Dicoias Equation — formulation intelligence by Panacea Bio Chem
Computational formulation intelligence · Panacea Bio Chem

One equation that reads a molecule and tells you what a formulation still needs.

Every substance — peptide, excipient, unknown powder — becomes a vector. Every recipe becomes a need. The Dicoias Ψ equation measures the gap between them: structure in, property out, certainty honestly capped. This is the private mathematical language Panacea Bio Chem built for its formulation work — mathematics as the universal formulation language.

Ψ(S | P, H) = Σj πj [ α·Δj² + β·(C⁻j)² + γ·Ej² ]
Score = 100 × e−Ψ/τ  ·  β > α — the do-no-harm weighting is written into the objective itself.
Φ → Λ 26 descriptors Ψ
Scientific illustration — not experimental imagery. An amino-acid chain assembles; structure becomes the Φ descriptor vector; Ψ reads it.

What Ψ computes

Ψ is a structure→property intelligence. It does not simulate a beaker — it maps molecular structure into a shared descriptor space, places the result on capability axes, and measures how far a whole formulation sits from what its peptide actually needs. Four pillars carry the weight.

Structure → vector

The W6 "Mendeleev" engine takes a canonical SMILES and derives a 26-dimensional descriptor vector Φ — electrostatics, hydrophobic patterning, chemical liability, solid-state behaviour — then maps Φ onto capability axes Λ with a ridge-NNLS model calibrated on 27 curated anchor substances. 68 of 101 periodic-table entries are derived this way.

Whole-recipe scoring

A formulation is a set, not a molecule. Ψ combines every ingredient's effective capability into Λ_S and scores the shortfall Δ, the harmful overshoot C⁻ and the oversupply E against the peptide's need vector N — with β > α, so doing harm costs more than falling short. The score maps through 100 × e−Ψ/τ with τ = 0.615.

Quantum descriptors, for real

A display-only sidecar computes frontier-orbital descriptors — HOMO, LUMO, gap, ionisation potential, electron affinity — using a real GFN2-xTB backend (xtb 6.7.1, run as a subprocess with hard timeouts and typed failure states). Empirical RDKit descriptors and SMARTS-based proxies are labelled as exactly what they are.

Bayesian memory

W7 treats every literature prior as a Beta distribution and updates it conjugately with each real Panacea run — posteriors live in an overlay, frozen literature priors are never mutated, and the 90% credible interval is the headline number. Certainty is earned run by run, never asserted.

The Dicoias Φ descriptor space: 27 curated anchors connected in known space, a distant query point certainty-capped — Panacea Bio Chem
Scientific illustration — not experimental imagery. The 26-dimensional Φ descriptor space projected to two dimensions: 27 curated anchors define known space; a structurally distant substance falls outside it, and its certainty is capped by distance.

The eight waves of the stack

Ψ is not one formula but a layered architecture, built in eight waves — each one display-only and advisory, each one gated. The full depth is on The Equation page.

W1

Surface the numbers

Every vector, every score, every driver made visible — a periodic table of substances where nothing is computed-then-hidden.

W2

Set scoring

Whole recipes scored as sets with an Occam tie-break — the mixture, not the molecule, is the unit of truth.

W3

τ calibration

The score scale τ = 0.615, with provenance, turning raw Ψ into an interpretable 0–100 readout.

W4

Ψ-driven search

Greedy forward selection refined by deterministic simulated annealing, under an osmolality budget — advisory, off by default.

W5

What to add

Backward Λ-space inversion: which excipient, and roughly how much, closes each gap — a proposal, never an instruction.

W6

Mendeleev engine

Structure→Λ derivation for substances Ψ has never seen — plus the honest electronic sidecar and the real GFN2-xTB backend.

W7

Bayesian posteriors

Beta-Bernoulli conjugate updates from real runs; the credible interval tightens as evidence accumulates.

W8

Health direction

Per-axis structural directions with certainty hard-capped by known-space distance — guidance, never a verdict about people.

W1b

Ask Dicoias

A question engine: Ψ becomes a family of question operators — stability, solubility, redox, comfort, fitness, similarity, health direction, unknown risk — each returning score, certainty, drivers, missing data and next action.

