Decision Trees to Characterise the Roles of Permeability and Solubility on the Prediction of Oral Absorption
Total Page：16
File Type：pdf, Size：1020Kb
Accepted Manuscript Decision trees to characterise the roles of permeability and solubility on the prediction of oral absorption Danielle Newby , Alex. A. Freitas , Taravat Ghafourian PII: S0223-5234(14)01115-5 DOI: 10.1016/j.ejmech.2014.12.006 Reference: EJMECH 7566 To appear in: European Journal of Medicinal Chemistry Received Date: 24 April 2014 Revised Date: 2 December 2014 Accepted Date: 3 December 2014 Please cite this article as: D. Newby, A.A. Freitas, T. Ghafourian, Decision trees to characterise the roles of permeability and solubility on the prediction of oral absorption, European Journal of Medicinal Chemistry (2015), doi: 10.1016/j.ejmech.2014.12.006. This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain. ACCEPTED MANUSCRIPT MANUSCRIPT All Data SdssS SdssS _acnt _acnt >1 >1 logP> logP= 5 <5 High or Low Oral ACCEPTEDAbsorption 1 Decision trees to characterise the roles of permeability and solubility on the prediction 2 of oral absorption 3 Danielle Newby a, Alex. A. Freitas b, Taravat Ghafourian a,c* 4 5 aMedway School of Pharmacy, Universities of Kent and Greenwich, Chatham, Kent, ME4 6 4TB, UK b 7 School of Computing, University of Kent, CanterburyA,CCEPTED Kent, CT2 MANUSCRIPT7NF, UK 8 c Drug Applied Research Centre and Faculty of Pharmacy, Tabriz University of Medical 9 Sciences, Tabriz, Iran 10 *Corresponding Author: Email: [email protected]; Tel +44(0)1634 202952; Fax 11 +44 (0)1634 883927 12 13 14 15 16 17 MANUSCRIPT 18 19 20 21 22 ACCEPTED 23 24 1 25 ABSTRACT 26 Oral absorption depends on many physiological, physiochemical and formulation factors. 27 Two important properties that govern oral absorption are in vitro permeability and solubility, 28 which are commonly used as indicators of human intestinal absorption. Despite this, the 29 nature and exact characteristics of the relationship between these parameters are not well 30 understood. In this study a large dataset of human intestinal absorption was collated along 31 with in vitro permeability, aqueous solubility, meltingA CCEPTEDpoint, and maximumMANUSCRIPT dose for the same 32 compounds. The dataset allowed a permeability threshold to be established objectively to 33 predict high or low intestinal absorption. Using this permeability threshold, classification 34 decision trees incorporating a solubility-related parameter such as experimental or predicted 35 solubility, or the melting point based absorption potential (MPbAP), along with structural 36 molecular descriptors were developed and validated to predict oral absorption class. The 37 decision trees were able to determine the individual roles of permeability and solubility in 38 oral absorption process. Poorly permeable compounds with high solubility show low 39 intestinal absorption, whereas poorly water soluble compounds with high or low permeability 40 may have high intestinal absorption provided that they have certain molecular characteristics 41 such as a small polar surface or specific topology. 42 KEYWORDS 43 Intestinal absorption, permeability, solubility, decision trees, QSAR 44 ABBREVIATIONS USED 45 %HIA, percentage human intestinal absorption; BCS, BiopharmaceuticsMANUSCRIPT Classification 46 System; CART, Classification and regression trees; Caco-2, Human colon adenocarcinoma 47 cell line; FN, false negative; FP, false positive; GSE, general solubility equation; MDCK, 48 Madin-Darby Canine Kidney; MPbAP, melting point based absorption potential; QSAR, 49 Quantitative Structure-Activity Relationship; SE, sensitivity; SP, specificity; TN, true 50 negative; TP, true positive 51 ACCEPTED 2 52 1. INTRODUCTION 53 The assessment of pharmacokinetic properties, especially absorption, is now well established 54 in early drug discovery. The need to determine absorption of new chemical entities is 55 essential for successful orally administered compounds, as well as efficacy, toxicity and other 56 ADME (absorption, distribution, metabolism, excretion) properties [1]. The prediction of oral 57 absorption can be carried out with experimental assays and/or the use of in silico models. 58 These experimental and computer models can beA CCEPTEDused as anMANUSCRIPT indication of intestinal 59 absorption in humans, which is carried out later on in drug development. By testing drug 60 compounds using these models, compounds with undesirable properties can be removed 61 earlier, therefore improving cost effectiveness [2, 3]. 