Ashebo Method Biophysics is a computational drug discovery platform that predicts protein druggability from atomic structure alone. No training data. No pattern matching. Pure first-principles physics.
The platform does not just score druggability — it discriminates. It tells you yes, but carefully for structurally tractable hard targets, no, clearly for genuinely undruggable proteins, and this domain, not that one for targets where only specific regions are viable.
The pharmaceutical industry spends over $2 billion per approved drug, and the single largest source of failure is target selection. Programs spend years and hundreds of millions of dollars pursuing protein targets that turn out to be structurally undruggable — or druggable only in specific conformational states that were never assessed.
Existing computational tools (FPocket, SiteMap, DoGSiteScorer) use geometric heuristics or ML-trained models that cannot distinguish between a genuinely undruggable protein and a hard-but-tractable one. They produce false positives on flat surfaces and false negatives on cryptic pockets. They cannot tell you which domain of a multi-domain protein to target, or whether a rescued mutant pocket is real.
The result: billions of dollars wasted on targets that a physics-based analysis could have triaged in minutes.
The Ashebo Engine V5 processes a protein structure through six sequential decision layers. Each layer consumes the outputs of the layers above it and adds a new dimension of analysis. The result is a complete druggability assessment that separates structural evidence from biological context from chemistry strategy from therapeutic feasibility.
Local Potency Density (LPD) field computed from 10 proprietary atomic interaction constants. Peak LPD, binding site discovery, sigmoid scoring. This is the frozen physics core — no ML, no training data.
Deep pocket characterization: enclosure, depth, volume, stability, chemistry balance, coherence. Each pocket scored independently with geometry classification (cavity, groove, surface).
Target class recognition, modality risk assessment, known ligand context. Determines whether the structural signal translates to biological tractability.
Strategy classification: traditional small molecule, fragment-first, covalent warhead, PPI disruption, allosteric, biologics-only, or non-small-molecule constrained. Pocket-level chemistry summaries.
Disease fit, modality realism, clinical precedent overlay. Connects structural findings to therapeutic development paths.
Formula-level drug candidate hypotheses with molecular weight, element composition, scaffold direction, and strategy coherence scoring. Primary and alternative paths ranked by strategy-adjusted score.
The platform's value is not in producing high scores — it is in producing the right answer. The validation portfolio demonstrates three distinct discrimination behaviors across hard-target proteins that conventional tools cannot distinguish.
5 conformational states analyzed (AlphaFold, BCL9/TCF4, hTcf-4, Compound 6, Axin). Scores 65–88/100. Recurring 299–312 Da compact inhibitor hypothesis across all states. Strong coherence. The hardest positive in oncology — and the engine says yes.
Score 12/100. Top pocket 0/100. Non-small-molecule constrained. The engine correctly identifies MYC as structurally undruggable for conventional small molecules — consistent with 30 years of failed MYC drug programs.
Score 3/100. Intrinsically disordered protein with no stable binding pocket. Non-small-molecule constrained. Biologics or degrader approaches required.
RRM1 domain is the strongest small-molecule-compatible region. Other domains are weaker or alternative-modality constrained. The engine prioritizes where to invest, not just whether to invest.
Apo mutant (6SHZ): 93/100. Ligand-bound with rezatapopt (9BR4): 83/100. Same 338 Da primary lead in both states. The engine recovers the clinically validated Y220C pocket — a rescued-pocket benchmark.
| Dimension | Conventional Tools | Ashebo Method |
|---|---|---|
| Foundation | Geometric heuristics or ML | First-principles atomic physics |
| Training data | Thousands of known complexes | Based on new physics concept that also solved the many-body problem |
| Output | Pocket score | 6-layer decision stack: score → pocket → biology → chemistry → therapy → molecule |
| Hard targets | False positives / false negatives | Discriminates: yes / no / where |
| Novel proteins | Degrades without training data | Works on any atomic structure |
| Multi-state | Single structure | Cross-state consistency validated (5 β-catenin conformations) |
Independent researcher and theoretical physicist based in Alberta, Canada. Creator of The Ashebo Method — a unified theoretical framework that derives Newton's gravitational constant G from first principles with zero free parameters and 0.064% accuracy.
The same atomic interaction constants that power the gravity framework — the Potency Map — turned out to predict protein druggability with remarkable accuracy. The biophysics platform emerged from the insight that atomic coordinates and molecular geometry are mathematically inseparable — a principle Yohannes calls coordinate-geometry entanglement.
Today, the Ashebo Method spans two platforms: ashebophysics.com (7 publications in fundamental physics) and this platform — the computational drug discovery engine. Multiple provisional patent applications have been filed with the United States Patent and Trademark Office.
We run your protein through the full 6-layer decision stack and deliver a comprehensive report with structural assessment, chemistry strategy, and molecular hypotheses — typically within 48 hours.