Nine oncology-relevant proteins selected for their historical difficulty in small-molecule drug discovery. Each target analyzed in at least two structural states to test cross-state consistency and state-dependent behavior. All results produced by the Ashebo engine from first-principles physics — no external database lookups.
The engine produces clear no-go signals for intrinsically disordered proteins — BCL9 at 3/100, YAP1 at 0/100, NRF2 at 5/100. These scores prevent resource waste on targets that cannot be addressed with small molecules. Four of nine targets received unambiguous no-go classifications.
Two targets (TEAD4, PCNA) demonstrated that the same protein scores differently depending on its conformational state. TEAD4 swings from 44 to 59 between inhibitor-bound and YAP-bound states. PCNA improves from 41 to 53 when a ligand reshapes the surface. This guides structure selection for drug discovery campaigns.
The 312 Da C₁₀H₁₇N₂O₄ formula recurs across KEAP1, RB1, and STAT3 — and was previously observed in β-catenin analyses. This convergent molecular weight optimum for balanced drug-like scaffolds targeting medium-sized pockets warrants further investigation as a potential design principle for hard-target medicinal chemistry.
RB1 demonstrates the engine's ability to separate the physics question ("is there a pocket?") from the biology question ("can you use it therapeutically?"). The pocket domain scores 85–94/100 with strong coherence, but Layer C correctly flags it as a loss-of-function tumor suppressor requiring functional restoration — not inhibition. The pocket is real; the therapeutic strategy is complex.