chemistry-interaction-inventory-of-the-modelled-complex

# Interaction inventory of the modelled complex ## What goes wrong When the deliverable is a modelled molecular complex — a docked pose, a co-folded assembly, a predicted interface — the report stops at aggregate distances: RMSD, DockQ, lDDT, "fraction under the accepted cut". Those say how far the partner is from where it belongs. None of them says whether the model found the right pocket, reproduced the recognition chemistry, or kept the specific contacts that hold the complex together. A reader who wants to know what the model got right about the chemistry finds nothing to read. Three symptoms: - **The only structural figure is a whole-complex superposition.** The interface is 10–15 Å across inside a 60–80 Å frame, so it occupies a few percent of the canvas. No residue is labelled and no contact is drawn. The figure shows that two clouds overlap. - **The contact-level number was computed and dropped.** The fraction of native contacts is a term inside DockQ, so any run reporting DockQ already has it, and burial counts fall out of any interface routine. This is `publish-what-the-run-already-computed`'s sweep, applied to the interface. - **The prose says the partner is "placed correctly but imprecisely".** That is an eyeball claim about a picture with no measurement under it. A figure list generated from your own hypotheses will not contain this slot, because no hypothesis asks for it. Reserve it at study design as a deliverable of the subject matter, not o...