Multipose Binding in Molecular Docking

Molecular docking has been extensively applied in virtual screening of small molecule libraries for lead identification and optimization. A necessary prerequisite for successful differentiation between active and non-active ligands is the accurate prediction of their binding affinities in the complex by use of docking scoring functions. However, many studies have shown rather poor correlations between docking scores and experimental binding affinities. Our work aimed to improve this correlation by implementing a multipose binding concept in the docking scoring scheme. Multipose binding, i.e., the property of certain protein-ligand complexes to exhibit different ligand binding modes, has been shown to occur in nature for a variety of molecules. We conducted a high-throughput docking study and implemented multipose binding in the scoring procedure by considering multiple docking solutions in binding affinity prediction. In general, improvement of the agreement between docking scores and experimental data was observed, and this was most pronounced in complexes with large and flexible ligands and high binding affinities. Further developments of the selection criteria for docking solutions for each individual complex are still necessary for a general utilization of the multipose binding concept for accurate binding affinity prediction by molecular docking.     

A Cascade Graph Convolutional Network for Predicting Protein–Ligand Binding Affinity

Accurate prediction of binding affinity between protein and ligand is a very important step in the field of drug discovery. Although there are many methods based on different assumptions and rules do exist, prediction performance of protein–ligand binding affinity is not satisfactory so far. This paper proposes a new cascade graph-based convolutional neural network architecture by dealing with non-Euclidean irregular data. We represent the molecule as a graph, and use a simple linear transformation to deal with the sparsity problem of the one-hot encoding of original data. The first stage adopts ARMA graph convolutional neural network to learn the characteristics of atomic space in the protein–ligand complex. In the second stage, one variant of the MPNN graph convolutional neural network is introduced with chemical bond information and interactive atomic features. Finally, the architecture passes through the global add pool and the fully connected layer, and outputs a constant value as the predicted binding affinity. Experiments on the PDBbind v2016 data set showed that our method is better than most of the current methods. Our method is also comparable to the state-of-the-art method on the data set, and is more intuitive and simple.     

Towards Automated Binding Affinity Prediction Using an Iterative Linear Interaction Energy Approach

Binding affinity prediction of potential drugs to target and off-target proteins is an essential asset in drug development. These predictions require the calculation of binding free energies. In such calculations, it is a major challenge to properly account for both the dynamic nature of the protein and the possible variety of ligand-binding orientations, while keeping computational costs tractable. Recently, an iterative Linear Interaction Energy (LIE) approach was introduced, in which results from multiple simulations of a protein-ligand complex are combined into a single binding free energy using a Boltzmann weighting-based scheme. This method was shown to reach experimental accuracy for flexible proteins while retaining the computational efficiency of the general LIE approach. Here, we show that the iterative LIE approach can be used to predict binding affinities in an automated way. A workflow was designed using preselected protein conformations, automated ligand docking and clustering, and a (semi-)automated molecular dynamics simulation setup. We show that using this workflow, binding affinities of aryloxypropanolamines to the malleable Cytochrome P450 2D6 enzyme can be predicted without a priori knowledge of dominant protein-ligand conformations. In addition, we provide an outlook for an approach to assess the quality of the LIE predictions, based on simulation outcomes only.     

CYP 2D6 Binding Affinity Predictions Using Multiple Ligand and Protein Conformations

Because of the large flexibility and malleability of Cytochrome P450 enzymes (CYPs), in silico prediction of CYP binding affinities to drugs and other xenobiotic compounds is a true challenge. In the current work, we use an iterative linear interaction energy (LIE) approach to compute CYP binding affinities from molecular dynamics (MD) simulation. In order to improve sampling of conformational space, we combine results from simulations starting with different relevant protein-ligand geometries. For calculated binding free energies of a set of thiourea compounds binding to the flexible CYP 2D6 isoform, improved correlation with experiment was obtained by combining results ofMDruns starting from distinct protein conformations and ligand-binding orientations. This accuracy was obtained from relatively short MD simulations, which makes our approach computationally attractive for automated calculations of ligand-binding affinities to flexible proteins such as CYPs.     

