Data processing: artificial intelligence – Knowledge processing system – Creation or modification
Reexamination Certificate
2002-04-16
2004-08-31
Davis, George B. (Department: 2121)
Data processing: artificial intelligence
Knowledge processing system
Creation or modification
C706S045000, C702S183000
Reexamination Certificate
active
06785665
ABSTRACT:
BACKGROUND OF THE INVENTION
The invention relates to the field of Computer Aided Molecular Design (CAMD) and, in particular, to systems and methods those simulate molecular behavior and predict the chemical and physical properties of molecules using computers (Molecular Modeling). Specifically, the invention relates to an expert system that automatically generates molecular mechanics force field parameters for computer simulation of molecular systems.
A molecular mechanics force field is a set of mathematical functions containing adjustable parameters. By adjusting the parameters, a molecular mechanics force field can be made to accurately describe inter-atomic interaction energies as functions of the atomic coordinates for a molecular system. Because all physical properties of a molecular system are determined by the interaction energies, a molecular mechanics force field is a fundamental component in molecular modeling.
Several general molecular mechanics force fields have been developed (e.g., CHARMM, AMBER, MM2, MM3, MMFF, CFF, COMPASS)[1-7]. These force fields perform well for molecular systems that have been explicitly parameterized, but not well for many others that have not been so parameterized because the required parameters are substituted by transferring parameters from similar molecules. Due to the massive number of possible chemical structures that can be made, it is impractical to have one force field that provides a complete coverage for all applications. Frequently, incomplete force field coverage is the stumbling block in “real-life” molecular modeling in various fields such as pharmaceutics, chemicals, and materials.
Generally speaking, two approaches, empirical and ab initio, have been taken to parameterize a force field. One can manually adjust the force field parameters to fit empirical data (for example, bond lengths, angles, conformational energies, vibrational frequencies, crystal structures, lattice energies, etc.) in a trial-and-error manner. Obviously, this approach is limited by the availability of the experimental data and is a very tedious and time-consuming process. Another approach is to derive the parameters from quantum mechanics ab initio data calculated for a group of model compounds. The ab initio approach can be used for any molecules prior to synthesizing the compounds so that it is ideal for applications in designing new molecules (for example, in the drug discovery process).
However, the ab initio parameterization still requires considerable efforts and experiences because the process entails more than simply the use of fitting procedures. It requires several steps consisting of selecting molecular models, preparing data, specifying force field models, setting initial parameters, fitting the data to the models, and validating the fit results. More importantly, the fit process itself is usually a difficult task due to the complexity of the problem. Hundreds of variables may be required in order to fit tens of thousands data points. Usually, simply applying a least-squares method to fit ab initio data leads to poor or unstable results. Empirical controls on the fit process have been extensively used in developing ab initio force fields, but so far, only a small number of experts in this field know how to do the work successfully. For most research scientists or engineers specializing in different fields, it is not a manageable task.
SUMMARY OF THE INVENTION
The present invention recognizes that high quality molecular mechanics force fields are required for successful computer modeling in chemical, pharmaceutical and material industries, and that it is currently a very difficult and time-consuming task to derive force field parameters. This invention remedies these problems by providing a method and a computer implemented expert system that can be used to make accurate force field parameters rapidly and easily from quantum mechanics ab initio data. Various technologies, as described in this specification and in the references identified in the enclosed LIST OF REFERENCES and cross-referenced throughout this document, have been utilized in constructing a novel process that make this system practical.
According to the invention, there is provided a method for developing molecular mechanics force field parameters for computer simulations of molecular systems automatically, comprising the steps of: creating or importing molecular models that represent said molecular systems; searching a database for matches between said molecular models and stored molecular models and retrieving stored parameters; preparing input data for quantum mechanics ab initio calculations; importing calculated data; selecting force field type and functional forms, assigning atom types; estimating initial force field parameters based on said database or a set of mathematical formulas; optimizing said initial force field parameters to fit the input data; validating the optimized force field parameters; and exporting the optimized force field parameters in required formats to external molecular mechanics simulation packages and saving the molecular models, input data and optimize force field parameters to the database.
The present invention provides a computer implemented expert system that implements said method. The system includes a database that contains developed force field parameters and the data utilized for the development, and a procedure that retrieves, stores, and searches said database. According to the invention, a knowledge base is established utilizing said database—future development becomes easier and faster with the amount of data accumulated increases. The system has a mechanism that allows selection of potential energy functional forms and adjustable parameters so that a custom-built force field type can be made to achieve the best performance and accuracy. This is especially useful for dealing with not-so-common molecular systems such as an interface between organic and inorganic molecules. The system provides a procedure that assigns force field attributes and generates default parameters automatically. Coupled with said database, estimated parameters can be made more accurately than those made using other methods. The system contains a robust procedure that utilizes several numerical algorithms and knowledge-based controls to fit very complicated energy surfaces efficiently. In addition, the system is supported by a platform-independent graphic user interface.
REFERENCES:
patent: 5715371 (1998-02-01), Ahamed et al.
patent: 6625569 (2003-09-01), James et al.
Davis George B.
Sun Huai
LandOfFree
Method and expert system of molecular mechanics force fields... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Method and expert system of molecular mechanics force fields..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Method and expert system of molecular mechanics force fields... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3342559