Reduced point charge models of amino acids are used to model Ubiquitin （PDB： 1UBQ）. They are designed （i） from local ex- tremum positions in charge density （CD） distribution functions built from the Poisson equ...Reduced point charge models of amino acids are used to model Ubiquitin （PDB： 1UBQ）. They are designed （i） from local ex- tremum positions in charge density （CD） distribution functions built from the Poisson equation applied to smoothed molecular electrostatic potential functions, or （ii） from local maximum positions in promolecular electron density distribution （ED） func- tions. Charge values are fitted versus all-atom Amber99 molecular electrostatic potentials. The program GROMACS is used to generate molecular dynamics trajectories of the protein, under various implementation schemes, solvation, and temperature conditions. Point charges that are not located on atoms are considered as virtual sites with a null mass and radius. The results illustrate that secondary structure is best preserved with the CD-based model at low temperatures and in vacuum. This indi- cates that local potential energy wells are consistent with the all-atom model. However, at room temperature, the structure is best conserved when point charges are forced to be located on atoms, due to a better description of the Coulomb l-4 energy terms. The ED-based model, generated at a lower resolution, led to the largest discrepancies versus the all-atom case. The CD-based model allows the formation of protein-water H-bonds with geometrical properties similar to the all-atom ones. Con- trarily, intra-molecular H-bonds are not well described. Structural, thermodynamical, and dynamical properties of proteins modelled with reduced point charge models are also significantly affected by the choice of the solvent force field.展开更多
The recent years have witnessed a surge of interests in graph-based semi-supervised learning(GBSSL).In this paper,we will introduce a series of works done by our group on this topic including:1)a method called linear ...The recent years have witnessed a surge of interests in graph-based semi-supervised learning(GBSSL).In this paper,we will introduce a series of works done by our group on this topic including:1)a method called linear neighborhood propagation(LNP)which can automatically construct the optimal graph;2)a novel multilevel scheme to make our algorithm scalable for large data sets;3)a generalized point charge scheme for GBSSL;4)a multilabel GBSSL method by solving a Sylvester equation;5)an information fusion framework for GBSSL;and 6)an application of GBSSL on fMRI image segmentation.展开更多
文摘Reduced point charge models of amino acids are used to model Ubiquitin （PDB： 1UBQ）. They are designed （i） from local ex- tremum positions in charge density （CD） distribution functions built from the Poisson equation applied to smoothed molecular electrostatic potential functions, or （ii） from local maximum positions in promolecular electron density distribution （ED） func- tions. Charge values are fitted versus all-atom Amber99 molecular electrostatic potentials. The program GROMACS is used to generate molecular dynamics trajectories of the protein, under various implementation schemes, solvation, and temperature conditions. Point charges that are not located on atoms are considered as virtual sites with a null mass and radius. The results illustrate that secondary structure is best preserved with the CD-based model at low temperatures and in vacuum. This indi- cates that local potential energy wells are consistent with the all-atom model. However, at room temperature, the structure is best conserved when point charges are forced to be located on atoms, due to a better description of the Coulomb l-4 energy terms. The ED-based model, generated at a lower resolution, led to the largest discrepancies versus the all-atom case. The CD-based model allows the formation of protein-water H-bonds with geometrical properties similar to the all-atom ones. Con- trarily, intra-molecular H-bonds are not well described. Structural, thermodynamical, and dynamical properties of proteins modelled with reduced point charge models are also significantly affected by the choice of the solvent force field.
基金supported by the National Natural Science Foundation of China(Grant Nos.60835002,61075004).
文摘The recent years have witnessed a surge of interests in graph-based semi-supervised learning(GBSSL).In this paper,we will introduce a series of works done by our group on this topic including:1)a method called linear neighborhood propagation(LNP)which can automatically construct the optimal graph;2)a novel multilevel scheme to make our algorithm scalable for large data sets;3)a generalized point charge scheme for GBSSL;4)a multilabel GBSSL method by solving a Sylvester equation;5)an information fusion framework for GBSSL;and 6)an application of GBSSL on fMRI image segmentation.
