International Conference on Chemical Structures :: May 27-31, 2018 :: ACD/Labs
ACD Labs Logo
MENU

International Conference on Chemical Structures

May 27-31, 2018
Noordwijkerhout, Netherlands

Related Materials

Related presentations, posters, and scientific talks from this event have been posted here for your reference. Please click the associated link to download.

A Comprehensive Evaluation of ACD/LogD on a Pharmaceutical Compound Set, A. Sazonovas, K. Lanevskij, R. Didziapetris
Download

A Web-Based Informatics Platform for PhysChem/ADME/Tox Property Predictions, A. Sazonovas, K. Lanevskij, R. Didziapetris
Download



Poster Schedule

A Comprehensive Evaluation of ACD/LogD on a Pharmaceutical Compound Set
A. Sazonovas, K. Lanevskij, R. Didziapetris

Abstract:

Lipophilicity, which is often expressed in terms of 1-octanol/water partitioning coefficient logP, or the corresponding pH-dependent distribution coefficient logD, is one of the key physicochemical characteristics of any new drug candidates, as it has a major influence on a variety of the compounds' properties constituting their ADME, pharmacokinetic, and drug safety profiles. Widely available in silico tools for predicting these properties are mostly based on experimental data for simple organic chemicals and marketed drugs. Consequently, as drug discovery projects are moving to increasingly novel regions of chemical space, utility of existing methods becomes more and more questionable. In several previously published evaluation studies1,2, the mean logP prediction error for in house compound libraries of pharmaceutical companies was shown to exceed 1 log unit by almost all methods. Prediction of logD is even more challenging, as it requires accurate knowledge of both logP of neutral form and distribution of ionic forms of the compound in the relevant pH range.

With these considerations in mind, the following objectives were set for the current study:

  1. Collecting a data set of experimental logD values from recent publications dealing with novel congeneric compound series from drug discovery projects;
  2. Evaluating the performance of ACD/LogD predictor3 for the newly collected molecules using different combinations of available logP and pKa calculation algorithms;
  3. Investigating the potential for improving prediction accuracy for unknown compound classes by application of automated model training.

The compiled data set consisted of ~1200 logD values measured at physiological pH conditions. According to the initial validation results, the highest accuracy of predictions based on the models employing only built-in compound libraries can be achieved using a combination of ACD/LogP Consensus and ACD/pKa Classic algorithms, yelding RMSE slightly under 1 log unit. However, utilizing the automatic training feature of ACD/LogP GALAS algorithm by the means of stepwise addition of collected data to the model self-training library allowed decreasing the RMSE of predictions for the reserved validation set to as low as 0.6 log units. Moreover, a significant improvement (RMSE ≈ 0.8) was already evident after adding the first portion of training data constituting less than 20% of the entire data set. These results demonstrate that performing experimental measurements for a relatively small number of molecules belonging to a novel chemical series is often sufficient to adapt ACD/LogP and ACD/LogD predictors to provide reliable property estimates for the entire class of compounds.

  1. Mannhold R., Poda G. I., Ostermann C., Tetko I. V. Calculation of molecular lipophilicity: State-of-the-art and comparison of log P methods on more than 96,000 compounds. J. Pharm. Sci. 2009, 98, 861-893.
  2. Tetko I. V., Poda G. I., Ostermann C., Mannhold R. Large-scale evaluation of logP predictors: local corrections may compensate insufficient accuracy and need of experimentally testing every other compound. B. Chem. Biodivers. 2009, 6, 1837-1844.
  3. ACD/LogD (part of Percepta® platform), v. 2017, ACD/Labs, Inc. (http://www.acdlabs.com/products/percepta/)

A Web-Based Informatics Platform for PhysChem/ADME/Tox Property Predictions
A. Sazonovas, K. Lanevskij, R. Didziapetris

Abstract: ACD/Percepta Portal is a new platform that builds upon the well-established components of ACD/Percepta desktop software—reliable predictive algorithms for a multitude of physicochemical, ADME, and safety-related properties, powerful data mining, visualization, compound profiling and risk assessment capabilities, as well as ACD/Structure Design Engine for generating libraries of virtual analogs compatible with the desired characteristics. Percepta Portal combines these features with flexible network-based deployment, raising software interactivity to a new level and offering some exciting features. This work brings particular focus to the components of the web version of Percepta that leverage the power of high performance computing in a server environment. The server-side architecture relies on multiple calculation units (kernels) that enable parallel processing of very large amounts of data in real time. These capabilities paved the road for new developments in several key areas. The addition of a quick exploration of the predicted property values for a multitude of structural analogs of a compound enables on-the-fly liability checking, i.e. identifying the areas of the molecule potentially responsible for unfavorable ADME/Tox characteristics. Adaptation of the ACD/Structure Design Engine to the employed architecture gave rise to a new generation of this tool that enables extensive enumeration of substituent property space in accordance with specific user-defined constraints at up to four independently varying substituent positions at the same time. Along with a built-in database of more than 104 building blocks, this leads to exploration of up to 1016 virtual analogs, which is actually feasible in real time in Percepta Portal. Such broadened scope of the chemical space investigated, greatly enhances the potential of encountering new compounds with the most favorable property profiles.