Vascular endothelial growth factor (VEGF), along with its receptor tyrosine kinases VEGFR-2 or kinase insert domain receptor (KDR), are targets for development of novel anticancer agents. Accurately predicting the structural characteristics of the target and chemical features of ligands can greatly reduce the cost and shorten the cycle of designing selective KDR inhibitors with desired activity. In this study, a docking strategy and three dimensional holographic vector of atomic interaction field (3D-HoVAIF) were applied in QSAR analysis of KDR inhibitors. The optimal model was constructed by using stepwise regression combined with partial least squares regression (SMR-PLS). Integrating the results of QSAR analysis, ADMET, pharmacophore modeling and a reverse screening strategy, eight derivatives were identified as potential KDR inhibitors. Then molecular dynamics (MD) simulations and free energy calculations were employed to explore the detailed binding process, so as to compare the potential binding modes of inhibitors with different activities. By analyzing the key residues in the binding site, it was found that different KDR–ligand complexes had similar binding modes. The predicted binding affinities were highly correlated with the experimental biological activity. Free energy analysis indicated that van der Waals interactions provided the major driving force for the binding process. Furthermore, key residues, such as Leu840, Val848, Ala866, Lys868, Leu889, Val899, Thr916, Phe918, Cys919, Leu1035, Cys1045, Asp1046, and Phe1047 played a vital role in forming hydrogen bonds, salt bridges, and hydrophobic interactions with the conformation of KDR. The above results will help design more efficient KDR inhibitors.

•We developed CoMFA, CoMSIA models for a series of 53 3-Pyridine heterocyclic derivatives.•Molecular docking method was used to analyses possible interactions between receptors and the compounds.•The results of models based ligand alignment and receptor alignment support each other.•Ten derivatives as potential candidates of PI3K/mTOR inhibitors with good predicted activities were designed.Phosphoinosmde-3-kinase/ mammalian target of rapamycin (PI3K/mTOR) dual inhibitors have attracted a great deal of interest as antitumor drugs research. In order to design and optimize these dual inhibitors, two types of 3D-quantitative structure–activity relationship (3D-QSAR) studies based on the ligand alignment and receptor alignment were applied using the comparative molecular field analysis (CoMFA) and comparative molecular similarity indices analysis (CoMSIA). In the study based on ligands alignment, models of PI3K (CoMFA with r2, 0.770; q2, 0.622; CoMSIA with r2, 0.945; q2, 0.748) and mTOR (CoMFA with r2, 0.850; q2, 0.654; CoMSIA with r2, 0.983; q2, 0.676) have good predictability. And in the study based on receptor alignment, models of PI3K (CoMFA with r2, 0.745; q2, 0.538; CoMSIA with r2, 0.938; q2, 0.630) and mTOR (CoMFA with r2, 0.977; q2, 0.825; CoMSIA with r2, 0.985; q2, 0.728) also have good predictability. 3D contour maps and docking results suggested different groups on the core parts of the compounds could enhance the biological activities. Finally, ten derivatives as potential candidates of PI3K/mTOR inhibitors with good predicted activities were designed.

•We developed QSAR models for a series of 41 azacycles CCR5 antagonists using CoMFA, CoMSIA and Topomer CoMFA methods.•Molecular docking method was used to analyses possible interactions between CCR5 and azacycles derivatives.•The results of molecular docking and 3D-QSAR studies supported one another.•A set of novel derivatives with predicted activities were designed.The beta chemokine receptor 5 (CCR5) is an attractive target for pharmaceutical industry in the HIV-1, inflammation and cancer therapeutic areas. In this study, we have developed quantitative structure activity relationship (QSAR) models for a series of 41 azacycles CCR5 antagonists using comparative molecular field analysis (CoMFA), comparative molecular similarity indices analysis (CoMSIA), and Topomer CoMFA methods. The cross-validated coefficient q2 values of 3D-QASR (CoMFA, CoMSIA, and Topomer CoMFA) methods were 0.630, 0.758, and 0.852, respectively, the non-cross-validated R2 values were 0.979, 0.978, and 0.990, respectively. Docking studies were also employed to determine the most probable binding mode. 3D contour maps and docking results suggested that bulky groups and electron-withdrawing groups on the core part would decrease antiviral activity. Furthermore, docking results indicated that H-bonds and π bonds were favorable for antiviral activities. Finally, a set of novel derivatives with predicted activities were designed.

