AstroDynamiXOffline Scientific & Telemetry Software
Biophysics & Critical Infrastructure · Division 01 · Research Use Only (RUO)

PRISM Structural Variant Triage & Infrastructure Screening Solvers

Self-contained Pure-Rust analytical solvers for Linux and macOS (arm64 / x86_64; U.S. provisional patents pending; zero shared source code with Division 02) built without HTTP/TLS client crates for air-gapped customer enclaves to separate protein folding destabilization from catalytic-pocket disruption (Research Use Only) and screen power-grid / space-weather hazards.

Data-Source Key:Observational Public MaveDB, ClinVar & NASA OMNI2 recordsReference Model Deterministic physics on PDB / reduced grid modelsSynthetic Finite-element simulation benchmarks
MaveDB Core Rank EffectObservational
|ρ| = 0.20–0.43
Single-channel & core-abundance Spearman |ρ| across 48,628 variants in 8 human genes (pocket-vs.-activity partial |ρ| = 0.165–0.317 in 3 paired sets; all 11 CSVs ledgered; cutoffs set on 3,143 PTEN)
Epistasis AtlasReference Model
3.04M Pairs
Non-additive double-mutant & cis-haplotype screen across 45 PDB targets + 5.44M single-missense variants on arm64
OMNI2 Backward HoldoutObservational
0.9525 AUC
2005–2014 intense storm vs. quiet holdout (calibrated on 2015–2024; n=60; 95% CI [0.8977, 0.9911]; ~54 min L1 transit)
Reduced DC GridsReference Model
9.67M in 56 ms
Linearized DC screening across 5 reduced North American reference topologies (230 buses, 461 lines; 5.8 ns/eval) on arm64
01 // Biophysical Variant Triage Platform (RUO)

PRISM: Separating Folding Stability from Catalytic Tolerance

Engineered as an offline pre-filter alongside existing AlphaFold, Rosetta, SPR, and NanoDSF workflows: in self-reported internal studio benchmarks on 8-core arm64 (reproduction harness included in NDA evaluation pack), single-chain normal-mode scoring completes in ~0.2–0.6 s per structure, while a 10-variant countermeasure triage batch—combining multi-chain complex assembly, epistatic pair screening, 3-scaffold candidate docking, and 50-frame 3D relaxation trajectories—completes in under 15 minutes.

Observational MaveDB · 3 Post-Lock H1 Abundance CSVs + 3 Ledgered Oct 8 EDT Activity CSVsAnalytical Normal-Mode Solver (No Neural Network; Cutoffs Set Oct 14 on PTEN)

Dual-Channel Paired-Assay Effect Sizes Across 3 Pharmacogenes (n = 7,585 Human Variants)

Full 2×2 Spearman rank correlation magnitudes (|ρ|) from Division01_PRISM_MaveDB_Run_Log.txt (7aa8359d...): Core channel exceeds Pocket channel on VAMP-seq cellular abundance (H1: Core→Abund 0.3416–0.4282 vs. cross-term Pocket→Abund 0.0889 [CYP2C9], 0.2044 [NUDT15], 0.2183 [VKORC1]), while Pocket channel exceeds Core channel on enzymatic activity controlling for abundance (H2: partial Pocket→Act|Abund 0.1646–0.3172 vs. cross-term partial Core→Act|Abund 0.0552 [CYP2C9], 0.2425 [NUDT15], 0.1657 [VKORC1]). H1 uses the 3 abundance CSVs fetched Oct 16 (+12s to +14s post-lock, Items 11–13); H2 combines them with the 3 activity CSVs cached Oct 8 EDT (Items 1–3; self-reported studio workflow). Because variants within a protein share 3D neighborhoods, |ρ| and 2,000-resample variant-level bootstrap 95% CIs (unclustered; intervals may be optimistic under within-protein structural dependence) are primary over the 10,000-shuffle resolution floor (p = 1.0×10⁻⁴ via (0+1)/(10,000+1)). Bar widths round nearest-integer against 0.4282 max:

VKORC1 (n=650) · Core→Abund
|ρ| = 0.4282
VKORC1 · Pocket→Act|Abund
|ρ| = 0.3172
NUDT15 (n=2,655) · Core→Abund
|ρ| = 0.3622
NUDT15 · Pocket→Act|Abund
|ρ| = 0.3117
CYP2C9 (n=4,280) · Core→Abund
|ρ| = 0.3416
CYP2C9 · Pocket→Act|Abund
|ρ| = 0.1646
Table 1 — MaveDB Deep Mutational Scanning Variant Reconciliation & Self-Reported Local Caching Timeline (51,771 Experimental Variants)
Cohort & Local Caching TimelineTarget Disease & Pharmacogenes (All 11 Non-PTEN CSVs Ledgered)Variant Count (pLDDT ≥ 70)Primary & Cross-Term Spearman |ρ|, 95% CIs & Shuffle Floor
Paired Dual-Assay Cohort
H1 Abundance Fetched Oct 16 (+12s–14s Post-Lock)
CYP2C9 (4,280), NUDT15 (2,655), VKORC1 (650) — all 3 paired enzymatic activity CSVs cached Oct 8 EDT (Items 1–3); all 3 paired VAMP-seq abundance CSVs fetched Oct 16 (+12s to +14s post-lock, Items 11–13; self-reported studio workflow)7,585H1 Core→Abund ρ CIs: CYP2C9 [−0.370, −0.316], NUDT15 [−0.395, −0.327], VKORC1 [−0.484, −0.366] (vs. Pocket→Abund |ρ| 0.0889–0.2183); H2 partial Pocket→Act|Abund |ρ| = 0.1646–0.3172 (vs. Core→Act|Abund |ρ| 0.0552–0.2425; floor 1.0×10⁻⁴)
Oct 8 EDT Cached Single-Assay
All 5 CSVs Ledgered, Scored Oct 16
TPMT (3,510, abundance), HMBS (4,996, activity), CBS (6,584, activity), MSH2 (14,838, LOF), MTHFR (11,115, activity) — all 5 single-assay CSVs cached Oct 8 EDT (Items 4–8) prior to Oct 14 PTEN calibration and scored Oct 16 without cutoff changes41,043Core |ρ| = 0.1963 (MTHFR) to 0.4254 (TPMT, 95% CI [−0.455, −0.397]; shuffle floor 1.0×10⁻⁴ across all 5 targets)
Exploratory Calibration
Evaluated Oct 14 (Set Cutoffs)
PTEN dual-assay MaveDB — used on Oct 14 (Item 9) to fix 3-channel Z-normalization & interaction-shell cutoffs (internal cutoff parameters withheld) prior to Oct 16 evaluation3,143ρ(Core, Abund) = −0.3472 [−0.379, −0.316]; partial ρ(Pocket, Act | Abund) = −0.3285 [−0.361, −0.295]
Total MaveDB DMS Variants Scored9 Experimental Human Disease & Pharmacogene Targets51,771 (48,628 Non-PTEN + 3,143 PTEN)Exact arithmetic sum; separate 86,088-variant ClinVar exploratory screen reported below (excluded from 51,771 sum)
Observational · Exploratory

75-Gene Oncology ClinVar Atlas

Reported separately from the 51,771 MaveDB total: PRISM scored 86,088 ClinVar missense variants across 75 oncology targets on Oct 14. Across the n = 7,244 high-confidence (pLDDT ≥ 70) expert-reviewed Pathogenic (5,610) vs. Benign (1,634) subset (77.4% base rate), composite scoring achieved modest ROC-AUC 0.6377 (95% CI [0.6225, 0.6517]; (0+1)/(2,000+1) label-shuffle p = 5.0×10⁻⁴; continuous ROC-AUC only—binary PPV vs. the 77.4% base rate was not computed in this exploratory log).

Three-Scaffold Exclusivity

Disjoint Countermeasure Classes

For CBRN medical defense, bioscavenger stabilization, and Soldier host-resilience research, stabilization sites are prioritized in silico across three independent structural scaffold classes (Class A helical bundles, Class B macrocyclic peptidomimetics, Class C conformation-locked beta-hairpins), delivering ranked 3D coordinate dossiers for partner assay design.

