Pharmacology, computed from first principles.

Identity & exposure

What the molecule actually is in vivo: protonation state, tautomer, stereochemistry, and the dose-to-exposure relationship that sets every downstream concentration.

Where it goes

Passive permeability, active uptake and efflux (OATP1B1/1B3, BCRP, BSEP, P-gp), and the partition across compartments that determines what concentration each tissue actually sees.

How long it stays

Phase-I and Phase-II stability, intrinsic clearance, reactive-metabolite alerts, and a mechanism-aware CYP isoform panel — feeding the timecourse and the hepatic safety profile.

What it touches

The target's interaction surface as a physical ensemble, not a static crystal pose — and the mode of binding through which the molecule actually engages it.

What happens next

Downstream mechanism — activation, inhibition, displacement, degradation, allosteric coupling — and the functional readout that turns binding into a measured response.

What you can defend

Calibrated uncertainty per endpoint, a structured explanation report, and an audit-grade record that traces every prediction back to the physical events that produced it.

Pocket-bound binding

Classical small-molecule–protein engagement — the ATP pocket, the orthosteric site, the allosteric cleft. The deepest-validated mode in the platform.

Protein–protein interface disruption

Surface-PPI inhibitors that block large, flat interaction interfaces — the historically “undruggable” class.

Targeted protein degradation

Molecular glues and PROTACs that induce a ternary complex with an E3 ligase, leading to ubiquitination and proteasomal degradation.

Nucleic-acid binding

DNA, RNA, and G-quadruplex engagement — minor-groove binders, intercalators, splice modulators, RNA-targeting therapeutics.

Biomolecular condensates

Partitioning into membraneless organelles and condensate phases — an emerging mechanism for nuclear transcription factors and stress granules.

Amyloid & aggregate modulation

Engagement with fibril, oligomer, and aggregate states — the physics of neurodegenerative-disease therapeutics.

Membrane-active mechanisms

Molecules whose primary target is the lipid bilayer itself — antimicrobial peptides, ionophores, certain antifungals and antiparasitics.

Glycan & carbohydrate engagement

Carbohydrate-recognition interactions — lectin binding, viral-attachment inhibitors, glycan-mediated cell-surface targeting.

Nuclear receptors

PXR, AR, ER, GR, PPAR, RAR — physics-resolved ligand-binding-domain activation with full agonist/antagonist/inverse-agonist distinction.

Kinases (~70)

ATP-pocket and allosteric inhibitors across tyrosine, serine/threonine, and dual-specificity families — with DFG state and gatekeeper-residue physics.

Ion channels

Voltage- and ligand-gated channels with state-dependent binding — including a mechanism-aware hERG liability adapter at production readiness.

GPCRs (~160 across 5 subfamilies)

Aminergic, cannabinoid, purinergic, peptide-binding, and lipid-binding receptors — agonist/antagonist/biased-ligand resolution.

Catalytic enzymes

Cysteine proteases, coagulation factors, zinc metalloenzymes (including HDAC1–11), and heme dioxygenases — covalent and non-covalent inhibition.

Epigenetic readers

BRD-family bromodomains, EP300, CREBBP — acetyl-lysine recognition with conserved physics across the family.

Transporters

OATP1B1, OATP1B3, BCRP, BSEP, P-gp — substrate and inhibitor liability inferred from the same physics that drives distribution and clearance.

Cytochromes-as-targets

Direct CYP inhibition and time-dependent inhibition resolved through the same heme/isoform physics used for metabolism prediction.

Aspartyl proteases

Catalytic-dyad protease physics — HIV protease, BACE1, renin, and viral main-protease targets.

Structural proteins

Tubulin and actin binders — the physics of taxanes, vinca alkaloids, and actin-disrupting cytotoxic warheads.

Lysine demethylases

LSD1 (flavin-dependent) and the JmjC family (iron/α-ketoglutarate-dependent) — two distinct catalytic mechanisms, both modeled explicitly.

