Designing efficient drugs for curing diseases is of essential importance for the 21st century's life science. Computer-aided drug design and discovery has obtained a significant recognition recently. However, the geometric complexity of protein-drug interactions remains a major challenge to conventional methods. We integrate algebraic topology and deep learning algorithms for the predictions of protein-drug binding affinity, drug toxicity, drug solubility, and drug partition coefficient. We demonstrate that element specific topology offers the best result in high throughput drug screening, protein flexibility analysis and protein stability change upon mutation. I will discuss how deep learning and mathematics, including algebraic topology and graph theory, has led my team to be a top performer in recent two D3R Grand Challenges ( https://drugdesigndata.org/about/grand-challenge ), a worldwide competition series in computer-aided drug design and discovery.