Regulatory Boundary · RUO

Strict Dry-Lab Scope & Honesty

AstroDynamiX supplies offline computational physics software only: we never supply physical vials or wet-lab reagents, and all outputs constitute computational structural evidence for research triage—never standalone clinical diagnostic determinations.

Step 01 · Mutual NDA

10-Variant Blind Benchmark

Provide 10 blinded variants under NDA; receive stability, pocket, and epistasis rankings plus 50-frame in silico relaxation cards.

U.S. Provisional Patents Pending
Step 02 · Retrospective Audit

Head-to-Head Evidence Pack

Compare Spearman |ρ|, ROC-AUC, and bootstrap 95% CIs against standard homology baselines on your target family.

Deliverable: SHA-256 Report
Step 03 · NDA / CRADA / OTA

Air-Gapped Binary Pilot

Deploy a time-bounded, self-contained native Rust CLI binary inside your network-isolated enclave on proprietary PDB/VCF cohorts (built with zero HTTP/TLS client crates).

Targets: Linux / macOS (arm64, x86_64)
Step 04 · Commercial or Gov

RUO Seat or Scaffold Option

License annual offline R&D seats or target-specific exclusivity across Scaffold Class A, B, or C under commercial option or 10 U.S.C. § 4022 OTA.

100% Founder-Originated IP
02 // Space-Weather, Power-Grid & Finite-Element Systems

UMMO Space-Weather Backtest & Reduced-Grid Screening

Division 01's phase-transition and network solvers extend to reduced-model DC contingency screening, L1-to-ground geomagnetic storm discrimination, and explicitly labeled 3D finite-element structural simulations.

Observational Backtest

UMMO 20-Yr NASA OMNI2

Caveats & Design: Backward-decade holdout (calibrated on 2015–2024 n = 78 windows, tested on earlier 2005–2014 n = 60 windows across 175,320 NASA OMNI2 hrs; protocol digest 799d0751...) discriminating intense storms (20 windows Dst ≤ −100 nT) from quiet sun (40 windows |Dst| ≤ 15 nT, excluding intermediate moderate activity). The +54.17 min mean lead reflects physical L1-to-Earth solar-wind ballistic transit; skill lies in 84.21% precision (2.53× lift over 33.33% base rate) and 0.9525 ROC-AUC ([0.8977, 0.9911]). Across N = 200 Fourier surrogates (empirical floor 1/201 = 0.0050), standard Benjamini–Hochberg applied within each decade cohort separately (m = 60 in 2005–2014; m = 78 in 2015–2024) yields q_BH = 0.0373 ((60/8)×(1/201), 8-way rank tie) for May 15, 2005 (Dst −247 nT, DOY 135 H08) and 0.0663–0.0924 across the other 8 highlighted storms (from (60/18)×(4/201) = 0.0663 at rank 18/60 to (78/42)×(10/201) = 0.0924 at rank 42/78; plain BH restricted solely to those 9 excerpted storms gives 9×(1/201) = 0.0448, a narrow < 0.05 margin).

Reference Model + n=1 Case

Linearized EHV Grid Screening

Algorithmic kernel benchmark across 5 reduced North American reference topologies (230 total buses, 461 lines — reduced test models, not 10,000+-bus operational utility interconnects): Eastern (16B/28L), Western (16B/28L), ERCOT (8B/11L), Quebec (6B/7L), and Continental Multi-Scale EHV (184B/387L), screening 9,668,081 linearized DC contingencies (461 N-1, 75,523 N-2, 9,592,097 N-3) in 56 ms on arm64 (5.8 ns/eval; 5.84 ns/N-3), plus an n = 1 historical case study on the Aug 10, 1996 WSCC breakup (+107 min).

Synthetic FEA Simulation

Geodesic Lattice Mechanics

Simulation Benchmark: 6-DOF 3D Timoshenko beam-frame finite-element and WKB ray-tracing simulations on arm64 evaluating bio-inspired geodesic hull buckling depth, porous coastal wave-load attenuation, and acoustic scattering prior to physical scale-model fabrication.

