Peptide Drugs QSAR Study Based On The Molecular Structural Characterization Of Primary Structure Of Peptide

The study of quantitative structure activity relationship(QSAR) of peptide drug molecules aims to create a relationship between the biological sciences and chemistry, it is a correlations study between character of drug molecule structure and its biological activity via the theories and methods of a combination of mathematics and chemistry. To use QSAR method in new medicine research provide a new theoretical basis for predicting unknown drug and drug design. Study of quantitative structure activity relationship in the long run of pharmaceutical researchis a epoch-making. In the QSAR researching, modeling in structure information reflected from a series of feature parameter and biological activity is a essential links.This paper starts with the structure parameter in QSAR. Four novel structural features parameters of amino acid are proposed by using different calculation method of descriptors with three-dimensional structure of 20 natural amino acids: SVWGM、SVICE、SVGT and 3D-Ho VAIF. The novel amino acids descriptors presented in this thesis are used to characterize some typical polypeptide drugs. Here some quantitative structure activity relationship(QSAR) models were built by partial least square regression(PLS) or multiple linear regression(MLR) with model validation and preliminary studies, and make sure that proposed amino acids descriptors have practical application value. The research contents of this paper includes as follows:The first part is using descriptor computing software, starting from spatial three-dimensional of molecules of natural amino acids. Extracts three 3D structure parameter: Whim descriptor 、 Getaway descriptor and Mo RSE descriptor. A new descriptor-SVWGM was extracted from principle component analysis in feature extraction. Applying SVWGM to angiotensin-converting enzyme(ACE) dipeptide, oxytocin analogues(OT) and bitter tasting dipeptides(BTT), then QSAR models were built by partial least square regression combining stepwise linear regression. Satisfactory result can be obtained by comprehensive analyzing of fitting ability、stability、predictive ability of the models, and it shows SVWGM covering nearly all the information of Whim descriptor、Getaway descriptor and Mo RSE descriptor. What is more, to build effectual simulation models in the structural information and bioactivity of the polypeptide medicine, SVWGM is proven to be a new structural parameters which is valuable to be popularized.In the second part of this dissertation, extracts three 2D structure parameter: Information indices、Connectivity indices、Eigenvalue-based indices from spatial three-dimensional of molecules of natural amino acids. A new descriptor-SVICE was extracted from principle component analysis in feature extraction. Applying SVWGM to angiotensin-converting enzyme(ACE) tripeptide, bitter tasting dipeptides(BTT) and antimicrobial peptides(AMP), then QSAR models were built by multiple linear regression combining stepwise linear regression. Also using related statistics parameter and biological activity prediction to analyze the fitting ability、stability、predictive ability of the models and the results are very promising. This study result shows that SVICE have powerful abilities of characterizing peptide drugs. This descriptor is valuable and helpful for further researching and development of novel type of peptide drugs.The third part is employing a new descriptor-SVGT, it contains Topological indices information, a 2D descriptor, and Geometrical information, a 3D descriptor. Applying SVGT to AMP, OT and ACE dipeptide, then QSAR models were built by multiple linear regression combining stepwise linear regression. From the results of the model, it can actually be seen SVGT is able to characterize structural information of peptide molecular medicine. The cause may be that it contains two dimensions structural information. Its proposal undoubtedly provides the theoretical basis improvement and innovation of medicines.The fourth section is used three-dimensional holographic vector of atomic interaction field(3D-Ho VAIF),which contains three common non-bonding interactions, that is, electrostatic interaction, steric interaction and hydrophobic interaction related with atomic relative distance and atomic self-properties to characterize molecular structure. 3D-Ho VAIF is separately injected into BTT, OT and ACE dipeptide to establish QSAR model. The new analysis shows that models developed have higher fitting capacity and prediction accuracy, that is to say, 3D-Ho VAIF is a great structure descriptor, which offer a new way of thinking and references of drug research.