The accurate identification of cytotoxic T lymphocyte epitopes is becoming increasingly important in peptide vaccine design. The ubiquitin–proteasome system plays a key role in processing and presenting major histocompatibility complex class I restricted epitopes by degrading the antigenic protein. To enhance the specificity and efficiency of epitope prediction and identification, the recognition mode between the ubiquitin–proteasome complex and the protein antigen must be considered. Hence, a model that accurately predicts proteasomal cleavage must be established. This study proposes a new set of parameters to characterize the cleavage window and uses a backpropagation neural network algorithm to build a model that accurately predicts proteasomal cleavage. The accuracy of the prediction model, which depends on the window sizes of the cleavage, reaches 95.454 % for the N-terminus and 95.011 % for the C-terminus. The results show that the identification of proteasomal cleavage sites depends on the sequence next to it and that the prediction performance of the C-terminus is better than that of the N-terminus on average. Thus, models based on the properties of amino acids can be highly reliable and reflect the structural features of interactions between proteasomes and peptide sequences.

Phenols and its analogues are known to induce caspase-mediated apoptosis activity and cytotoxicity on various cancer cell lines. In the current work, two types of molecular field analysis techniques were used to perform the three dimension quantitative structure activity relationship (3D-QSAR) modeling between structural characters and anticancer activity of two sets of phenolic compounds, which are comparative molecular field analysis (CoMFA) and comparative molecular similarity indices analysis (CoMSIA). Then two 3D-QSAR models for two sets of phenolic analogues were obtained with good results. The first QSAR model, which was derived from CoMFA for phenols with caspase-mediated apoptosis activity against L1210 cells, had good predictability (q2 = 0.874, r2 = 0.930), and the other one was derived from CoMSIA for electron-attracting phenols with cytotoxicity in L1210 cell (q2 = 0.836, r2 = 0.950). In addition, the CoMFA and CoMSIA contour maps provide valuable guidance for designing highly active phenolic compounds.

A novel set of descriptors G-scale was derived from 457 physicochemical properties of the natural amino acids. The descriptors were then applied to study on quantitative structure-activity relationships (QSARs) of nine peptide datasets of angiotensin-converting enzyme inhibitor (ACE-inhibitor) oligopeptides (between dipeptides and decapeptides) by using partial least square (PLS) regression. The multiple correlation coefficients (R2) and leave one out cross validation values (Q2) of PLS models are better than or close to the results of references. The results show that the descriptors proposed here may be a useful structural expression method, and they may be hopefully used in biological activity study of ACE-inhibitor oligopeptides.

Protein–protein interaction plays a critical role in signal transduction and many other key biological processes. The present study evaluated four parameters selected from among 554 physiochemical variables of 20 natural amino acids listed in AAindex, namely, hydrophobicity, electronic properties, steric properties, and hydrogen-bond properties. Human amphiphysin-1 Src homology 3 (SH3) domain-binding decapeptides were the object of analysis. A quantitative structure–activity relationship model of the SH3 domain-binding peptides was constructed using multivariate linear regression. The results showed that the four parameters ably characterize the structure of SH3 domain-binding decapeptides, have definitive physicochemical properties and a low level of computational complexity, are accessible, and may be used in integrated prediction models for other protein–peptide interactions.

The potencies of natural adipokinetic hormones and synthetic variants (short peptides) have been obtained in Locusta migratoria. This short peptides (a total of sixty-nine analogues) data was used to construct the mathematical models of the hormone potencies. The sequence variations of amino acids in both natural and artificial adipokinetic hormone analogues were characterized by using a set of descriptors proposed in our laboratory, then QSAR models were developed successfully by using orthogonal signal correction combines with partial least squares (OSC-PLS) method. The cross validation correlation coefficients (Q2) were up to 0.942. The results show that the descriptors proposed in this study will be useful in structure characterization and activity prediction of biological molecules.