MHCII3D—Robust Structure Based Prediction of MHC II Binding Peptides

Knowledge of MHC II binding peptides is highly desired in immunological research, particularly in the context of cancer, autoimmune diseases, or allergies. The most successful prediction methods are based on machine learning methods trained on sequences of experimentally characterized binding peptides. Here, we describe a complementary approach called MHCII3D, which is based on structural scaffolds of MHC II-peptide complexes and statistical scoring functions (SSFs). The MHC II alleles reported in the Immuno Polymorphism Database are processed in a dedicated 3D-modeling pipeline providing a set of scaffold complexes for each distinct allotype sequence. Antigen protein sequences are threaded through the scaffolds and evaluated by optimized SSFs. We compared the predictive power of MHCII3D with different sequence-based machine learning methods. The Pearson correlation to experimentally determine IC₅₀ values for MHC II Automated Server Benchmarks data sets from IEDB (Immune Epitope Database) is 0.42, which is in the competitor methods range. We show that MHCII3D is quite robust in leaving one molecule out tests and is therefore not prone to overfitting. Finally, we provide evidence that MHCII3D can complement the current sequence-based methods and help to identify problematic entries in IEDB. Scaffolds and MHCII3D executables can be freely downloaded from our web pages.     

From knowledge-based potentials to combinatorial lead design in silico

Computational methods are becoming increasingly used in the drug discovery process. In this Account, we review a novel computational method for lead discovery. This method, called CombiSMoG for "combinatorial small molecule growth", is based on two components: a fast and accurate knowledge-based scoring function used to predict binding affinities of protein-ligand complexes, and a Monte Carlo combinatorial growth algorithm that generates large numbers of low-free-energy ligands in the binding site of a protein. We illustrate the advantages of the method by describing its application in the design of picomolar inhibitors for human carbonic anhydrase.     

Trastuzumab-Peptide Interactions: Mechanism and Application in Structure-Based Ligand Design

Understanding of protein-ligand interactions and its influences on protein stability is necessary in the research on all biological processes and correlative applications, for instance, the appropriate affinity ligand design for the purification of bio-drugs. In this study, computational methods were applied to identify binding site interaction details between trastuzumab and its natural receptor. Trastuzumab is an approved antibody used in the treatment of human breast cancer for patients whose tumors overexpress the HER2 (human epidermal growth factor receptor 2) protein. However, rational design of affinity ligands to keep the stability of protein during the binding process is still a challenge. Herein, molecular simulations and quantum mechanics were used on protein-ligand interaction analysis and protein ligand design. We analyzed the structure of the HER2-trastuzumab complex by molecular dynamics (MD) simulations. The interaction energies of the mutated peptides indicate that trastuzumab binds to ligand through electrostatic and hydrophobic interactions. Quantitative investigation of interactions shows that electrostatic interactions play the most important role in the binding of the peptide ligand. Prime/MM-GBSA calculations were carried out to predict the binding affinity of the designed peptide ligands. A high binding affinity and specificity peptide ligand is designed rationally with equivalent interaction energy to the wild-type octadecapeptide. The results offer new insights into affinity ligand design.     

Neural network prediction of cetane number and density of diesel fuel from its chemical composition determined by LC and GC-MS

Neural networks were used to correlate and predict the cetane number and the density of diesel fuel from its chemical composition. Cetane number (CN) and density were correlated with 12 hydrocarbon groups in diesel fuel determined by liquid chromatography (LC) and gas chromatography-mass spectrometry (GC-MS). In total, 69 diesel fuels were available for this study: 48 diesel fuels were included in the training data set and 21 in the test data set. Various neural network architectures were trained using the training data set, and the accuracy of the model obtained was examined by using the test data set. For correlating both CN and density in this study, the best neural network architecture was a general regression neural network (GRNN). With the test data set, the mean absolute errors were 1.23 (CN) and 0.002 g/cm³ for the CN and density, respectively. Predictive equations for CN and density of diesel fuel from its chemical composition were also developed with a standard multiple linear regression method. The comparison of the neural network method with the multiple linear regression method, using this data set, revealed that for complex nonlinear problems such as the correlation of the CN with the hydrocarbon type characterization, the neural network approach could provide a better model. However, for a simpler correlation problem like the density of a diesel fuel, which is approximated well by the sum of the contributions of individual components, the predictive equations produced by multiple linear regression and neural network methods gave similar results.     

