How would you evaluate a selectivity panel containing many weak off-target activities?

Why

A panel produces a long list, and the list by itself is not a risk assessment. Three things turn a number into a judgement. First, the margin between the off-target potency and the unbound concentration achieved in plasma and in the relevant tissue.

A 50-fold margin on a protein the body tolerates is different from a 50-fold margin on a cardiac ion channel. Second, the consequence of hitting that protein, which is where known liability lists matter, such as hERG for cardiac repolarization or certain kinases for known toxicities. Third, whether the activity is real, since panels run at a single concentration produce false positives that a full concentration-response curve resolves.

Arrowsmith and colleagues note how often an inhibitor's reputation for selectivity outlives the data, with LY294002 and staurosporine as examples, so the panel is worth more than the label. Breadth also matters. A panel of 50 kinases says little about a compound's behavior across the whole kinome or across unrelated families.

Some promiscuity is deliberate, and a few drug classes work through multiple targets, which is why the question is risk against benefit rather than purity for its own sake.