Conformational changes occurring in proteins are closely related to their biological functions.　Ligands, when located near the active sites of enzymes, induce conformational changes that stabilize the ligand-enzyme complex and then catalyze appropriate chemical reactions. In this study, we use linear response theory to predict the direction of conformational changes that occur upon ligand binding. In contrast with the previous studies, forces upon ligand binding are obtained from all-atom force field with parameters of CHARMM22. In addition, we use covariance matrix obtained from ANM on unbound structure (ligand-free), thus developing a fast (compared to the use of molecular dynamics simulations) and accurate method to predict conformational changes upon ligand binding. Our dataset consists of proteins that undergo coupled and independent domain motions upon ligand binding. We find that LRT cannot be used to predict conformational changes of the latter. In the case of proteins that undergo coupled domain motion, we are able to predict the conformational changes upon ligand binding without any information from the bound structure. In addition, following results are observed for proteins with coupled domain motion. The conformational changes predicted with linear response theory using either MD derived forces or simple attractive forces are not significantly different from each other. Also, we see that the average correlation coefficient in the case of LRT predicted conformational changes is higher than that from ANM first mode. If MD forces are used in LRT, energy minimization has no significant effect on the correlation coefficient. When simple perturbation forces are used in LRT, the average correlation coefficient is higher when the initial ligand/free-protein complex is energy-minimized. Hence, we speculate that LRT results are sensitive to the position of ligand, especially when simple perturbation forces are used.