Fast and accurate virtual screening of over 10 million compounds.

A completely data-driven, domain knowledge-free, deep learning method for absorption, distribution, metabolism, and excretion (ADME) property prediction.