Consider a virtual screen of fifty thousand candidate molecules against a target protein, an early and increasingly standard step in modern drug discovery. A computational model predicts each candidate's binding affinity, how strongly and specifically it is likely to bind the target, long before any physical compound gets synthesized in a lab.
Why the threshold is the whole economics of screening
Only candidates scoring above a chosen threshold advance to wet-lab testing, where the true hit rate finally gets confirmed through real chemistry, a process that remains slow and expensive per candidate regardless of how sophisticated the upstream screening was. The threshold determines how large, and how enriched with genuine hits, that expensive shortlist actually is.
Set it too low, and cost balloons without proportional benefit
A low threshold advances a large number of candidates, most of which will not turn out to be genuine hits, driving up wet-lab costs without a proportional increase in actual discoveries. At the scale of tens of thousands of candidates, that inefficiency compounds fast.
Set it too high, and real hits get filtered out
A high threshold produces a smaller, cheaper, more enriched shortlist, but risks discarding genuine hits whose predicted affinity score, while real, simply did not clear an overly conservative bar, a false negative that a purely computational screen can never fully protect against.
Try it yourself
The AI Drug Discovery Lab simulates a pool of candidates with a realistic gap between predicted affinity and true hit status, letting you move the threshold and watch shortlist size, true hits captured, and estimated wet-lab cost respond.
🧪 Try it yourself: the AI Drug Discovery Lab simulation lets you experiment with everything described above directly in your browser.