ALOGPS + logD data = logD program!
The ALOGPS was developed to predict 1-octanol/water partition coefficients, logP, and aqueous solubility of neutral compounds only. However, the self-learning properties of Associative Neural Networks used in this program make possible to predict new properties that were not used to develop the program. One of such important properties is pH-dependent distribution coefficient, logD. Of course, since no logD data were used to develop the ALOGPS program, it can't really predict this value in "as is" mode, i.e. without any additional training. However, if the user has experimental values for (few) such compounds, he/she can get excellent results with our program.
To use this option the user needs to create and upload the file with SMILES followed by experimental logD values (see What are the steps to create my own LIBRARY?) in the same way as he does it for the logP library. The logD library will be created in the memory of program and the program will be able to predict logD values for similar compounds, i.e. compounds from the same chemical series.
The success rate will critically depend on the amount of compounds with logD values. However, sometimes (particularly if number of compounds with logD values is small) better results can be calculated if internal logP library, that is used by default in the ALOGPS program, will be discarded. This possibility is only available for standalone version that can be downloaded from our web site too. To use this option simply substitute logp.bin file with the empty logp.bin file provided on this site and restart the standalone version of the program (either applet or exe file). Actually, you can try both ways (with the original logP library or with the empty library) and see which one is more suitable to you. In both cases you should have a dramatic improvement in your results. Interested? Try it now!
Some benchmarking results of the ALOGPS and some commercially available programs for logD prediction could be found in
This page was created by Igor V. Tetko on 11/08/2004.
Copyright 2001 -- 2023 https://vcclab.org. All rights reserved.