DELFICTUS IO — COMPUTATIONAL STRUCTURAL BIOLOGY INFRASTRUCTURE
—— Companion Artifact 02 · Manifold Observatory

Advanced Capabilities

A coupled interferometric translational engine, rendered as a living manifold observatory.

PRISM-DSTW is not a pipeline. It is a four-act observability engine: twin interferometric manifolds under perturbation, bifurcation topology with attractor basins and saddles, BALD active uncertainty collapse, and chronology-bound decision artifacts. The interactive artifact below renders all four acts and their cross-cutting layers — thermal hysteresis, UV excitation, LIF event-collapse, causal propagation — as they actually operate in the engine.

Manifold Observatory
Interactive · WebGL · 60 fps
Twin manifolds Bifurcation topology BALD vortex Chronology lock Open standalone
Advanced capability
Bridge contracts
Wet-lab modalities remain evidence, not physics.
HDX, DMS, BRET, SPR, cryo-EM occupancy, and internalization data enter through typed bridge contracts. Each record keeps provenance, source tier, assay state, license posture, and uncertainty semantics intact.
Projection layer
TSOs stay topological and residue-aware.
PRISM-DSTW maps PRISM-4D telemetry into residue-indexed and peptide-integrated observables, preserving dynamic contact graphs and transfer-entropy structure rather than relying on absolute voxel coordinates.
Calibration
Bayesian bridges quantify what biology supports.
Layer 4 consumes physically plausible TSOs under shrinkage, censored likelihoods, and errors-in-variables uncertainty. Biological relevance is inferred; the registry only controls admissible evidence flow.
Four-act architecture

The observatory above is the architecture, not a metaphor for it. Each act in the visualisation corresponds to a real subsystem in PRISM-DSTW: twin manifolds under coupled perturbation, bifurcation topology with attractor basins, BALD adaptive uncertainty collapse, and a chronology-bound decision lattice. The cards below name what is happening in each act and what it means for translation.

1 · State tensor
PRISM-4D exports aligned physics, not biological labels
The physics substrate remains blind to assay outcomes. It emits state manifolds, transfer-entropy channels, thermal hysteresis, hydration variance, and dynamic contact graphs through the handshake contract.
2 · Observation bridge
Assays become likelihood blocks with provenance
Each assay is treated as a window onto the same latent thermodynamic geometry. Processed tables, raw-derived spectra, and validation references keep separate evidence kinds so the model can express disagreement instead of hiding it.
3 · Active learning
BALD selects the next physics questions
Posterior uncertainty is routed back into PRISM-4D as a targeted simulation request. The loop spends compute on variants and states that are expected to collapse the most uncertainty.
4 · Firewall
Decision artifacts are chronology-bound
Registries, handshake contracts, priors, normalized evidence, and exported decision matrices are hashable as a single evidence state. The downstream claim is tied to the manifold state that produced it.
bioRxiv MS BIORXIV/2026/725676 Zenodo DOI 10.5281/zenodo.20247900 USPTO Prov. 64/067,538 (2026-05-16) UEI LXT3B9GMY4N8 CAGE 13H70 Manifest SHA-256 73d178…5930