Anderson in Immunological Signaling

Date: March 2, 2026 (Sessions 105–108) Status: Computational implementation COMPLETE — all nS-601..605 experiments validated. 💧♨️ wetSpring V92D+: Exp273-279 (157/157 immunological Anderson) + Gonzales reproductions (Exp280-286: 202/202) full three-tier. Paper-math chain complete: Exp291 Paper Control v4 (45/45) includes Gonzales P42-P47 (IC50, PK, IL-31, pruritus, three-compartment, selectivity). CPU v22 validates Hill/PK/Anderson in 0.8ms. GPU v9 proves portability. Streaming v9 confirms W↔P(QS) r=-0.924. 🔨🔥 metalForge v14 validates cross-system. Gonzales modeling (Hill dose-response, PK decay, pruritus time-series), 3D tissue lattice (multi-layer Hamiltonian, barrier promotion spectrum, three-compartment disorder), and Fajgenbaum MATRIX scoring (6 drug candidates, pathway × geometry × disorder) fully implemented and cross-validated (Python 48/48 + Rust 240/240 + 27 unit tests, GPU 4, dispatch 3, mixed hardware 7). S108: module refactored (1023→3 files: mod.rs + lattice.rs

• matrix.rs), provenance wired, scripts synced, doc sweep complete. Domain: Immunology × condensed matter physics × pharmacology × drug repurposing Novelty: No prior work applies Anderson localization to cytokine signal propagation in tissue; no prior work adds spatial geometry to drug repurposing scoring Cross-Spring: 💧♨️ wetSpring (Anderson spectral) × 🧠♨️ neuralSpring (ESN regime classifier, LSTM time series) × ⛰️♨️ groundSpring (transport, uncertainty, spectral validation)

Abstract

We extend the Anderson localization framework from microbial quorum sensing (Papers 01, 05, 06) to immunological cytokine signaling in skin tissue. The core observation: Th2 cytokines (IL-4, IL-13, IL-31) are diffusible signals propagating through a disordered biological medium (heterogeneous skin tissue with mixed cell populations). The same physics that governs autoinducer propagation through microbial communities governs cytokine propagation through inflamed tissue.

We map the atopic dermatitis (AD) disease cycle — allergen exposure → Th2 activation → cytokine release → neuro-immune itch signaling → barrier disruption → amplification — onto the Anderson framework and show that barrier disruption constitutes a dimensional promotion (inverse of the tillage dimensional collapse in Paper 06): scratching opens 3D diffusion channels through normally 2D-barrier skin, enabling cytokine signal delocalization.

We then connect this to the Fajgenbaum drug repurposing paradigm (MATRIX, ARPA-H $48.3M) by adding a spatial geometry dimension to pathway-based drug-disease scoring: a drug must both (a) target the right pathway AND (b) physically reach its target through tissue geometry. Anderson localization quantifies condition (b).

1. Source Literature — The Gonzales Catalog

1.1 Publications to Ingest

All authored or co-authored by Andrea J. Gonzales (Zoetis → MSU Pharmacology & Toxicology, 2025–present). These constitute the experimental foundation for the immunological Anderson extension.

1.2 Companion Literature (Not Gonzales-Authored)

2. The Anderson Mapping

2.1 Tissue as Anderson Lattice

2.2 Skin Layer Geometry

2.3 The AD Disease Cycle as Anderson Phase Transitions

Healthy skin:
  Epidermis (2D) → cytokines LOCALIZED (contained, homeostatic)
  Dermis (3D) → cytokines EXTENDED (but low production = no pathology)

AD initiation:
  Allergen → Langerhans → Th2 → IL-4, IL-13, IL-31 production in dermis
  Dermis (3D) → cytokines PROPAGATE to sensory nerve endings → ITCH

Barrier disruption (scratching):
  Epidermis physically breached → NEW 3D channels through barrier
  d_eff of epidermal layer INCREASES (2D → quasi-3D)
  Cytokines now propagate from dermis THROUGH barrier to surface
  External allergens now penetrate INTO dermis
  = DIMENSIONAL PROMOTION (inverse of Paper 06 tillage collapse)

