By transfer learning between 10,000s of assays, millions of data points, and 10 millions of experimental measurements, massively multitask modeling makes predictions comparable to 4-concentration experimental biochemical and cellular EC50s.
Predictions for millions of compounds on up to 10,000 assays provides unique information.

• What are the strengths and weaknesses of various MMRM algorithms?
• What are successful applications in drug discovery with examples?
• Are there other approaches, experimental or computational, besides MMRMs?