The Dicoias Ψ stack — eight waves W1 to W8 linked by the Ψ objective — Panacea Bio Chem
Scientific illustration — not experimental imagery. The eight-wave Ψ architecture feeding one objective.

The honesty doctrine

A number you cannot defend is worse than no number. Ψ is built so that false certainty is structurally impossible — three mechanisms do the work, and the full doctrine lives on the Honesty page.

Honest declines

Unresolvable structure, ambiguous identity, missing quantum backend, uncurated biological reference set — each returns a typed decline with the reason attached. The system says "I cannot know this" in the same voice it says everything else. Never a fabricated number.

Certainty caps

The known-space distance gate measures how far a substance sits from the 27 curated anchors in descriptor space. Farther away means a harder ceiling on certainty — down to a 0.15 cap for structurally alien substances. Novelty automatically deflates the readout.

The display-only firewall

The entire Ψ spine is isolated from the formulation engine: it writes nothing, applies nothing, mutates no seed and touches no actuator. Ψ informs the operator; the operator — and only the operator — decides.

Evidence, labelled

Every number carries its provenance: literature prior, derived-from-structure, or bench-calibrated from real Panacea runs. The label travels with the value wherever it is shown.

The Dicoias Ψ equation stack — Panacea Bio Chem private formulation intelligence

A private computational alchemy

Panacea Bio Chem does not sell software. Ψ exists for one reason: so that when Panacea formulates — a peptide, an amino-acid chain, a lyophilized cake — the question "what does this molecule still need?" has a mathematical answer before the first bench run, and an honest boundary drawn around what mathematics cannot yet say. It is a private computational alchemy: structure transmuted into property, property into guidance, guidance into better questions for the bench.

It runs inside S3Pulse, Panacea's formulation environment, beside the instruments it advises — and it stays in its lane: computational intelligence that informs, never an oracle that decides.

Private system · displayed publicly for scientific context · advisory by construction
Bogdan Dicoias — Biochemist, AAC Designer, Director of Panacea Bio Chem
"A formulation is a sentence written in molecules. For years we read those sentences by intuition and by bench time. Ψ is the grammar we wrote down: every substance a vector, every need a direction, every claim carrying the evidence that earned it. The part I am proudest of is not what it computes — it is that when the data is not there, the system says so, plainly, instead of inventing a number."

Bogdan Dicoias — Biochemist · AAC Designer · Panacea Bio Chem Ltd

Frequently asked questions

What is the Dicoias Ψ equation?

Panacea Bio Chem's formulation-compatibility objective: every substance is reduced to a capability vector, a recipe's needs become a need vector, and Ψ measures the squared distance between them. The score is 100 × e−Ψ/τ.

How does Ψ handle a substance it has never seen?

Through the W6 Mendeleev engine: canonical SMILES → 26-dimensional descriptor vector → ridge-NNLS mapping calibrated on 27 curated anchors. Unresolvable structures get a typed decline, never a guessed vector.

Does Ψ use quantum chemistry?

Yes — a display-only sidecar computes HOMO/LUMO and derived quantities with a real GFN2-xTB backend (xtb 6.7.1). Without a backend or a structure, those fields are honest declines, not numbers.

Does Ψ decide the recipes?

No. The whole spine is display-only and advisory, firewalled from the engine. It informs the operator; it never writes a recipe. More on the Honesty page and the Questions page.

Trending in the field

The Panacea Technology Universe

24 technologies, each the leader of its class

Proprietary Panacea Bio Chem Ltd technologies, invented by Bogdan Dicoias — what each one does, and why it leads its class.