Chronic AD:
  Persistent 3D channels → persistent signal delocalization
  Th2 amplification loop → increasing W (more immune cell types infiltrate)
  BUT still below W_c in 3D → signals KEEP propagating → chronic inflammation

Treatment:
  Cytopoint: removes IL-31 molecule → no signal to propagate (signal elimination)
  Apoquel: blocks JAK1 receptor → cells can't respond even if signal arrives (transduction block)
  Barrier repair: restores 2D epidermis → Anderson localization re-confines signals (geometry intervention)
  Dupilumab: blocks IL-4Rα → eliminates IL-4 + IL-13 simultaneously (receptor block)

2.4 The Dimensional Promotion–Collapse Duality

Paper 06 (no-till): Tillage is dimensional COLLAPSE (3D → 2D) → QS fails → soil ecosystem services collapse.

Paper 12 (AD): Scratching is dimensional PROMOTION (2D → 3D) → cytokine signaling delocalizes → inflammatory cascade amplifies.

Same physics, opposite direction, opposite outcome:

• In soil: losing 3D = losing coordination = BAD
• In AD skin: gaining 3D = gaining pathological propagation = BAD

The Anderson framework is agnostic — it predicts signal propagation. Whether propagation is beneficial or pathological depends on the biological context.

3. The Fajgenbaum Bridge — Geometry-Aware Drug Repurposing

3.1 Standard MATRIX Score

Fajgenbaum’s MATRIX: Score(drug, disease) = f(pathway overlap, literature evidence, molecular similarity, clinical data)

This is pathway-only. It asks: “Does the drug hit a relevant target?”

3.2 Anderson-Augmented Score

Anderson extension: Score(drug, disease, tissue) = f(pathway overlap) × g(tissue geometry, drug delivery route, molecular size)

The geometry factor g() encodes:

• Can the drug physically reach the target cell in the relevant tissue?
• What is the effective Anderson dimension of the target tissue?
• Does the drug need to cross a 2D barrier (epidermis) to reach a 3D compartment (dermis)?
• Large molecules (mAbs like Cytopoint): systemic delivery → 3D dermal access → good. Topical delivery → 2D barrier blocks → poor.
• Small molecules (oclacitinib): oral → systemic → 3D dermal access. Topical → can penetrate barrier → reaches both compartments.

3.3 Repurposing Targets for AD (Anderson-Filtered)

4. Spring Integration Plan

4.1 wetSpring Experiments (Proposed)

4.2 neuralSpring Connections

4.3 groundSpring Connections

4.4 Reproduction Targets

5. Computational Results (Session 107)

5.1 nS-601: Gonzales Dose-Response Modeling

All 6 Gonzales cytokine pathways (G2) modeled via generalized Hill equation response = E_max × [drug]^n / ([drug]^n + IC50^n). Validated n=1 (standard) and n=2 (cooperative) forms. Cytokine-specific barrier heights computed as W = ln(IC50) × scale:

Result: Barrier height ordering JAK1 < IL-31 < IL-13 confirmed computationally. All 6 dose-response sweeps monotonically increasing. Saturation at 1000× IC50 > 99.9%. Python 5/5 checks, Rust 80+ cross-language parity checks — all PASS.

5.2 nS-602: Pruritus Time-Series (Gonzales 2016 G3)

Treatment effect modeled as exponential recovery from initial suppression: score(t) = nadir + (baseline - nadir) × (1 - exp(-decay_rate × t))

• Baseline: 8.0 (untreated clinical score)
• Suppression: 70% (oclacitinib peak effect)
• Nadir: 2.4 at t=0 post-dose
• Asymptote → baseline at t→∞

Time-series validated at 0, 24, 72, 168, 336, 672 hours: monotonically recovering toward baseline. Cross-language parity to 1e-10.

