The aim of this paper is to develop predictive models that can determine, whether a particular compound is active (1) or not (0). In this, we have different datasets along with the drugs. A dataset contains a number of features. Every feature will be processed by using feature selection algorithms and particular compound also be found with the help of these features. We have taken the dataset from Tox21 databases, clean the data and then apply the different feature selection algorithms in order to predict the desired result by applying different models whether the particular drug will be beneficial or injurious to the health of our patient. All the models will be compared with each other and the suitable one will be selected whose accuracy will be more.