| The accidently discovery of penicillin at 1940 indicated the breakthrough and development of antibiotic molecules. However, the antibiotic drugs used in clinic were bounded to the resistance. Since then, this problem hindered the development of antibiotic drugs. As the only solution to solve this problem was the discovery or design of new molecules. In this paper, the method of quantitative structure-activity relationship(QSAR) was applied to research three different classes of antibiotic molecules: antimicrobial peptides, β-lactams and sulfonamide antibiotic. By the models built by different 2D or 3D QSAR studies, we thereby analyzed the affection of structures on the antimicrobial activities and designed new antibiotics. The results were listed as follows:①The Quantitative structure–activity relationships researches to antimicrobial peptides. To build and develop descriptors was the mainstay of the Quantitative structure–activity relationships research. And until now numerous of sequences representation have been constructed and developed. Some classical descriptors were meaningful for the whole industry of QSAR research including ISA-ECI and z-scales descriptors et al. While in recent years, great numerous amino acids descriptors, ST-scale, T-scale, VHSE, and VSTV et al have been reported. In this work for 12 polypeptides, we used and comparatively estimated 10 different descriptors-FASGAI, NNAAIndex, ST-scale, ISA-ECI, z-scales, MS-WHIM, SZOTT, T-scale, VHSE and VSTV. The variables were selected by the genetic algorithm(GA) and models were generated by multiple linear regression models-the partial least squares(PLS). As shown in the statistic results, PLS models of FASGAI, VHSE, Z-Scales descriptors were qualified. And FASGAI-based QSAR model had the best fitting ability, stabilities and prediction capability, with determination coefficient R2=0.805, cross-validated Q2cv=0.633 and predictive Q2 ext =0.658. Finally, by the analysis to GA-PLS models, we could conclude that: The GA-PLS model generated by the FASGAI representations proved to have the best fitting ability and predictive capability. Analysis to the FASGAI GA-PLS model, we could find out that the factors important for antimicrobial activity were: the bulky property at fifth and eighth amino acids; hydrophobic property at sixth, seventh, eighth and eleventh amino aicds; electronic property at sixth and twelfth amino acids. We applied the I-TASSER to predict the secondary structure of 20 best active antimicrobial peptides in the training set and the secondary structures predicted by I-TASSER show that the peptides tended to present strand structures. In addition, I-TASSER also provided the prediction to the biding sites. For the most active antimicrobial peptides VRLRIRVAVIRK, the prediction to binding sites could be the sixth and seventh amino acids which conformed to results of the predominant and influential position analyzed by the FASGAI-based GA-PLS model.②Topomer Co MFA was employed in the 3D-QSAR research to the β-lactams antibiotics. First, based on the data of 14 the clinic cephalosporins drugs, Topomer Co MFA model was constructed, and the optimal principal components, R2 and Q2 cv of model are 2, 0.792 and 0.607 respectively. While the model obtained suggested that: to enhance the antimicrobial activity, the negative group should be substituted at C-3 near to the core structure, and the small positive molecules be substituted at C-3 remoted to the core structure. For the R2 substitution at C-7, minor and electronegative group could be substituted near to the core structure, and binding with the giant positive group.In addition, a Topomer Co MFA model was constructed based on the data of 23 reported cephalosporins. The optimal principal components, R2, Q2 cv and Q2 ext were 5, 0.973, 0.908 and 0.622 respectively. Therefore the optimal antimicrobial molecules could be designed by the Topomer search. R-groups search to ZINC database was conducted by comparing R-group with activity contribution. By screeing structures by topomer, there were six R2-groups selected. We used them to alternately substitute for the original R2 groups of the best active molecule. As a result, we got a total of 2 new compounds with same best activity. In addition, the mechanism of antimicrobial activity was further researched by the molecular docking. The best active no.8 sample molecule was docked into Penicillin-binding protein 2a from MRSA, and the result indicated that: the biding area appeared to be deep and narrow. Besides the hydrogen bonding interaction of no.8 molecule and PBP2 a, the guanidyl of R-group at C-3 could also enhanced the electrostatic interaction. So the bulky and electropositive groups substituted at C-3 could provide the β-lactams with better antimicrobial activities.③Topomer Co MFA was employed in the 3D-QSAR research to the sulfonamide antibiotics. The data of 20 sulfonamide molecules reported earlier was used to constructed 3D-QSAR model which was applied to predict five molecules as test set. And the optimal principal components, R2 and Q2 cv of this model are 4, 0.793, 0.872 and 0.690 respectively. According to the Topomer Co MFA model, the C-5 at phenyl group of R1 groups could be substituted to or binding with the small electropositive groups to obtain better active molecules. As for the R2-group, the bulky and electropositive groups near to the core structure biding with the electronegative groups on the terminal could efficiently enhance the activity for these molecules. Key words: antimicrobial peptides, β-lactams, Sulfonamide antibiotic, QSAR,... |
PDF Full Text Request