5.3 nS-603: Lokivetmab Pharmacokinetics (Fleck/Gonzales 2021 G4)

• PK decay: C(t) = C_0 × exp(-k × t) where k = ln(2)/half_life
• Duration regression: duration = 10.10 × ln(dose) + 35.00
  • G4 data is perfectly log-linear (equal spacing in ln-dose and duration)
  • Exact fit: R² = 1.0, zero residual at all 3 dose levels
  • Monotonically increasing with dose (validated 0.05–4.0 mg/kg)

Errata: Prior version reported intercept = 33.28 with constant ~1.7-day bias (R² = 0.971). The G4 dose-duration data is perfectly collinear in log-space — 3 points, 2 parameters, zero residual is expected. The bias was a regression initialization error (intercept off by 1.72). Corrected to slope = 10.10, intercept = 35.00.

5.4 nS-604: Three-Compartment Tissue Lattice (3D Systems)

McCandless (2014) G6 three-compartment extension implemented:

Cross-compartment variance = 5.31 (healthy) vs 0.03 (inflamed) — inflammation homogenizes disorder across compartments, enabling cross-compartment cytokine propagation.

Tissue lattice Hamiltonian: Multi-layer Anderson matrices with configurable layer sizes and per-layer disorder. Symmetric, real eigenvalues, finite.

Barrier promotion spectrum: 5-step sweep from intact (d=2.0) to fully breached (d=3.0). Level spacing ratio r transitions through the Anderson critical region. All d_eff ∈ [2, 3], all r ∈ [0, 1].

5.5 nS-605: Fajgenbaum MATRIX Drug Repurposing

Anderson-augmented scoring: combined = pathway × geometry × (1 - 0.3 × W). Six drug candidates evaluated against AD flare and chronic profiles.

AD Flare Profile (barrier_breach=0.4, d_eff=2.7, W=0.75):

Key findings:

• Tofacitinib ranks #1 for both flare and chronic AD — direct pathway match (human equivalent of Apoquel) combined with small molecule geometry advantage
• Large mAbs (Tanezumab 148kDa, Nemolizumab 145kDa) penalized by geometry factor despite strong pathway scores
• Crisaborole (topical, 0.251kDa) benefits from chronic barrier breach (chronic geom 0.730 > flare geom 0.713)
• Trametinib ranks #5 — MEK pathway mismatch reduces overall score
• Score factorization combined = pathway × geometry_eff verified for all candidates to 1e-10

Integrated score: Dose-response at 100 nM × MATRIX = 0.909 × 0.713 = 0.648 for Tofacitinib — demonstrating full pipeline from concentration to repurposing recommendation.

5.6 Validation Summary

6. Cross-Paper Connections

Paper 01 → Paper 12

Anderson QS in microbial communities → Anderson cytokine signaling in tissue. Same math, different biology. The level spacing ratio r, disorder W, dimension d, and W_c all transfer directly.

Paper 04 → Paper 12

Sentinel microbes detect environmental perturbation via Anderson regime shift. Paper 12 extends: immune cell populations detect disease perturbation (AD flare) via the same Anderson regime shift. The ESN classifier (validated on AKD1000) can classify AD tissue state from cytokine measurements.

Paper 05 → Paper 12

Cross-species signaling in symbiotic systems (lichen, coral, rhizobia). Paper 12 extends: cross-cell-type signaling in immunological systems (Th2 → neuron, mast cell → keratinocyte, eosinophil → fibroblast). Same Anderson geometry governs whether signals reach their cross-type targets.

Paper 06 → Paper 12

No-till = dimensional collapse → QS fails → ecosystem services lost. AD scratching = dimensional promotion → cytokine delocalization → pathological cascade. Same physics, opposite direction. Duality documented.

7. Gonzales Lab + MSU Drug Discovery Collaboration

The Gonzales lab is the biological validation layer. The MSU Drug Discovery program (ADDRC) is the screening infrastructure. Together they create a complete pipeline from computational prediction to experimental validation.

