MARX SYSTEMS LLC — TECHNICAL WHITEPAPER
Recursive Coherence Field Theory
A Unified Framework for Stability, Coherence, and Phase Transitions
Eric L. Marx|Marx Systems LLC|2025
Abstract
Recursive Coherence Field Theory (RCFT) introduces a unified mathematical framework for modeling stability, coherence, and phase transitions across physical, biological, and computational systems. At its core, RCFT defines a single coherence metric — Ω (Omega) — computed from an 11-dimensional state vector that captures the interplay between structural integrity, thermodynamic favorability, and recursive self-consistency. This paper presents the theoretical foundations, the computational architecture (GAS → LIQUID → SOLID pipeline), and initial applications to protein stability prediction through the Angelika Fold system.
1. Introduction
The fundamental challenge in computational science is not the lack of data but the lack of a unifying language for describing how systems maintain coherence under perturbation. Molecular dynamics simulations, machine learning models, and empirical scoring functions each capture fragments of reality, but none provides a single, interpretable metric that answers the question: "How stable is this system, and why?"
RCFT addresses this gap by proposing that stability is not a property to be measured but a field to be computed. Drawing on principles from differential geometry, information theory, and thermodynamics, RCFT constructs a recursive evaluation framework where each measurement refines the next, converging on a coherence score Ω ∈ [0, 1] that quantifies the degree to which a system's internal states are mutually consistent.
The framework is deliberately general. While the initial application targets protein stability prediction, the mathematical architecture makes no assumptions specific to biochemistry. The same 11-dimensional state vector and three-phase pipeline can, in principle, be applied to any system where stability emerges from the interaction of multiple coupled variables.
2. Mathematical Foundations
RCFT operates on an 11-dimensional state vector Ψ that encodes the complete coherence state of a system at any given evaluation point. The components of Ψ are:
2.1 The State Vector Ψ
The 11 dimensions of Ψ are not arbitrary. Each corresponds to a measurable or computable property that contributes to overall system coherence. In the protein stability domain, these map to: hydrophobic packing density, electrostatic balance, hydrogen bond network saturation, backbone torsion strain, solvent accessibility profile, secondary structure propensity, disulfide bridge potential, sequence-structure compatibility, thermodynamic free energy estimate, evolutionary conservation signal, and recursive self-consistency (the degree to which the other 10 dimensions agree with each other).
The eleventh dimension — recursive self-consistency — is what makes RCFT recursive. It is computed as a function of the other ten, but it also feeds back into the evaluation of each, creating a fixed-point iteration that converges to the final Ω score.
2.2 The Coherence Metric Ω
Ω is defined as the normalized fixed-point of the recursive evaluation:
Ω = lim(n→∞) R(Ψₙ)
where R is the recursive coherence operator and Ψₙ is the state vector at iteration n. In practice, convergence is achieved within 3–7 iterations for most protein systems. The golden ratio φ = (1 + √5)/2 appears naturally in the convergence rate, a mathematical property that emerges from the recursive structure rather than being imposed.
Ω = 1.0 represents perfect coherence (all dimensions mutually consistent), while Ω = 0.0 represents complete incoherence. Real systems fall between these extremes, with most stable proteins scoring Ω > 0.7 and most unstable or disordered proteins scoring Ω < 0.4.
2.3 The Void Null Operator ∅⁺
A distinctive feature of RCFT is the Void Null Operator ∅⁺, which represents the baseline coherence of empty space — the minimum energy configuration against which all systems are measured. ∅⁺ is not zero; it is the coherence floor of the vacuum state, analogous to zero-point energy in quantum mechanics.
This operator serves a practical purpose: it prevents the Ω metric from being dominated by any single dimension and ensures that the recursive evaluation has a well-defined convergence target.
