|
The LIBRARY mode FAQ
-
What is the LIBRARY?
-
What are the advantages to use my own LIBRARY?
-
What do I need to create my own LIBRARY?
-
What are the steps to create my own LIBRARY?
-
How can I use the created LIBRARY?
-
What are the tricks to get maximal performance of
my LIBRARY?
-
Where can I find some literature about the LIBRARY
mode?
What is the LIBRARY?
LIBRARY or user's LIBRARY is a set of compounds with known experimental
values that was processed by ALOGPS program plus a set of neural network
parameters optimized to provide maximal prediction ability of the method
for new user's compounds.
What are the advantages to use my own LIBRARY?
The use of the user's LIBRARY can increase prediction ability of
ALOGPS program in 2-5 times. The more
compounds are used in the LIBRARY, the better results will be calculated.
However, a use even of a single compound can increase predictive ability
of ALOGPS program in hundreds of percent.
What do I need to create my own LIBRARY?
You need compounds with measured logP or logS values. Even 1 or
2 such compounds could be enough to dramatically improve prediction ability
of ALOGPS for similar compounds from your series.
What are the steps to create my own LIBRARY?
-
Convert your structures to SMILES representation. On WWW you can upload
such structures in a batch mode by ALOGPS and use the resulting file as
a source of SMILES.
-
Provide logP or logS {in log10(mol/L)
units } after each SMILES, e.g.
CC 1.81
CI 1.51
BrC 1.19
C=C 1.13
-
Upload your data using corresponding "upload data" button on the front
page of the ALOGPS program.
-
Wait until a corresponding message will inform you that your library has
been created.
Your library is ready! You can examine it by clicking on LogP_LIBRARY or
LogS_LIBRARY link on the front page of the applet. If library was successfully
created, you will see instead of this message a set of processed compounds.
How can I use the created LIBRARY?
You can provide compounds in SINGLE or BATCH mode as before. Your
LIBRARY will be automatically used. The LIBRARY will be destroyed each
time if you quit the browser or applet.
What are the tricks to get maximal performance of the LIBRARY mode?
The compounds in LIBRARY and the compounds to be predicted should
be similar. If you have two chemically different series of compounds (i.e.,
if each series has some large common fragment), create a LIBRARY for each
series independently.
Where can I find some literature about the LIBRARY mode?
This article provides general description of the main properties
of the Associative Neural Network (ASNN), the method that is used in the
LIBRARY mode. It explains how the nearest neighbors are calculated in the
space of neural networks models and how this method can be used to provide
clustering in the output space of parameters. Several examples are used
to demonstrate the properties of the method, including testing of ALOGPS
program at BASF.
-
Tetko, I.V. Associative Neural Network, Neural Processing Letters, 2002,
16, 187-199.
This is more theoretical article that demonstrates properties of
ASNN for regression and classification tasks using theoretical examples.
-
Tetko, I.V. and Tanchuk, V.Yu, Application of Associative Neural Networks
for Prediction of Lipophilicity in ALOGPS 2.1 program, J.
Chem. Inf. Comput. Sci., 2002, 42, 1136-1145.
This theoretical article analyses in details properties of the LIBRARY
mode in ALOGPS 2.1 program. It is shown that in some cases the predictive
ability of this program can be increased in 2-5 times using only few molecules
from the user's set.
-
Tetko, I.V. Associative Neural Network, CogPrints
archive code: cog00001441.
This is the first article that introduced ASNN and that was published on
WWW in April 2001. The idea of ASNN was created on January 7, 2001 and
the name of the method was given in March 2001.
This page was created by Igor V. Tetko on 06/06/2002.
See FAQ if you have questions. How to cite this applet? Are you looking for a new job in chemoinformatics? |
|
