AlphaFold Substrate Alignments#

Terminology#

This appendix defines the terminology used throughout the AlphaFold Substrate Alignments artifact. Terms are presented in a substrate‑agnostic, model‑independent manner and apply to any biological inference engine with comparable structural and latent‑space characteristics.

1. Substrate Terms#

Substrate#

A structured, model‑agnostic framework that defines stable axes, primitives, and invariants for interpreting high‑dimensional inference systems.

Substrate Axis#

One of the three primary axes (S‑axis, I‑axis, R‑axis) used to map structural, latent‑space, and resonance‑time behavior.

Substrate Primitive#

A minimal structural unit used to define substrate behavior, including Dimensional Cores, Resonance Primitives, and Inference Projections.

Substrate Invariant#

A property that remains stable across inference cycles, model versions, and dimensional projections.

2. Axis Definitions#

S‑axis (Structural Axis)#

Represents geometric and topological features of predicted protein structures, including backbone orientation, motif‑level coherence, and residue‑interaction geometry.

I‑axis (Inference Axis)#

Represents latent‑space structures such as attention‑map coherence, pairwise embeddings, and track‑level alignment signals.

R‑axis (Resonance‑Time Axis)#

Represents inference‑cycle behavior, including convergence, oscillation, divergence, and regime transitions.

3. Dimensional Terms#

Dimensional Core#

A stable geometric substrate used for projection and interpretation.
Includes the 3D structural core, 6D interaction core, and 9D coherence core.

3D Core#

Captures physical geometry and motif‑level structure.

6D Core#

Captures interaction‑level structure and residue‑pair coherence.

9D Core#

Captures folding‑pathway coherence and resonance‑time alignment.

High‑Dimensional Extension#

Projection of latent‑space structures into dimensions beyond 9D (e.g., 32D–128D) while preserving substrate invariants.

4. Regime Terms#

Folding Regime#

A classification of folding behavior into stable, transitional, or high‑uncertainty states.

R₁ (Stable Regime)#

A regime where structural and latent‑space signals converge consistently.

R₂ (Transition Regime)#

A regime characterized by oscillatory or branching behavior during structural reorientation.

R₃ (High‑Uncertainty Regime)#

A regime where predictions exhibit high variance and weak motif‑level coherence.

5. Inference Terms#

Latent‑Space Structure#

Internal model representations such as attention maps, pairwise embeddings, and track‑level signals.

Inference Cycle#

A single pass or iteration of the model’s internal reasoning process.

Coherence Surface#

A stable region in latent space where folding predictions converge.

Inference Projection#

A mapping from latent‑space structures onto substrate axes.

6. Validation Terms#

vST (Validation‑Space‑Time)#

A validation framework that evaluates structural coherence, latent‑space stability, regime behavior, and dimensional‑core alignment.

Validation Layer (V₁–V₄)#

A structured evaluation step within vST, each corresponding to a specific substrate property.

7. Drift Terms#

Drift#

A deviation from expected substrate behavior, indicating model degradation or instability.

Drift Category (D₁–D₄)#

Classification of drift into structural, latent‑space, regime, or dimensional‑core drift.

Drift Severity#

A measure of the magnitude of drift, ranging from low to high.

8. Example Terms#

Dimensional Projection#

The process of mapping high‑dimensional inference structures into 3D–9D cores.

Regime‑Aware Projection#

A projection that preserves regime identity (R₁, R₂, R₃).

Alignment Walkthrough#

A step‑by‑step demonstration of substrate alignment applied to a specific inference output.