Introduction
LocalSCF is a novel linear scaling quantum mechanics method implemented at present in the semiempirical framework. The computer program named after the method is designed for fast electronic structure calculations of large, complex proteins.
Although the program is very fast the linear scaling algorithm inevitably involves some additional computational overhead to make computations scaling linearly with number of atoms. Therefore on small molecules e.g., 10-50 atoms conventional SCF matrix diagonalization techniques can be faster and thus the use of LocalSCF becomes justified for molecules containing 50 and more atoms.
The nature of the LocalSCF linear scaling algorithm lays in use of localized molecular orbitals. Therefore molecular systems characterized by highly delocalized electronic structure, e.g. fullerenes, are not recommended for treatment with the LocalSCF program.
Features
- Runs on popular PC platforms
- Ultra large 100,000+ atoms protein systems
- Very fast advanced geometry optimization specially tuned for proteins
- Powerful control options for balancing between speed and accuracy
- Linear scaling COSMO solvation model
- Fast Multipole Method for evaluation of Coulomb integrals
- True variational linear scalability
- Semi-empirical Hamiltonians: MNDO, AM1, PM3, and PM5
- Recognition of protein structure from Cartesian coordinates
- Structure quality checking and verification
- Identification of various molecular fragments: amino acid backbone, side-chain, terminal atoms, water molecules and counterions
- Intuitive and easy to use interface for specification of geometry optimization modes by defining particular fragments or amino acid numbers
- Keyword based recognition and on-the-fly optimization of drug molecules in the enzyme cavity
- Assures very low memory requirement, extremely fast calculation of large systems and high accuracy of evaluation of Coulomb interactions
- Provides flexible control over the resource consumption
- Retains high accuracy for short localized molecular orbitals (the shorter LMOs the less RAM is consumed)
- Provides an optimal user-controllable balance between speed and accuracy
- The built-in mechanism for accuracy validation allows comparison of molecular properties in connection with particular keyword options
New Features
- Transition metal complexes are supported through d-orbital extension of general purpose MNDO, AM1, PM3 and PM5 semiempirical Hamiltonians
- All atom quantum-mechanical modeling of enzymes and DNA - transition metal complexes is made available for hundred thousand atoms systems
- Energy calculation, geometry optimization in gas-phase and solvent environment are supported
- Low resource requirement is provided by the use of localized molecular orbitals for the protein part and delocalized orbitals for the transition metal portion of the system
- Computationally efficient initial guess generation for DNA/RNA systems is implemented
- DNA/RNA structure recognition allows reliable identification of structure errors
- New analytical gradients for geometry optimization are implemented
- Support for d-orbitals systems is added
- A considerable speed up is provided for local geometry optimization in solvent environment
- High-throughput screening of thousand compounds libraries of drug candidates is implemented
- QM scoring of all atom protein-ligand complexes is provided for gas phase and solvent environment
- QM re-docking of preliminary docked ligands is available through ligand relaxation in the protein-ligand complexes
- Resource saving QM/QM mode allows freezing molecular orbitals of a portion of the system
- New improved initial guess generator
- More reliable SCF procedure
- Automatic control for memory consumption - orbital tidying
- Restart of a terminated job from last saved topology and density matrix
- PDB input/output filter which allows running calculation from an external regular PDB file and saving results in the PDB format
- A powerful Cartesian coordinate - PDB converter is implemented
- Molecular topology input/output filter allows initial guess generation based on the information imported from the external topology file
- Topology preservation mode allows applying restriction to bond breaking/creation
- Bond order analysis for very large systems is implemented
The relationship between Fujitsu MOPAC and LocalSCF
Both computer programs LocalSCF and MOPAC share the same semiempirical core but have different areas of application that allow better optimization of the program code for particular application. Thus Fujitsu MOPAC is a general computational tool, while LocalSCF is a semiempirical program specially optimized for proteins. The programs are complimentary in the sense that LocalSCF is better for geometry optimization of proteins while MOZYME, as a part of MOPAC, allows faster evaluation of absolute heat of formation of large molecules.
System Requirement
Operating System: Windows 2000/XP, and Linux
CPU: Intel Pentium 4 2.0 GHz or higher.
RAM:1GB or more is recommended for 50,000-atom proteins. There is no hard-coded limit on the number of atoms. However, 2GB memory allocation limit on the 32-bit platforms will restrict the molecule size to approximately 150,000 atoms.
HDD space: 500MB HDD or more, because the HDD requirements will grow with size of the molecule.
|
|
