Leading Edge Predictors for Drug Discovery

CSLogWS Home

CSLogD Home

CSLogP Home

CSpKa Home



CSGenoTox Home



     ...Predictions and pH Profiles

Prediction of Log D property values

CSLogD Method and Design

CSLogD is based strictly on the use of derived expressions ( see CSLogD Introduction ) for LogD, which involve predicted values of LogP, pKa's, and the partition coefficients of ionizable species of a given chemical entity. It is able to deliver LogD for numerous ampholytes (12 in all) and simple monoprotic compounds, and their important LogD pH-profiles.

Experimental Data Used in the Modeling Process

A comparison of LogD calculated vs. experimental values was made using a diverse set of known drugs measured under the same method and conditions (1).

Method:  Rapid shake-flask at pH: 7.4

(1).M. Kansy, H. Fischer, K. Kratzat, F. Senner,  B. Wagner, and I. Parrilla  “High-Throughput Artificial Membrane Permeability Studies in Early Lead Discovery and Development” in Pharmacokinetic Optimization in Drug Research, Eds. B. Testa, H. van de Waterbeemd, G. Folkers, and R. Guy, Publ. Wily-VCH (2001).

Predicted Data Resulting From the Modeling Process

All predicted LogD values were made using CSLogD and the 2D structures of the drugs. The latter was obtained from the Merck Index (ver.13).  Observed LogP values were for neutral compounds taken from several sources (2,3,4) and used to compare calculated LogP results employed internally within the CSLogD program.

(2). Kow from Sangster Research Laboratories, Montreal Quebec, Canada.

(3). CLogP, Biobyte StarList

(4). PhysProp, Syracuse Research Corp., Syracuse, NY.

Percent Flux vs LogD experimental

The scatter plot to the right shows variation in %Flux vs LogD(experimental) using PAMPA, an artificial passive permeability membrane technology. The range in LogD(experimental) extended from -1.3 to 4.0, covering a range of 5.3 log units. The lowest %Flux corresponded quite well with negative or high values of LogD, with highest %Flux in the  0.5 to 2.0 range for LogD

CSLogD was used to predict the LogD for the 78 drugs and those predictions are shown in the following plot.  These predictions were made over the entire experimental range of LogD using pKa's and LogP computed internally by CSLogD prediction software.

External Validation of Log D Predictions

A correlation of the CSLogD (predicted) values with the known LogD(experimental) values (taken at pH 7.4) gave the following statistics for this external validation set:

Q2valid = 0.80

MAE = 0.52 (mean absolute error)

Given the great diversity of chemical space represented by these compounds (see Representative Compounds from CSLogD Validation Set), the results are demonstrative of the predictive abilities of CSLogD

CSLogD is not built on LogD values but only pKa's and LogP.

It is important to note that pH-profiles of LogD are equally important for oral absorption of therapeutic agents.  The pH-profiles for three acid-base ampholytic drugs are presented below.

LogD pH Profiles


Hydrochlorothiazide is a widely used diuretic, that is representative of the family of benzothiadiazine sulfonamide derivatives, commonly known as thiazides.

LogD experimental  =  0.04 at pH 7.4

The hydrochlorothiazide pH-profile has the characteristic U-shape of base-acid ampholyte.  Here there is a protonable secondary amine with the sulfonamide being the acid group, pKa ~ 9.3.


It is naturally occurring alkaloid that is commonly used as bronchodilator

LogD experimental = 0.13 at pH 7.4 (1)

The pH-lipophilicity profile is shown below.  The flatness in the profile is typical of an acid-base ampholyte with a large difference in pKa's for basic and acidic groups.  Here the large difference in pka's is between the aromatic N and the NH group in imidazole ring.


Acrivastinf is an H1-receptor antagonist acting as an antihistamine.  Acrivastine has two basic groups and a carboxylic acid, which suggests a pH profile similar to Theophylline due to its large differences in its pKa values.

LogD experimental = 0.20 at pH 7.4

The shape of the pH-lipophilicity profile is quite different from the normal U-shaped curves expected for this type of ampholyte as shown below.  The zwitterion was found in both the octanol and water phases (2)

CSLogD did reasonably well in characterizing not only the general shape found experimentally but provides reasonable values for logD at pH's 2.0 and 11.0 given that the zwitterion partitions into both phases.

(1). R. Scherrer (2001) “Biolipid pKa Values and the Lipophilicity of Ampholytes and Ion Pairs” in Pharmacokinetic Optimization in Drug Research, Eds. B. Testa, H. van de Waterbeemd, G. Folkers, and R. Guy, Publ. Wily-VCH.

(2). A. Pagliara, P.A. Carrupt, P. Gaillard, B. Testa, (1997), Chem. Rev. 97, 3385

CSLogD  Representative Compounds

Follow the link below to a set of 34 representative compounds  of the 86 used in external validation testing of CSLogD.  Each structure is given with a comparison of known experimental values with predicted LogD and LogP.

Go to: CSLogD  Compounds

Back to: CSLogD  Home Page

user login
contact us

To contact us:

Phone: (888) 636-8777

Fax: 781-275-5197

Email:  sales@chemsilico.com

Copyright © 2003 ChemSilico LLC All Rights Reserved

Terms and Conditions of Use | Privacy Policy

ChemSilico is a registered trademark of ChemSilico LLC, Tewksbury, MA 01876