For using ML in virtual screening, a training set of known active and inactive compounds are required. This is training data is then used to train a model using supervised learning algorithms. The model is then used on new data sets to screen compounds with the desired activity against a target.