Different QSAR models of 40 natural compounds as neuraminidase inhibitors (NIs) are developed to comprehend chemical–biological interactions and predict activities against neuraminidase (NA) from Clostridium perfringens. Based on the constitutional, topological and conformational descriptors, R² and Q² values of the obtained SRA model are 0.931 and 0.856. The R² and Q² values of the constructed HQSAR and almond models are 0.903 and 0.767, 0.904 and 0.511, respectively. Based on the pharmacophore alignment, R² and Q² values of the optimal CoMSIA model are 0.936 and 0.654. Moreover, R_test ² and Q_ext ² of values of SRA, HQSAR, almond and CoMSIA models are 0.611 and 0.565, 0.753 and 0.750, 0.612 and 0.582, 0.582 and 0.571, respectively. So, QSAR models have good predictive capability. They can be further used to evaluate and screen new compounds. Moreover, hydrogen bonds and electrostatic factors have high contributions to activities. To understand molecular interactions between natural compounds and NA from Clostridium perfringens, molecular docking is investigated. The docking results elucidate that Arg266, Asp291, Asp328, Tyr485, Glu493, Arg555, Arg615 and Tyr655 are especially the key residues in the active site of 2bf6. Hydrogen bonds and electrostatics are key factors, which impact the interactions between NIs and NA. So, the influential factors of interactions between NIs and NA in the docking results are in agreement with the QSAR results.