Gonzales Lab:

• Empirical cytokine data (IL-31 pruritus scores, dose-response curves)
• iPSC-derived skin models across species (canine, feline, human)
• Plate-based screening infrastructure
• Drug discovery pipeline and regulatory expertise
• 18 years of Zoetis data and institutional knowledge

MSU Drug Discovery — ADDRC (Erika Lisabeth, Director):

• High-throughput screening facility in the same department (Pharm & Tox)
• 8,000+ compound library, liquid-handling robots, plate readers
• HTS assay development and drug repurposing expertise
• GREENScreen informatics for compound management and data analysis
• Gonzales referred contact (March 2026 interview)

MSU Drug Discovery — Additional (Richard Neubig, Edmund Ellsworth):

• Neubig: Rho/MRTF/SRF inhibitors for skin fibrosis and melanoma — potential cross-talk with JAK/STAT in AD barrier models
• Ellsworth: Medicinal chemistry optimization downstream of HTS hits

What we bring:

• Anderson localization spatial modeling (no one else has this for immunology)
• Fajgenbaum MATRIX drug repurposing with tissue geometry scoring (nS-605)
• Statistical analysis and ML (data science for screening data)
• Bioinformatics (sequencing analysis, NCBI pipelines)
• Automated data workflows
• GPU/NPU-accelerated diversity + regime classification

The pipeline:

Anderson-augmented MATRIX scores (nS-605, 6 candidates scored)
    → ADDRC high-throughput screening (Lisabeth, 8,000+ compounds)
    → iPSC skin model validation (Gonzales, canine/feline/human)
    → Medicinal chemistry optimization (Ellsworth)
    → Pre-clinical development

As RA III in Gonzales Lab:

• Dual contribution: bench work (molecular biology, cell culture, assays)
  • computational (Anderson modeling, data analysis, ML)
• Bridge between Gonzales lab (biology) and ADDRC (screening)
• Build internal Spring experiments from real lab data
• First opportunity to test Anderson predictions in immunological tissue
• Data science analytics for HTS data coming from ADDRC collaborations

8. Open Questions for Spring Evolution

1. What is W for inflamed vs healthy dermal tissue? (Need: single-cell transcriptomics data to compute Pielou evenness of cell populations)
2. What is the effective d_eff of barrier-disrupted epidermis? (Need: 3D imaging of AD skin to quantify channel geometry)
3. Does the Anderson W_c hold for cytokine propagation as it does for QS autoinducers? (Need: diffusion coefficient data for IL-31 in ECM)
4. Can the ESN regime classifier distinguish AD flare from healthy skin using cytokine panel data? (Need: published cytokine profiling datasets)
5. Does rapamycin’s efficacy in cytokine storms (Fajgenbaum) predict efficacy in AD via the mTOR/JAK cross-talk? (Testable with Gonzales’s iPSC models + ADDRC screening infrastructure)
6. Can the Anderson-augmented MATRIX scoring (nS-605) guide compound selection from the ADDRC’s 8,000+ library for AD-specific screening? (Need: ADDRC compound metadata + Gonzales iPSC validation assay)
7. Does Neubig’s Rho/MRTF/SRF skin fibrosis pathway cross-talk with JAK/STAT in AD barrier disruption? (Testable: screen Rho inhibitors in Gonzales iPSC models, score with Anderson geometry)
8. Can the PK duration model (nS-603) be refined with additional dose levels from companion-animal clinical data? (Need: lokivetmab outcomes at doses beyond the 3 published in G4)

9. Reading Order for This Paper

For an immunologist: §2 (Anderson mapping) → §3 (Fajgenbaum bridge) → §1 (source literature) → §4 (Spring experiments)

For a Spring developer: §4 (integration plan) → §2 (the mapping) → §5 (cross-paper connections) → §1 (papers to reproduce)

For a drug discovery researcher: §3 (Fajgenbaum bridge) → §2.3 (AD disease cycle) → §6 (collaboration potential) → §1 (Gonzales catalog)

For the interview: focus/gonzales_anderson_fajgenbaum_bridge.md (talking points) → §2.3 (the cycle) → §3.3 (repurposing table)