ORIGINAL RESEARCH Challenges in docking 2 0 -hydroxy and 2 0 ,4 0 -dihydroxychalcones into the binding site of ALR2 Sorin I. Avram • Luminita Crisan • Liliana M. Pacureanu • Alina Bora • Edward Seclaman • Monica Balint • Ludovic G. Kurunczi Received: 25 June 2012 / Accepted: 13 November 2012 / Published online: 28 November 2012 Ó Springer Science+Business Media New York 2012 Abstract Polyhydroxylated chalcones are able to effi- ciently inhibit aldose reductase (ALR2; EC 1.1.1.21), an enzyme implicated in the development of diabetic compli- cations. In this study, we performed docking experiments on a series of 38 2 0 -hydroxy and 2 0 ,4 0 -dihydroxychalcones with measured inhibitory activities. We demonstrate that docking the chalcones into a single ALR2 active site conformation oversimplifies the attempt to find a probable binding model. Scoring functions exerted relative low agreement concerning a common ALR2 conformation, and suggest that relative to the substituents, chalcones bind to different ALR2 confor- mations. We found plausible binding modes of 2 0 ,4 0 -dihydr- oxychalones for both closed- and open-state conformations of the ALR2 active site, while 5 0 -chloro, 2 0 -hydroxychalcones would bind into an ITB (2-(carboxymethyl)-1,3,3-trioxo- benzo[e][1,2]benzothiazole-4-carboxylic acid)-like binding site conformation. The relative poor reliability of docking programs to recognize a ligand binding into a closed- or open- form conformation of the ALR2 binding site was highlighted by a cross-docking experiment of the native ligands. In this context, we employed conditional interference forests to build classification models aiming to predict the conformational state of the ALR2 ligand-binding site in complex with chal- cone derivatives. Internally and externally validated models suggest that 2 0 -hydroxy and 2 0 ,4 0 -dihydroxichalones leave the specificity pocket closed. Another issue discussed in the current paper concerns the errors related to the average position of the atoms in the crystals, which can heavily influence the docking poses. The diffraction-component pre- cision index (DPI) is able to reflect these errors, but some structures have no calculated values. In a comparative experiment on 55 ALR2 crystal structures, we found the Blow DPI to better approximate the Cruickshank DPI compared to Goto. Keywords Aldose reductase  ALR2  Chalcones  Flavonoids  Docking  Conditional interference forest Abbreviations ALR2 Aldose reductase ALR1 Aldehyde reductase DPI Diffraction-component precision index ADV AutoDock Vina CIF Conditional interference forest OOB-Acc ‘‘out-of-bag’’ prediction accuracy LD-Acc Learning data accuracy Introduction Chalcones are 1,3-diphenyl-2-propene-1-one derivatives with the two aromatic rings joined by a three-carbon a,b- unsaturated carbonyl system. In nature, they occur as Electronic supplementary material The online version of this article (doi:10.1007/s00044-012-0367-5) contains supplementary material, which is available to authorized users. S. I. Avram  L. Crisan  L. M. Pacureanu  A. Bora Department of Computational Chemistry, Institute of Chemistry of Romanian Academy, Timisoara, 24 Mihai Viteazul Avenue, 300223 Timisoara, Romania E. Seclaman Department of Biochemistry, Faculty of Medicine, University of Medicine and Pharmacy ‘‘Victor Babes’’, 2 Eftimie Murgu, 300041 Timisoara, Romania M. Balint  L. G. Kurunczi (&) Department of Pharmacy I, Faculty of Pharmacy, University of Medicine and Pharmacy ‘‘Victor Babes’’, 2 Eftimie Murgu, 300041 Timisoara, Romania e-mail: kurunczi@umft.ro 123 Med Chem Res (2013) 22:3589–3605 DOI 10.1007/s00044-012-0367-5 MEDICINAL CHEMISTR Y RESEARCH