Back to the Top
The following message was posted to: PharmPK
I'd like to inform PharmPK subscribers that Simulations Plus, Inc. has
renamed its popular QMPRPlus(TM) and QMPRchitect(TM) software programs.
The new names are ADMET Predictor(TM) and ADMET Modeler(TM).
We have made these changes for the convenience of both our scientific
and investor communities. New versions of both programs will be released
shortly with the new names. Of course, all of our customers who have
current licenses for QMPRPlus and QMPRchitect will receive updated
versions when they are released.
ADMET Predictor is a structure-to-property prediction program that
accepts structures in SMILES or MDL format (.mol, .sdf, or .rdf) and
uses artificial neural network ensembles to predict about 50 properties,
including pKa (based on complete microequilibria theory, with
microspecies distributions identified at each charge state), logP,
solubility (native, intrinsic, and at user-specified pH, all either with
or without melting point information), human effective jejunal
permeability, MDCK apparent permeability, likelihood of blood-brain
barrier permeation, human volume of distribution, human plasma protein
binding, and fraction absorbed (by simulation) at 1, 10, 100, and 1000
mg. For reference, solubility prediction using the Meylan-Howard model
and logP using the Moriguchi model are also provided. The program is
extremely fast, with the ability to generate all predictions in batch
mode for millions of compounds in a day, and is compatible with the
SciTegic Pipeline Pilot(TM) software. Most model RMSEs are in the range
of 0.3-0.4 log units, with some less than 0.3 and a few above 0.4.
In the upcoming release, structure depiction has been added, along with
prediction of logD at user-specified pH. In addition, a new optional
Toxicity Module, to be released shortly, will include predictions for
Maximum Recommended Therapeutic Dose, salmonella mutagenicity, rat
estrogen receptor based toxicity, fathead minnow acute toxicity, rat
chronic carcinogenicity, and mouse chronic carcinogenicity. Additional
toxicity predictions are in development and will be released as upgrades
to the Toxicity Module as they are completed.
ADMET Modeler is a program that automates the complex process of
building the artificial neural network ensemble models employed in ADMET
Predictor, allowing researchers to quickly build high-quality models for
any property from their own data and add them to the ADMET Predictor
outputs. The program automatically performs descriptor filtering,
Kohonen mapping, selecting of training, verification, and test sets,
genetic algorithm, and training a series of artificial neural network
ensembles across a user-specified range of architectures (numbers of
descriptors and numbers of neurons). The user selects the
best-performing architecture for the final model, and exports it without
ADMET Predictor with a mouse click.
In our hands, once an accurate input data set is available, ADMET
Modeler has reduced the time required to build high quality models from
2-3 months to less than a day. The upcoming release will include the
ability to train both artificial neural network ensembles and support
vector machine ensemble models.
Chairman & CEO
Simulations Plus, Inc. (AMEX: SLP)
1220 W. Avenue J
Lancaster, CA 93534-2902
Phone: (661) 723-7723
FAX: (661) 723-5524
PharmPK Discussion List Archive Index page
Copyright 1995-2010 David W. A. Bourne (email@example.com)