ML isn't really what's being used since it's too unreliable. The reigning models are all differential equations based, using ODEs and PDEs along with the clinical side using PBPK with maximum-likelihood estimation.