Quantitative structure-activity relationship (QSAR) analysis of the twenty-six perfluoroisopropyl-dinitrobenzene (PFIPDNB) derivatives was performed to explain their ability to suppress photochemical activity of the plants photosystem II using chloroplasts and subchloroplast thylakoid membranes enriched in photosystem II, called DT-20. Compounds were optimized by semi-empirical PM3 and DFT/B3LYP/6-31G methods. The Heuristic and the Best Multi-Linear Regression (BMLR) method in CODESSA were used to select the most appropriate molecular descriptors and to develop a linear QSAR model between experimental pl(50) values and the most significant set of the descriptors. The obtained models were validated by cross-validation (R-cv(2)) and internal validation to confirm the stability and good predictive ability. The obtained eight models with five-parameter show that: (a) coefficient (R-2) value of the chloroplast samples are slightly higher than that of the DT-20 samples both of Heuristic and BMLR models: (b) the coefficients of the BMLR models are slightly higher than that of Heuristic models both of chloroplasts and DT-20 samples; (c) The YZ shadow parameter and the indicator parameter, for presence of NO2 substituent in the ring, are the most important descriptor at PM3-based and DFT-based QSAR models, respectively. This article is part of a Special Issue entitled: Photosynthesis Research for Sustainability: from Natural to Artificial. (C) 2012 Elsevier B.V. All rights reserved.