Protein modeling has been a very challenging problem in drug discovery and computational biology. The latest advances and progress in computational power have helped to solve this problem to a considerable extent; however, predicting accurate three-dimensional structure of proteins has always been and remains a complicated assignment. Of the two common methods of protein structure prediction, template-based modeling has become more popular than ab initio modeling. In this review, we summarize the developments in methodology and of understanding for comparative protein modeling during the last three years, including for homologue search, fold recognition, secondary structure prediction, model building, loop building, side-chain prediction and model quality assessment.