1 2 Research paper 4 In silico modelling of permeation enhancement potency in Caco-2 5 monolayers based on molecular descriptors and random forest 6 7 8 Søren H. Welling a,b , Line K.H. Clemmensen b , Stephen T. Buckley a , Lars Hovgaard a , Per B. Brockhoff b , 9 Hanne H.F. Refsgaard a,⇑ 10 a Global Research, Novo Nordisk A/S, Novo Nordisk Park, 2760 Måløv, Denmark 11 b Technical University of Denmark, DTU Compute, 2800 Kgs. Lyngby, Denmark 12 13 15 article info 16 Article history: 17 Received 30 January 2015 18 Revised 14 May 2015 19 Accepted in revised form 17 May 2015 20 Available online xxxx 21 Keywords: 22 Permeation enhancers 23 Caco-2 24 Random forest 25 QSAR 26 Surfactants 27 28 abstract 29 Structural traits of permeation enhancers are important determinants of their capacity to promote 30 enhanced drug absorption. Therefore, in order to obtain a better understanding of structure–activity rela- 31 tionships for permeation enhancers, a Quantitative Structural Activity Relationship (QSAR) model has 32 been developed. 33 The random forest-QSAR model was based upon Caco-2 data for 41 surfactant-like permeation enhan- 34 cers from Whitehead et al. (2008) and molecular descriptors calculated from their structure. 35 The QSAR model was validated by two test-sets: (i) an eleven compound experimental set with Caco-2 36 data and (ii) nine compounds with Caco-2 data from literature. Feature contributions, a recent developed 37 diagnostic tool, was applied to elucidate the contribution of individual molecular descriptors to the pre- 38 dicted potency. Feature contributions provided easy interpretable suggestions of important structural 39 properties for potent permeation enhancers such as segregation of hydrophilic and lipophilic domains. 40 Focusing on surfactant-like properties, it is possible to model the potency of the complex pharmaceutical 41 excipients, permeation enhancers. For the first time, a QSAR model has been developed for permeation 42 enhancement. The model is a valuable in silico approach for both screening of new permeation enhancers 43 and physicochemical optimisation of surfactant enhancer systems. 44 Ó 2015 Published by Elsevier B.V. 45 46 47 48 49 1. Introduction 50 Development of oral delivery systems for proteins and peptides 51 offers the promise of improved patient compliance compared to 52 conventional parenteral administration. However, bioavailability 53 is, in part, limited due to poor absorption of proteins across the 54 intestinal epithelial barrier. To effectively deliver a protein 55 systemically this barrier can be modulated by the presence of 56 permeation enhancers [1]. 57 Quantitative Structural Activity Relationship, QSAR methods 58 have been applied extensively for exploration of structural proper- 59 ties of importance for oral absorption of new chemical entities, e.g., 60 QSAR models have been developed for permeability [2] and solu- 61 bility [3–5]. To our knowledge, no QSAR model for permeation 62 enhancement has previously been published. 63 Some permeation enhancers have specific mechanisms of 64 action, e.g., modulating the function of tight junctions in the 65 plasma membrane such as zona-occludens-toxin [6], EDTA [7] or 66 melittin [8]. However, the majority of permeation enhancers are 67 primarily surfactants and will non-specifically disrupt the lipid 68 bilayer packing of phospholipids in the epithelial membrane [1]. 69 Surfactants are molecules having segregated lipophilic and hydro- 70 philic domains. Water soluble surfactants tend to pool in the sur- 71 faces of water/air and water/lipid, lowering the surface tension. 72 Lowering of surface tensions of water/air surfaces and the ability 73 to enhance the permeability across lipid bilayers correlated well http://dx.doi.org/10.1016/j.ejpb.2015.05.012 0939-6411/Ó 2015 Published by Elsevier B.V. Abbreviations: C6, sodium hexanoate; C8, sodium octanoate; c8G, octylglucoside; C10, sodium decanoate/caprate; c12PC, dodecanoylphosphocholine; c12GPC, dode canoylglycerophosphocholine; c14GP, myristoylglycerophosphate; CART, classification and regression tree; CDC, chenodeoxycholate; DDM, dodecylmaltoside; EDTA, ethyldiaminetetraaceticacid; GCC, glycochenocholate; GH, glycyrrhizinate; LCC, lauroylcarnitinechloride; LOO-CV, leave-one-out cross validation; MOE, molecular operating environment; PCC, palmitoyl carnitine chloride; QSAR, quantitative structural activity relationship; SM, simomenine; RMSE, root mean square error; r p , Pearsons correlation coefficient; r s , Spearman rank correlation coefficient; SD, standard deviation; TEER, transepithelial electrical resistance; TDM, tetradecylmaltoside; TDS, sodium tetradecyl sulphate; TDM, tetradodecyl maltoside; TC, taurocholate; T pot , TEER potency; UC, Ursocholate. ⇑ Corresponding author at: Insulin Pharmacology Research, Novo Nordisk A/S, Novo Nordisk Park, 2760 Måløv, Denmark. E-mail address: hare@novonordisk.com (H.H.F. Refsgaard). European Journal of Pharmaceutics and Biopharmaceutics xxx (2015) xxx–xxx Contents lists available at ScienceDirect European Journal of Pharmaceutics and Biopharmaceutics journal homepage: www.elsevier.com/locate/ejpb EJPB 11942 No. of Pages 8, Model 5G 26 May 2015 Please cite this article in press as: S.H. Welling et al., In silico modelling of permeation enhancement potency in Caco-2 monolayers based on molecular descriptors and random forest, Eur. J. Pharm. Biopharm. (2015), http://dx.doi.org/10.1016/j.ejpb.2015.05.012