Study And Application Of Quantitative Structure-Spectrum Relationship

Serving as a significant calculating method and common technique for drug design, quantitative structure-activity relationship (QSAR) plays an important role in preparing and developing new drugs. In half a century, QSAR study has greatly improved developments of many subjects as organic synthesis chemistry, medicinal chemistry and drug design etc., being potent to investigate molecular physicochemical properties and bioactivities and to seek reasonable interpretations. While quantitative structure-spectrum relationship (QSSR) is referred to the process that spectrum data derived from instrumental analysis are theoretically simulated by QSAR approaches. However, spectrum data, due to their own complexity and diversity, are not merely in linear relation with structures, thus being difficult to be correctly predicted and simulated. In this context, a helpful discussion has been attempted, i.e. several kinds of spectrum behaviors of organic compounds and biomolecules are deeply researched by some new molecular structural characterization (MSC) methods. Major works consist of the following.â‘ Starting with molecular two-dimensinal structures, atomic electronegativity interaction vector(AEIV) and atomic hybridation state index (AHSI), which indicate atomic microscopic environment in molecules and atomic hybridation states, are resulted from interatomic interaction manners and hybrid states. Applying these descriptors to characterize a great deal of equivalent resonant carbon atoms of 48 pyran and 25 furan glycoses, multiple linear regreesion (MLR) model is constructed to simulate nuclear magnetic resonance (NMR) chemical shifts of 13C atoms for these glycoses, with correlation coefficient r and cross validation q both above 0.9. Then by strict statistical diagnosis, the model is confirmed to be stable and predictable.â‘¡36 atomic fragment types of organic compound have been defined, and the multilevel atom-pair frequency matrix has been constructed according to the occurrence number in pairs of atomic fragments with different bond lengths in the molecule. On the basis of them, a novel molecular coding technique: atom-pair holographic code (APH), is obtained. To some extent, this method exhibits a large number of benefits at the same time. For example, it can calculate 2D molecular topological descriptor easily, operate without difficulty and possess definite physicochemical meaning of 3D molecular structural characterization methods, and may fetch the complicated information of molecule, etc. Therefore, it is appropriate for the study on quantitative structure-retention relationship (QSRR) of medicines and biological molecules. We attempt in this paper to utilize the method of APH to the quantitative prediction of reversed-phase liquid chromatogram (RPLC) retention data of 33 purine derivatives and 24 steroids. The fitting multiple correlation coefficient r², cross-validated multiple correlation coefficient q² and predicted ability r²_pred over test set's samples of obtained partial least-square (PLS) regression model are respectively 0.990, 0.893 and 0.977, 0.897, 0.941.â‘¢The concept of facing to consumer together with the idea of self-regulation is introduced into molecular structural representation. Upon molecular topological structure and branch-connecting type, a novel molecular representing method generalized correlative index (GCI) arises from such definitions as generalized correlative function (GCF), property correlative parameters (PCP) and distance-relational function (DRF). Applying GCI into retention behavior studies on several molecules, e.g. 115 polychlorinated dibenzofurans, 41 polychlorinated dibenzo-p-dioxins, 62 polychlorinated naphthalenes and 210 polychlorinated biphenyls, the resulting QSRR models both have their correlative coefficients r and cross-validation correlative coefficients q above 0.98. GCI is thus deemed to be superior in molecular structural representation and adaptable to diverse kinds of molecular properties.â‘£Collected from reference reports, a large-scale ion mobility spectrometry collision cross section database comprising 819 samples has been performed quantitative structure-spectrometry relationship (QSSR) studies by the APH. Deeply testing modeling stabilities and generalized abilities by both internal and external exams, it is confirmed that APH is in outstanding relation with peptide collision cross sections, while involving in partially nonlinear factors for few polypeptides. The model is deemed to assist in quantitative computer-aided predictions for peptide collision cross sections.⑤A novel electrotopological descriptor, molecular electronegativity-interaction vector (MEIV), has been developed in this paper to characterize the structure of 420 singly protonated peptides. In order to find the link between structure and collision cross section of peptide ion, three quantitative structure-property relationship (QSPR) models were built with excellent fitness and were further proved to be of stability and predictability by applying both internal and external validation. The results show that MEIV correlates well with collision cross section, mainly linear and somewhat nonlinear relationship.â‘¥Considering that ion mobility spectrometry (IMS) applications in future may mainly focus on intricate drug/biological systems, a novel molecular structural characterization method referring to molecular graphic fingerprint (MoGF) is proposed in this paper. By the way of mapping intrinsic interatomic correlations into a certain coordinates system according to a certain reasonable sampling interval by radical distribution function, a characteristic graph curve is obtained to carry on abundant information about molecular structural characteristics, possessing of great merits in easy calculation, independent of experiments, rich information contents, explicit structural meanings and intuitive expressions, etc. In an attempt to apply MoGF into QSPR studies for 579 singly-protonated peptide collision cross sections, the constructed partial least square (PLS) regression model are subject to rigorous double internal and external validations, indicating the model is robust and predictable, with statistics on both training and test set as r²=0.991, q²=0.990, RMSEE=5.526, RMSCV=5.572, q_ext²=0.990, r_ext²=0.990, r_0,ext²=0.990, r_0,ext^'2=0.990, k=1.003, k′=0.996 and RMSEP=5.561, respectively.