Phase 2 Architectural Slot · One-Way Read-Only FeedStandby

UMMO Public Space-Weather Summary Slot (Scheduled Phase 2)

Reserved for static, SHA-256-checksummed public NOAA/NASA L1 space-weather summary snapshots generated offline—exposing zero interactive solver endpoints or proprietary parameters.

Division 01 SHA-256 Raw Input, Spec & Output Ledger (20 Digests)Public MaveDB Inputs Independently Verifiable · Studio Artifacts Under NDA

Complete 11-CSV Public MaveDB Input Anchors, Self-Sealed Spec & Aliased Studio Outputs

100% Non-PTEN MaveDB Input Coverage & Alias Disclosure: Items 1–8 and 11–13 are all 11 unmodified public MaveDB CSV files spanning 100% of the 48,628 non-PTEN human variants (displayed under their exact upstream urn_mavedb_*_scores.csv filenames, downloadable from mavedb.org for independent third-party shasum -a 256 verification): Items 1–3 are all 3 paired enzymatic activity CSVs and Items 4–8 are all 5 single-assay CSVs cached locally on Oct 8 EDT (2026-10-09 01:21:51–03:54:40 UTC = Oct 8 21:21–23:54 EDT, prior to Oct 14 PTEN calibration in Item 9), while Items 11–13 are all 3 paired H1 VAMP-seq abundance CSVs fetched on Oct 16 at 17:30:40–17:30:42 UTC (+12s to +14s after locking Item 10 at 17:30:28 UTC, prior to the 17:33:52 UTC run log in Item 14). Internal studio logs and dossiers (Items 9–10 and 14–20) use uniform Division01_* public display aliases for private studio workspace paths; SHA-256 digests, exact byte sizes, and local UTC mtimes are recorded below so public MaveDB inputs can be independently verified via shasum -a 256 while internal studio artifacts are supplied under mutual NDA. Local filesystem mtimes are mutable OS metadata corroborating our self-reported studio workflow rather than external RFC 3161 cryptographic timestamps:

1. Public MaveDB CYP2C9 Activity CSV (Oct 8 EDT Cache)
urn_mavedb_00000095-a-1_scores.csv · 1,487,212 B · 2026-10-09 01:21:51 UTC5eadb2da3afb6ab8eb820baad71070707c4d1dcf6ea441ab8e19d85193a5655e
2. Public MaveDB NUDT15 Activity CSV (Oct 8 EDT Cache)
urn_mavedb_00000055-b-1_scores.csv · 255,738 B · 2026-10-09 03:47:22 UTC5685a9f55cd5d9235efe293cad15e1faee029104ff43e4e8a345a3c375d59447
3. Public MaveDB VKORC1 Activity CSV (Oct 8 EDT Cache)
urn_mavedb_00000078-a-1_scores.csv · 75,167 B · 2026-10-09 03:51:57 UTC630c3965325635b14bc53b61d1e9f9adbe3f4639691e7c8bd7ae5340b0196a56
4. Public MaveDB MSH2 Single-Assay CSV (Oct 8 EDT Cache)
urn_mavedb_00000050-a-1_scores.csv · 1,394,308 B · 2026-10-09 03:52:01 UTC83b2a4a6692b32b7f14b89589eb858ceb5aae2e5941065b387be058adca2e1b9
5. Public MaveDB HMBS Single-Assay CSV (Oct 8 EDT Cache)
urn_mavedb_00000108-a-1_scores.csv · 553,566 B · 2026-10-09 03:52:05 UTC8f116060e70d0e0d7efd12c0a253c5667440b5f439378babe52f29d7d71c2ddc
6. Public MaveDB CBS Single-Assay CSV (Oct 8 EDT Cache)
urn_mavedb_00000005-a-5_scores.csv · 1,336,001 B · 2026-10-09 03:52:11 UTC47d2da4bd1c1368b5664c5f854444ca579cadea8a716eb549cfb59c12853fde5
7. Public MaveDB MTHFR Single-Assay CSV (Oct 8 EDT Cache)
urn_mavedb_00000049-a-4_scores.csv · 1,287,131 B · 2026-10-09 03:54:13 UTC42885269f1c222a37451bc0880010c33f54f42ce70e731b09cbfdfe557488561
8. Public MaveDB TPMT Single-Assay CSV (Oct 8 EDT Cache)
urn_mavedb_00000013-b-1_scores.csv · 1,541,324 B · 2026-10-09 03:54:40 UTCed21a34678559f21919339d982bfc2f297b27a7078eb55e50393afa984cedf66
9. Exploratory PTEN [n=3,143] & ClinVar Log (Alias)
Division01_PTEN_ClinVar_Calibration_Log.txt · 2,888 B · 2026-10-14 18:57:21 UTCee58bff7b12150a1857de7a6bdb3cac1c2c189848489777329001a6ed487610e
10. PRISM Self-Sealed Specification [Oct 16 Lock] (Alias)
Division01_PRISM_Self_Sealed_Spec.md · 8,435 B · 2026-10-16 17:30:28 UTCcb367d60825c2c45e60c5ab7aa2535e4833a727e357eb644c9711492b2bc90c6
11. Public MaveDB CYP2C9 Abundance CSV [+12s Post-Lock]
urn_mavedb_00000095-b-1_scores.csv · 1,362,701 B · 2026-10-16 17:30:40 UTC773c9785b9f2a74c42ea5752369743815c02376d257d02c3764506835bfd6d90
12. Public MaveDB NUDT15 Abundance CSV [+14s Post-Lock]
urn_mavedb_00000055-a-1_scores.csv · 257,695 B · 2026-10-16 17:30:42 UTCf42c775fa969057a720e083a1c20453450d17a0016a1bcec32e5009201450acd
13. Public MaveDB VKORC1 Abundance CSV [+14s Post-Lock]
urn_mavedb_00000078-b-1_scores.csv · 295,802 B · 2026-10-16 17:30:42 UTCd0d7bb93680c80a16b5a31b63aa517736502d736203bb3fc45fd5fdeecc01e78
14. PRISM Non-PTEN MaveDB Run Log [48,628] (Alias)
Division01_PRISM_MaveDB_Run_Log.txt · 19,658 B · 2026-10-16 17:33:52 UTC7aa8359d4ae6e83e05f8df2874cf5fbd03be0840a951ab567014372701f53788
15. PRISM 3-Channel MaveDB & ClinVar Dossier (Alias)
Division01_PRISM_TriChannel_Dossier.md · 19,881 B · 2026-10-18 20:30:33 UTC16be02bf3b9de4deb32581c94e3f168d3b3d525f6d565db4de1843992edadd73
16. PRISM 3.04M-Pair Epistasis & 5.44M Atlas (Alias)
Division01_Epistasis_Atlas_Dossier.md · 14,735 B · 2026-10-21 02:47:23 UTC5e6d37fcd7612f1c9d80c713f7c4e7caf33c8288bb750dbf522fc9b4ae8f9f43
17. UMMO 20-Yr NASA OMNI2 Backtest Dossier (Alias)
Division01_UMMO_OMNI2_Holdout_Dossier.md · 17,463 B · 2026-10-18 20:33:40 UTCfc2ddb7afe32663d41db552acf792f52ba5b85fb60f743b00dc56190ed370b71
18. 5-Grid 9.67M-Contingency Run Log (Alias)
Division01_5Grid_DC_Contingency_Log.txt · 3,452 B · 2026-10-18 18:57:59 UTCd1404f89c63323c1505a73aa4c68ce71679e0ad80ab1995a24587877d3fd7378
19. Geodesic Hull & Acoustic FEA Suite (Alias)
Division01_Geodesic_FEA_Simulation_Dossier.md · 12,639 B · 2026-10-21 02:04:04 UTC020117c7093bce76f34cc80bcfc17c7b8edcb2194a2d75f8159a8ba1249137cd
20. UMMO 2005–2014 Blind Holdout Protocol Lock
Division01_UMMO_2005_2014_Holdout_Protocol.md · 3,299 B · 2026-10-15 00:42:36 UTC799d075169489b21cf5604c306404e13118a15334bdedda5c43103a376ec61ef