Lyoprester® — Panacea Bio Chem technology by Bogdan DicoiasLyoprester®The only dual-chamber cartridge that is autoreconstitution-enabled, vacuum-sealed and argon-fillback.lyoprester.com ↗P-EARLs — Panacea Bio Chem technology by Bogdan DicoiasP-EARLs™Panacea-Engineered Aseptic Reconstitution Liquid(s) — each tuned to the peptide it wakes.p-earls.com ↗Peptourbillon — Panacea Bio Chem technology by Bogdan DicoiasPeptourbillon™The layered peptide formulation architecture — single- or multi-layer, never a blend.peptourbillon.com ↗RF Tunnel — Panacea Bio Chem technology by Bogdan DicoiasRF Tunnel™The RF-formed central channel through the cake.rftunnel.com ↗TgShift — Panacea Bio Chem technology by Bogdan DicoiasTgShift™Raises the cake’s glass-transition temperature with RF — instead of chilling below it.tgshift.com ↗Cryolapse — Panacea Bio Chem technology by Bogdan DicoiasCryolapse™Cryogenic pressure collapse — and the machine that pushes plungers and crimps.cryolapse.com ↗LyoLevit — Panacea Bio Chem technology by Bogdan DicoiasLyoLevit™The cake levitates and spins in high orbit — driven by ultrasound and RF.lyolevit.com ↗Lyochrysalis — Panacea Bio Chem technology by Bogdan DicoiasLyochrysalis™The integrated chamber housing the whole drying stack.lyochrysalis.com ↗S3Pulse — Panacea Bio Chem technology by Bogdan DicoiasS3Pulse™The control brain for every piece of Panacea hardware.s3pulse.com ↗Liquiprester — Panacea Bio Chem technology by Bogdan DicoiasLiquiprester™The single-liquid cartridge engineered so multiple peptide APIs coexist in one shared vehicle.liquiprester.com ↗Syntheseract — Panacea Bio Chem technology by Bogdan DicoiasSyntheseract™Continuous-flow peptide synthesis in a special, very fast and economical way.syntheseract.com ↗CFSPPS — Panacea Bio Chem technology by Bogdan DicoiasCFSPPS™Continuous-flow solid-phase peptide synthesis, written as its own category.cfspps.com ↗OxyDeplete — Panacea Bio Chem technology by Bogdan DicoiasOxyDeplete™Degassing plus no-headspace doctrine — the oxygen-starved seal.oxydeplete.com ↗ArgonLock — Panacea Bio Chem technology by Bogdan DicoiasArgonLock™The final inert-atmosphere lock under argon.argonlock.com ↗RedoxVault — Panacea Bio Chem technology by Bogdan DicoiasRedoxVault™Separation, not merely suppression — redox isolation in lipid micro-reservoirs.redoxvault.com ↗PleniDose — Panacea Bio Chem technology by Bogdan DicoiasPleniDose™The shared filling gantry — one machine filling both the dual-chamber Lyoprester and the liquid Liquiprester.plenidose.com ↗IncreSure — Panacea Bio Chem technology by Bogdan DicoiasIncreSure™The dose-metrology layer — verified API per pen increment.incresure.com ↗ElimiVoid — Panacea Bio Chem technology by Bogdan DicoiasElimiVoid™Front-void elimination without touching the metered dose.elimivoid.com ↗Cryoviscous — Panacea Bio Chem technology by Bogdan DicoiasCryoviscous™The characterised cold, high-viscosity, low-mobility conditioning state.cryoviscous.com ↗
Vana Machine — Panacea Bio Chem technology by Bogdan DicoiasVana Machine™Vacuum Assisted Needle Accessory — vacuum conditioning and plunger-locking for the cartridge.
EZnject — Panacea Bio Chem technology by Bogdan DicoiasEZnject™The disposable auto-injector pen built around the Lyoprester.panaceaeznject.com ↗Dicoias Ψ — Panacea Bio Chem technology by Bogdan DicoiasDicoias ΨThe computed-chemistry advisory — every substance reduced to a vector across physical, electronic and formulation space.dcppsi.com ↗SealoPrester — Panacea Bio Chem technology by Bogdan DicoiasSealoPrester™Aseptic Cartridge Closure System — Seal o’ Precision + Sterility.sealoprester.com ↗Peptidic Liquid — Panacea Bio Chem technology by Bogdan DicoiasPeptidic LiquidThe peptide formulation in solution — the active plus its buffers, cryoprotectants, lyoprotectants and scaffolders.peptidicliquid.com ↗

Weekly review — 7–13 Sep 2026

Publications indexed in PubMed in the last 30 days for ("applicability domain" OR "excipient compatibility" OR "drug-excipient") AND (prediction OR "machine learning" OR model) — refreshed weekly.