do Better

CapabilityGapPath to Parity
Raw throughput (CUDA optimized kernels)WGSL overhead for some operationsImproving with wgpu maturity
Ecosystem (cuBLAS, cuFFT, cuDNN)barraCuda covers science ops; no ML frameworkNot competing with ML frameworks
Multi-GPU scalingSingle GPU per dispatchtoadStool multi-device dispatch planned
Tensor cores / mixed precisionf64 focus, not fp16/bf16Science needs f64; ML-style mixed precision not a priority
HPC job scheduler integration (Slurm)No Slurm scriptsNUCLEUS bonding model is an alternative; Slurm adaptor trivial

8. Aggregate Ecosystem Metrics

MetricValue
Springs (validation domains)7 (wet, hot, air, ground, neural, health, ludo)
Primals (infrastructure)14 ( barraCuda, toadStool, coralReef, biomeOS, BearDog, NestGate, Songbird, sweetGrass, rhizoCrypt, loamSpine, petalTongue, Squirrel, bingoCube, FIELDMOUSE)
Total tests27,000+ across all springs
Total validation checks15,334+
Papers reproduced70+ (63 wetSpring + healthSpring + neuralSpring + hotSpring + others)
WGSL shaders ( barraCuda)806+
LanguagesPure Rust (zero C/C++/Fortran in application code)
Unsafe codeZero (#![forbid(unsafe_code)] in all spring lib crates)
External dependenciesMinimal; all pure Rust or explicit rust_backend
LicenseAGPL-3.0-or-later (code), CC-BY-SA-4.0 (docs)
Development methodologyK-Nome (Knowledge-Numeric Observed & Mentored Evolutionary Programming)
Development history69,000+ AI invocations, 51B tokens, 185-day streak
Hardware investment~$15,000 (consumer hardware, zero cloud)

9. What No Proprietary Tool Offers

These capabilities have no commercial equivalent:

CapabilityWhat It DoesSpring
Anderson localization for community structureMaps microbial diversity onto condensed matter physics; predicts signal propagation vs confinementwetSpring, groundSpring, neuralSpring
Geometry-aware drug repurposingAdds spatial tissue penetration to pathway-based drug scoring (extends MATRIX)wetSpring Track 3, neuralSpring nS-06, groundSpring
Cross-species Anderson translationSame physics, different tissue parameters — species-agnostic by constructionhealthSpring Track 6
Cryptographically signed scientific resultsEd25519 signature on every diversity index, ASV table, drug scoreBearDog
Provenance DAG for biological samplesField-to-publication chain-of-custody with fraud detectionrhizoCrypt + loamSpine + sweetGrass
NPU edge classificationBrainChip AKD1000 at 18.8K Hz, coin-cell power, pure Rust drivertoadStool + neuralSpring
Sovereign shader compilerCompile WGSL without NVIDIA/AMD/Intel toolchaincoralReef
Zero-knowledge medical provenancePatient-owned records with consent certificatesPaper 22

10. Rebuilding From Source

Every spring is a self-contained Cargo workspace. To rebuild any capability:

# Prerequisites: Rust 1.87+ (rustup.rs), git
# Optional: GPU (any Vulkan-capable card for GPU tests)

# Clone the spring you need
git clone [email protected]:syntheticChemistry/<spring>.git
cd <spring>

# Build and test
cargo test --workspace              # All library tests
cargo clippy --all-targets          # Zero warnings guaranteed
cargo run --release --bin <binary>  # Run specific validation

# GPU tests (requires Vulkan-capable GPU)
cargo test --workspace --features gpu

Build dependencies: Rust toolchain only. No Python, no R, no conda, no Docker, no pip, no npm. One cargo build compiles everything from source.

barraCuda (the math primal) is a path dependency. Clone it alongside the spring:

cd Development/ecoPrimals
git clone [email protected]:ecoPrimals/barraCuda.git  # renamed from barraCUDA
git clone [email protected]:syntheticChemistry/wetSpring.git
# Cargo.toml points to ../../barraCuda/crates/barracuda

Document History

DateChange
2026-03-17Initial capability & parity assessment (V127 wetSpring, V35 healthSpring)