Use of classification regression tree in predicting oral absorption in humans.

The purpose of this study is to explore the use of classification regression trees (CART) in predicting, in the dose-independent range, the fraction dose absorbed in humans. Since the results from clinical formulations in humans were used for training the model, a hypothetical state of drug molecules already dissolved in the intestinal fluid was adopted. Therefore, the molecular attributes affecting dissolution were not considered in the model. As a result, the model projects the highest achievable fraction dose absorbed, providing a reference point for manipulating the formulations or solid states to optimize oral clinical efficacy. A set of approximately 1260 structures and their human oral pharmacokinetic data, including bioavailability and/or absorption and/or radio-labeled studies, were used, with 899 compounds as the training set and 362 the test set. The numerical range of the fraction dose absorbed, 0 to 1, was divided into 6 classes with each class having a size of approximately 0.16. A set of 28 structural descriptors was used for modeling oral absorption without considering active transport. Then, a separate branch was created for modeling oral absorption involving active transport. The AAE of the training set was 0.12 and those of five test sets ranged from 0.17 to 0.2. In terms of classification, two test sets of unpublished, proprietary compounds showed 79% to 86% prediction when the predicted values fallen within +/- one class of real values were considered predicted. Overall, the computational errors from all the test sets of diverse structures were similar and reasonably acceptable. As compared to artificial membranes for ranking drug absorption potential, prediction by the CART model is considered fast and reasonably accurate for accelerating drug discovery. One can not only improve continuously the accuracy of CART computations by expanding the chemical space of the training set but also calculate the statistical errors associated with individual decision paths resulting from the training set to determine whether to accept individual computations of any test sets.

Permeation prediction of M100240 using the parallel artificial membrane permeability assay.

PURPOSE: Kansy et al first introduced the Parallel artificial membrane permeation assay (PAMPA) in 1998. In this system, the permeability through a membrane formed by a mixture of lecithin and an inert organic solvent on a filter support is assessed. PAMPA shows definite trends in the ability of molecules to permeate membranes by transcellular passive diffusion. Its simplicity, low cost, high throughput, and wide pH range make it very attractive in modern drug discovery. Based on this concept, Whohnsland et al, Sugano et al and Zhu et al modified the assay and used it to screen compound permeability. We used PAMPA for the permeation prediction of M100240, which was unable to be determined by cell-based assays due to compound instability. METHODS: In this study, 92 commercially available agents provided the structural diversity used to generate a mathematical prediction model for human fraction absorbed, M100240--an acetate thioester of MDL 100,173. Permeation of M100240 and MDL 100,173 was evaluated using the parallel artificial membrane permeability assay (PAMPA). The donor and recipient solutions consisted of 0.5N HCl (pH 1.5) or phosphate-buffered saline (pH 5.5 or 7.4) with 2% dimethyl sulfoxide. The donor solution also contained 200 mM M100240 or MDL 100,173. RESULTS: M100240 had a medium permeation at pH 5.5 (2.99%), corresponding to a high predicted Fa in humans (92%). Permeation of MDL 100,173 was low at this pH (0.72%), corresponding to a medium-to-low predicted Fa (46). At pH 7.4, the permeation of M100240 was low (approximately 1%) and no permeation was apparent for MDL 100,173. CONCLUSIONS: We predicted M100240 is likely to be well absorbed via passive diffusion across the human gastrointestinal tract following oral administration.

A comparative study of artificial membrane permeability assay for high throughput profiling of drug absorption potential.

Artificial membrane permeability measurement is a potentially high throughput and low cost alternative for in vitro assessment of drug absorption potential. It will be an ideal screening/profiling tool in the lead generation program of drug discovery research if it is proven to be generally applicable for classifying drug absorption potential and is advantageous over other in vitro or in silico methods. This study provides an in-depth evaluation of the method in close comparison to Caco-2, LogD, LogP, polar surface area (PSA), and quantitative structure-property relationship (QSPR) predictions using a large and diverse compound set. It showed that the accuracy of using artificial membrane permeability in assessing drug absorption is comparable to Caco-2, but significantly better than LogP, LogD, PSA, and QSPR predictions. This study also explored the artificial membrane composition by adopting a hydrophilic filter membrane for artificial membrane (lecithin-dodecane) support. The use of hydrophilic filter membrane increased the rate of permeation significantly and reduced the transport time to 2 h or less as compared with over 10 h when a hydrophobic filter membrane is used.