Selectivity panel

Score one molecule across every target class FluxMateria covers — intended target, related family members, off-target liabilities, anti-targets, safety-critical proteins — in one call. Returns a ranked binding profile with per-target uncertainty and integrated safety flags (hERG, CYP inhibition, transporter risk).

Used in lead-optimisation cycles to catch cross-reactivity before it surfaces in a tox study.

Repurposing panel (~260 targets)

Drop in any approved drug, tool compound, or natural product. Get back a ranked off-target engagement profile across ~260 targets spanning kinases, GPCRs, nuclear receptors, ion channels, transporters, and catalytic enzymes — with mechanism attribution for every hit.

Built for indication-expansion programs, mechanism-of-action discovery, and adverse-event hypothesis generation.

Published benchmark accuracy

Use as primary screening criterion.

Published numbers against public leaderboards or curated leave-one-out panels of 14K+ compounds.

• 5 of 8 ADMET endpoints at public-benchmark SOTA: Solubility, Metabolism, PPB strict #1; DILI AUROC 0.9597 on the comparable TDC binary task; Caco-2 MAE 0.277 matching the public TDC reference SOTA
• 3 of 8 ADMET endpoints in the high-accuracy class: BBB classification, hERG mechanism, secondary permeability tier
• Pocket-bound binding on the PXR nuclear receptor (production reference, bit-equality preserved across the cascade rewire)
• hERG ion-channel liability — mechanism-aware, production reference
• CASF-2016 binding-affinity benchmark across diverse target classes, Pearson r = 0.772

Same physics as validated targets

Usable as primary screening criterion with the per-endpoint validation caveat.

The cascade runs end-to-end and dispatches to family-specific physics. Validation depth is rolling out family by family on customer and held-back datasets.

• Pocket-bound binding across the remaining target families: ~70 kinases, ~160 GPCRs across 5 subfamilies, catalytic enzymes (incl. HDAC1–11), epigenetic readers, transporters, CYP-as-target, aspartyl proteases, structural proteins, KDM lysine demethylases
• Selectivity panel across the wired family set — intended target + off-targets + safety flags in one call
• Repurposing panel across ~260 targets, with mechanism attribution per hit
• Cross-target safety integration: CYP, hERG, and transporter risk fused with the ADMET cascade

Newer physics, wider uncertainty

Treat output as directional input; cross-check experimentally.

Code paths run end-to-end with deliberate scope caps and elevated per-prediction uncertainty, so the engine flags where it is leaning on the prototype tier.

• The seven non-pocket modes of binding physics: protein–protein interface disruption, targeted protein degradation (molecular glues / PROTACs), nucleic-acid binding (DNA / RNA / G-quadruplex), biomolecular condensates, amyloid & aggregate modulation, membrane-active mechanisms, glycan / carbohydrate engagement
• Fully calibrated cross-mode uncertainty across all eight modes
• End-to-end explanation reports with full inter-step mechanism trace
• Public-benchmark validation depth for the in-pilot family adapters

Bacterial

Today: PBPs · MurA-F · DNA gyrase · topo IV · RNA pol · DHFR/DHPS · FAS-II reachable as binding targets. Roadmap: ribosome 50S/30S · ATP synthase · cytochrome bc1 · MIC + resistance outputs.

Viral

Today: polymerases (RdRp, DNA pol, RT) · proteases (HIV PR, HCV NS3/4A, SARS-CoV-2 Mpro) · integrase · helicase · NA reachable. Roadmap: capsid · entry proteins · lifecycle-stage selectivity.

Fungal

Today: CYP51/Erg11 (azoles) · squalene epoxidase · Erg6/2 · flucytosine targets · β-tubulin reachable. Roadmap: FKS1 β-glucan synthase · chitin synthase · host-CYP51 selectivity composition.

Parasitic

Today: PfDHFR-TS · PfCytB · PfATP4 · PfActin/PI4K · T. cruzi CYP51 · trypanothione reductase · helminth β-tubulin reachable. Roadmap: parasite Cl channels · lifecycle-stage selectivity.