A potency gain needs a chemical history

A potency number becomes more useful when a reader can retrace the chemistry that produced it. For this launch analysis, I would put three records beside the strongest compound: the starting structure, the selection decision and the measurement. A leaderboard of final activities would leave too much of the discovery story out.

Fourteen compounds, several different achievements

Nippa and colleagues combined reaction prediction with property assessment and structure-based scoring to optimize inhibitors of monoacylglycerol lipase, or MAGL. Their 2025 study produced fourteen new inhibitors. In the detailed Results, all improved on the starting hit, while six had human-enzyme IC50 values between 0.1 and 10 nanomolar. This is a biochemical series result, not fourteen demonstrated medicines.¹

The structural evidence is particularly informative: selected alkyl additions changed how a pyridine ring sat in the binding site, while preserving other interactions. The authors also distinguish predicted properties used during selection from subsequent measurements on selected compounds.¹

Preserve the rejected alternative

Here is the decision record I would ask a future optimization campaign to publish. For each proposed analogue, show the molecular change in a consistent representation. Record why it was selected, what the team expected to improve and which existing property it wanted to preserve. Then attach the actual assay result and the corresponding synthesis outcome. A missing measurement should remain visible as missing.

Consider a hypothetical pair of candidates. One is expected to be more potent but is difficult to prepare. Another is less exciting in the potency ranking but would provide a clean test of the proposed binding explanation. I would want to know which question determined the choice. Making the second molecule can be editorially interesting even when it does not deliver the lowest final number: it can challenge the reason the first molecule was attractive.

This is an argument for documenting decisions, not a claim that these authors performed that hypothetical comparison. I would also resist assigning the entire outcome to the model. A reviewable account should preserve where a person selected a structure, interpreted a result or changed a priority.

Keep the axes separate

For my medicinal chemistry beat, the next reporting question would concern the compounds that are attractive for different reasons. I would ask which analogue earned its position through measured potency, which through a predicted property and which through practical access. Combining those labels into one score may help select work, but it makes a poor substitute for explaining the work afterwards.

I would use the same discipline when describing a trade-off. The useful sentence identifies the changed group, the particular assay and the remaining uncertainty. It should not promote an attractive biochemical result into a claim about exposure, safety or treatment benefit.

The publication provides a concrete series to examine.¹ Our proposed reporting standard is deliberately more demanding than naming the best compound: keep enough chemical history that another reader can challenge the next design decision. That is where a model-assisted optimization story becomes an inspectable medicinal chemistry argument.