Graphical model for estimating oral bioavailability of drugs in humans and other species from their Caco-2 permeability and in vitro liver enzyme metabolic stability rates.

This paper describes a graphical model for simplifying in vitro absorption, metabolism, distribution, and elimination (ADME) data analysis through the estimation of oral bioavailability (%F) of drugs in humans and other species. This model integrates existing in vitro ADME data, such as Caco-2 permeability (P(app)) and metabolic stability (percent remaining - %R) in liver S9 or microsomes, to estimate %F into groups of low, medium, or high regions. To test the predictive accuracy of our model, we examined 21 drugs and drug candidates with a wide range of oral bioavailability values, which represent approximately 10 different therapeutic areas in humans, rats, dogs, and guinea pigs. In vitro data from model compounds were used to define the boundaries of the low, medium, and high regions of the %F estimation plot. On the basis of the in vitro data, warfarin (93%), indomethacin (98%), timolol (50%), and carbamazepine (70%) were assigned to the high %F region; propranolol (26%) and metoprolol (38%) to medium %F region; and verapamil (22%) and mannitol (18%) to the low %F region. Similarly, the %F of 11 drug candidates from Elastase Inhibitor, NK1/NK2 antagonist, and anti-viral projects in rats, guinea pigs, and dogs were correctly estimated. This model estimates the oral bioavailability ranges of neutral, polar, esters, acidic, and basic drugs in all species. For a large number of drug candidates, this graphical model provides a tool to estimate human oral bioavailability from in vitro ADME data. When combined with the high throughput in vitro ADME screening process, it has the potential to significantly accelerate the processes of lead identification and optimization.

Site-specific absorption of M100240 and MDL 100,173 in rats evaluated using Sweetana-Grass diffusion chamber technology.

INTRODUCTION: M100240--an acetate thioester of MDL 100,173--a dual angiotensin-converting enzyme (ACE)/neutral endopeptidase (NEP) inhibitor is in Phase II development for cardiovascular conditions. The absorption characteristics of M100240 and MDL 100,173 were evaluated in specific regions of the rat gastrointestinal (GI) tract. METHODS: Solutions of M100240 and MDL 100,173 were prepared using mucosal Kreb's solution. Four male Sprague-Dawley rats, fasted before tissue procurement, were sacrificed using 95% CO2 and 5% O2. The entire small and large intestine were removed, rinsed in serosal Kreb's solution, and segments mounted onto the Sweetana-Grass diffusion chamber. Test compounds were added to the donor compartment and drug-free oxygenated serosal Kreb's solution to the receiving compartment. The temperature of the chambers was maintained at 37 degrees C and supplied with 95% O2/5% CO2. Samples were removed (0.5 ml) at 0, 30, 60, 90, 120, 180, and 240 min from the serosal side. The volume was maintained with drug-free warm serosal Kreb's solution. Samples were diluted with rabbit plasma and analyzed by liquid chromatography/mass spectrometry. Apparent permeation values [Papp (cm/min)] of M100240 and MDL 100,173 through the duodenum, jejunum, ileum, and colon were determined. RESULTS: The mean Papp of M100240 and MDL 100,173 was highest in the duodenum: 2.29 x 10(-4) +/- 2.40 x 10(-4) and 1.66 x 10(-4) cm/min +/- 8.33 x 10(-5), respectively. The mean Papp was lowest in the colon: 3.61 x 10(-6) +/- 3.44 x 10(-6) and 1.62 x 10(-5) +/- 3.21 x 10(-6) cm/min for M100240 and MDL 100,173, respectively. Absorption of MDL 100,173, however, was evident throughout the rat GI tract. DISCUSSION: M100240 and MDL 100,173 are predominantly absorbed from the duodenum in the rat GI tract. MDL 100,173 is also absorbed from the jejunum, ileum, and colon. These results, consistent with data obtained in humans, demonstrate the potential predictive value of the Sweetana-Grass model for site of absorption assessments.