Fragment-based drug design has developed significantly over the past ten years and is now recognized as a successful method of lead compound generation and optimization[1]. Computational approaches to fragment-based drug discovery have the potential to dramatically mitigate the costs of experimental based approach. However, the majority of computational methods suffer from limited representation of protein flexibility and solvation effects. Recently, a fragment-based approach based on explicit solvent all-atom molecular dynamics simulations (SILCS: Site Identification by Ligand Competitive Saturation) was developed in our lab to overcome these drawbacks[2]. As a test case, we applied this method to the BTB domain of BCL6 protein. Good agreements between calculated three-dimensional probability maps of fragment binding and the X-ray structure of the BCL6 inhibitor 79-6[3] were found. In addition, a more specific SILCS simulation was performed and several potential functional group binding sites around the location of 79-6 were identified. Based on the location of these functional groups, modifications of 79-6 were proposed. Free energy perturbation (FEP) calculations, including FEP using the orthogonal space random walk (OSRW) approach[4], were applied to obtain quantitative computational estimates of the relative free energy of binding associated with these modifications. Two FEP methods show consistent results that the modified compound has higher binding affinity to BCL6 than parent compound 79-6. This result indicates that the SILCS method has the ability to qualitatively inform the optimization of small-molecule inhibitors.[1] Schulz, M.; Hubbard, R. Curr. Opin. Pharmacol. 2009, 5, 615.[2] Guvench, O.; MacKerell, A. D., Jr. PLoS Comput. Biol. 2009, 5, e1000435.[3] Cerchietti, L. C.; Ghetu, A. F.; Zhu, X. and et al. Cancer Cell 2010, 17, 1.[4] Zheng, L.; Chen, M.; Yang, W. Proc. Natl. Acad. Sci. U.S.A. 2008, 105, 20227.