1 research article on Compound Screening — drawn from 53 integrated data sources with computational validation.
After evaluating 120+ invention candidates across 5 rounds of a structured generate-critique-disprove-audit cycle, we found that service/API ideas survived at ~27% while compound-level ideas survived at 0%. Every compound promotion was killed as prior art. The anti-portfolio reveals systematic patterns in what kills ideas.