3. The GAS → LIQUID → SOLID Pipeline
RCFT processes data through a three-phase pipeline that mirrors physical phase transitions:
3.1 GAS Phase — Generation & Abstraction
The GAS phase ingests raw input data and generates the initial state vector Ψ₀. For protein systems, this means parsing the amino acid sequence, computing initial property estimates, and populating all 11 dimensions with first-pass values. The GAS phase is deliberately approximate — it prioritizes coverage over precision, ensuring that no dimension is left uninitialized.
3.2 LIQUID Phase — Iterative Refinement
The LIQUID phase applies the recursive coherence operator R repeatedly, refining Ψ through successive iterations. Each pass updates the state vector based on the interactions between dimensions, with the eleventh dimension (recursive self-consistency) driving convergence. The LIQUID phase is where the "recursive" in RCFT does its work — each iteration produces a more internally consistent state vector.
3.3 SOLID Phase — Crystallization & Output
The SOLID phase extracts the final Ω score and generates interpretable outputs: stability classification, confidence intervals, per-dimension contribution analysis, and (where applicable) structural predictions. The SOLID phase also performs validation checks, flagging cases where convergence was slow or where individual dimensions show high variance.
4. Application: Angelika Fold
Angelika Fold is the first production implementation of RCFT, targeting protein stability prediction. Given an amino acid sequence, Angelika Fold computes the Ω coherence score and classifies the protein as STABLE, MARGINAL, or UNSTABLE.
4.1 Current Status
The system has been benchmarked against 902 proteins from the Protein Data Bank (PDB), spanning globular proteins, membrane proteins, enzymes, and structural proteins. The current implementation successfully computes Ω scores that correlate with known stability properties — hyperthermophilic proteins consistently score higher than mesophilic ones, and known unstable mutants score lower than wild-type.
The binary classification threshold is still being calibrated. The ranking signal is strong (the system correctly orders proteins by relative stability), but the absolute threshold for STABLE vs. UNSTABLE requires further refinement against experimental ΔG data.
4.2 What We Know and What We Don't
We know that the Ω metric captures meaningful structural information. The correlation between Ω and known stability properties is consistent and reproducible across runs. We know that the GAS → LIQUID → SOLID pipeline converges reliably for standard globular proteins.
We do not yet know how the system performs on intrinsically disordered proteins, multi-chain complexes, or proteins with non-standard amino acids. We do not yet have sufficient experimental validation against calorimetric ΔG measurements. These are active areas of development, not limitations of the theory — they are limitations of the current dataset and calibration.
4.3 Comparison to Existing Methods
Angelika Fold is not a competitor to AlphaFold. AlphaFold predicts 3D structure from sequence; Angelika Fold predicts stability coherence from sequence. They answer different questions. A more appropriate comparison is to stability prediction tools like Rosetta's ddG, FoldX, or machine learning approaches like DeepDDG. Formal benchmarking against these baselines is planned but not yet complete.
5. Implementation
The core RCFT engine is implemented in Python with performance-critical sections in C++. The system runs on standard hardware (no GPU required for inference) and processes a typical protein sequence in under 2 seconds. The codebase is modular: the mathematical core is separated from the protein-specific feature extraction, allowing the framework to be adapted to other domains.
Results are deterministic — given the same input sequence and parameters, the system produces identical Ω scores across runs. This is by design: RCFT is a mathematical framework, not a stochastic model.
6. Intellectual Property
RCFT and Angelika Fold are protected under provisional patent applications filed in February and March 2025, respectively. The core algorithms, the 11-dimensional state vector architecture, and the GAS → LIQUID → SOLID pipeline are proprietary. This whitepaper describes the framework at a level sufficient for scientific evaluation without exposing implementation details that would enable reproduction of the proprietary engine.
For collaboration inquiries, licensing discussions, or access to the full technical documentation under NDA, contact Marx Systems LLC.
© 2025 ERIC L. MARX — MARX SYSTEMS LLC — ALL RIGHTS RESERVED
PATENT PENDING — PROVISIONAL APPLICATIONS FILED FEB & MAR 2025