DMSO Solubility

pdf version (598KB)

Prediction of a compound's solubility in DMSO (S_DMSO) plays a vital role in sample management and global drug discovery. The distribution of samples as DMSO solutions is preferable to distribution of solids/powders, since these require an extra dissolution step when prepared for analytical assays. DMSO solutions can be easily diluted with water, thus excluding the need of sample weighing and dissolving before the analysis. Compounds that are insoluble in DMSO cannot be handled using standard laboratory protocols, so their development is costly. Furthermore, plating out these compounds is severely hampered and may even lead to incorrect test results.

AB/DMSO is an ideal tool for calculating S_DMSO values in batch mode. The AB/DMSO predictor uses a data set of >20,000 compounds with S_DMSO cut-off at 20 mM. This set represents a very diverse set of drug-like structures, which were taken from the Specs collection of compounds. This particular compound/data set has been obtained upon plating out more than 200,000 different compounds over a period of 10 years. All these compounds were checked for purity, and for the development of this particular algorithm only compounds with the highest (>90%) purity were selected. The data set that was eventually used (which is only a subset of the full data set) was divided into training (15,584) and validation (6,679) sets. Algorithms were developed entirely on the training set and their accuracy was tested on the validation set.

AB/DMSO was developed using the Algorithm Builder (AB) development application. The combination of high quality data and the powerful AB toolset (including proprietary statistical methods) led to the development of an accurate and robust predictor. The AB/DMSO algorithm is based on three sub-algorithms. Each sub-algorithm has a unique generality and control system that directs which sub-algorithm should be applied in every single case. This mechanism achieves very good results for compounds that are similar to the training set and avoids steep drops in accuracy for compounds that are less similar to the training set. In order to check generality and applicability of the AB/DMSO algorithm, a calculation of solubility and its reliability was executed on the World Drug Index. Analysis of results showed that the algorithm of highest reliability was valid for ca. 30% of compounds, while low reliability only applied for less than 1% of the structures. This example shows that AB/DMSO can be used with confidence on any data set of drug/lead-like molecules.

The Algorithm Builder development platform also presents an "open product" concept, e.g. it is not just a "black box" solution. Databases and project files are available as additional product options providing access to complete information about the AB/DMSO algorithm and the ability to customize and modify it utilizing in-house knowledge.

Features :

Additional Project Options :

• Full database with experimental solubility data.
• Project file, containing all workflow of the AB/DMSO algorithm. Modify or even rebuild the existing algorithm using Algorithm Builder.
• Distribute AB/DMSO algorithm from in-house applications, through the Pharma Algorithms - Algorithm Launching Pack (ALP).

Development partner :

More Information :

DMSO Solubility Prediction
P. Japertas, P. Maas and A. Petrauskas
Poster presentation at DDT World Congress, August 2003
pdf version (296KB)

DMSO Solubility Prediction
Powerpoint slides in addition to Dr. Chris Lipinski presentation
DDT World Congress, Scientific Session, August 13, 2003
ppt (2.3MB)

If you are interested in evaluating AB/DMSO, e-mail us for a trial version. Please include your full contact information.