Migrastatin, a macrolide natural product, and its structurally related analogs are potent inhibitors of cancer cell metastasis, invasion and migration. In the present work, a specialized fragment-based method was employed to develop QSAR models for a series of migrastatin and isomigrastatin analogs. Significant correlation coefficients were obtained (best model, q2 = 0.76 and r2 = 0.91) indicating that the QSAR models possess high internal consistency. The best model was then used to predict the potency of an external test set, and the predicted values were in good agreement with the experimental results (R2 pred = 0.85). The final model and the corresponding contribution maps, combined with molecular modeling studies, provided important insights into the key structural features for the anticancer activity of this family of synthetic compounds based on natural products.