62 Intestinal absorption depends on many physiological, physiochemical and formulation 63 factors. Two important properties that govern oral absorption are permeability and solubility 64 as utilised by the Biopharmaceutics Classification System (BCS) [4]. For a drug to be 65 absorbed it must firstly dissolve in the gastrointestinal fluid in order to then permeate the 66 intestinal membrane. The relationship between these properties is closely, usually inversely, 67 related [5, 6]. As an increasing number of new chemical entities (NCE) have high 68 lipophilicity and low solubility, predicting absorption of NCEs is problematic. Inadequate 69 aqueous solubility can lead to poor, erratic, variable absorption, so it is important to consider 70 the effects of solubility for the prediction of intestinal absorption [7] . 71 The importance of solubility on oral absorption is highlighted in the literature, but there are 72 few studies that incorporate both experimental solubility and permeabilityMANUSCRIPT values within a 73 model, in order to see the effect these two properties have on oral absorption [8, 9]. Early oral 74 absorption models are too small to effectively represent all the different biological processes 75 of absorption and the physiochemical properties including solubility [10, 11]. Most studies 76 have removed compounds with solubility issues when modelling oral absorption [12, 13], 77 which is not ideal due to the increasing number of poorly soluble drugs being developed. 78 Zhao and co-workers demonstrated that predicting BCS Class II compounds (low solubility 79 and high permeability) resulted in an overestimation of fraction absorbed by their model [12]. 80 Solubility itself is a complex parameter and in turn dependent on numerous factors, therefore 81 it is important to investigate what multipleACCEPTED elements such as those calculated from the 82 molecular structure may improve understanding of this property in relation to absorption. 3 83 Molecular descriptors that describe the process of solubilisation of the drug such as crystal 84 lattice energy, solvent cavity formation energy and solvation energy are utilised in the 85 prediction of solubility [14, 15]. The general solubility equation (GSE) is a simple method 86 that predicts aqueous solubility using only two parameters, logP and melting point [16]. 87 Other methods may employ more specific molecular descriptors to improve the prediction 88 accuracy [17, 18]. GSE and its variants have been used for the estimation of oral absorption- 89 related parameters termed absorption potential [19,A CCEPTED20]. Recently MANUSCRIPT a melting point based 90 absorption potential (MPbAP) has been proposed which is derived from the GSE and 91 includes maximum dose, to give an indication of oral absorption. In general, it was found that 92 the lower the melting point the higher the tendency the compound had to be highly-absorbed 93 and vice versa , and it was also found that for higher melting points absorption was limited by 94 dose [21]. 95 Permeability in drug discovery is routinely measured using in vitro cell based assays to give 96 an indication of permeability of drug compounds in the intestine, blood brain barrier, nasal 97 cavity and skin [22]. Apparent permeability (Papp ) is the rate of permeation across cell 98 monolayers and is usually measured in cm/s -1. The ideal permeability model for the small 99 intestine mimics the physical and biochemical processes of intestinal absorption [1]. There 100 are many different cell lines that can be used to measure permeability. Human colon 101 adenocarcinoma (Caco-2) is a commonly used cell line [23-25], which displays biological and 102 characteristic properties of the enterocytes of the small intestine such as the brush border and 103 tight junctions [1, 25-27]. These cells can express a variety of transporters and metabolic 104 enzymes, allowing other transport and metabolism mechanisms toMANUSCRIPT be investigated [28]. 105 Drawbacks of this particular cell line are inter-laboratory differences, variable transporter 106 expression, long culture time, tighter junctions compared with in vivo situation and lack of 107 mucus secreting goblet cells [1, 29, 30]. Some of these problems have been resolved by other 108 cell lines such as 2/4/A1, a rat intestinal epithelial cell line, which has leakier tight junctions 109 [31, 32]; also, the cell line HT29-MTX is a co-culture of Caco-2 cells with mucus secreting 110 goblet cells to study the effects of mucus on absorption [33]. Another cell line that has been 111 gaining popularity is MDCK II (Madin-Darby Canine Kidney strain II) cells, due to shorter 112 culture time (of 3-5 days), leakier tight junctions and low expression