It is a classification problem. In this study, machine learning models are developed for protein secondary structure prediction (PSSP), relative solvent accessibility prediction (RSA), and torsion angle prediction (TAP). PSSP aims to assign a secondary structure class to each amino acid of a protein. It can be predicted as 8-states or 3-states. In this work, the 8-state representation is transformed to 3-states. SA is the area that is accessible to solvent such as water and RSA is the SA normalized by the maximum accessible surface area. Similar to secondary structure, SA and RSA information is derived for each amino acid separately. It can be predicted as a real-valued quantity or it can be categorized and predicted as a discrete label.