Towards an integrated description of hydrogen bonding and dehydration: decreasing false positives in virtual screening with the HYDE scoring function

We developed a new empirical scoring function, HYDE, for the evaluation of protein-ligand complexes. HYDE estimates binding free energy based on two terms for dehydration and hydrogen bonding only. The essential feature of this scoring function is the integrated use of log P-derived atomic increments for the prediction of free dehydration energy and hydrogen bonding energy. Taking the dehydration of atoms within the interface into account shows that some atoms contribute favorably to the overall score, while others contribute unfavorably. For instance, hydrogen bond functions are penalized if they are dehydrated unless they can overcompensate this loss by forming a hydrogen bond with excellent geometry. The main stabilizing contribution represents the removal of apolar groups from the water: the hydrophobic effect. Initial studies using the DUD dataset show that with HYDE, there is a significant decrease in false positives, a reasonable categorization of compounds as either non-binders, weak, medium or strong binders, and in particular, there is a generally applicable and thermodynamically sensible cutoff score below which there is a high likelihood that the compound is indeed a binder.     

A novel approach to using neural networks to predict the colour of fibre blends

This work is concerned with the colour prediction of viscose fibre blends, comparing two conventional prediction models (the Stearns–Noechel model and the Friele model) and two neural network models. A total of 333 blended samples were prepared from eight primary colours, including two‐, three‐, and four‐colour mixtures. The performance of the prediction models was evaluated using 60 of the 333 blended samples. The other 273 samples were used to train the neural networks. It was found that the performance of both neural networks exceeded the performance of both conventional prediction models. When the neural networks were trained using the 273 training samples, the average CIELAB colour differences (between measured and predicted colour of blends) for the 60 samples in the test set were close to 1.0 for the neural network models. When the number of training samples was reduced to only 100, the performance of the neural networks degraded, but they still gave lower colour differences between measured and predicted colour than the conventional models. The first neural network was a conventional network similar to that which has been used by several other researchers; the second neural network was a novel application of a standard neural network where, rather than using a single network, a set of small neural networks was used, each of which predicted reflectance at a single wavelength. The single‐wavelength neural network was shown to be more robust than the conventional neural network when the number of training examples was small.     

The Functional Role of Loops and Flanking Sequences of G-Quadruplex Aptamer to the Hemagglutinin of Influenza a Virus

Nucleic acid aptamers are generally accepted as promising elements for the specific and high-affinity binding of various biomolecules. It has been shown for a number of aptamers that the complexes with several related proteins may possess a similar affinity. An outstanding example is the G-quadruplex DNA aptamer RHA0385, which binds to the hemagglutinins of various influenza A virus strains. These hemagglutinins have homologous tertiary structures but moderate-to-low amino acid sequence identities. Here, the experiment was inverted, targeting the same protein using a set of related, parallel G-quadruplexes. The 5′- and 3′-flanking sequences of RHA0385 were truncated to yield parallel G-quadruplex with three propeller loops that were 7, 1, and 1 nucleotides in length. Next, a set of minimal, parallel G-quadruplexes with three single-nucleotide loops was tested. These G-quadruplexes were characterized both structurally and functionally. All parallel G-quadruplexes had affinities for both recombinant hemagglutinin and influenza virions. In summary, the parallel G-quadruplex represents a minimal core structure with functional activity that binds influenza A hemagglutinin. The flanking sequences and loops represent additional features that can be used to modulate the affinity. Thus, the RHA0385–hemagglutinin complex serves as an excellent example of the hypothesis of a core structure that is decorated with additional recognizing elements capable of improving the binding properties of the aptamer.     

Advances and challenges in protein-ligand docking

Molecular docking is a widely-used computational tool for the study of molecular recognition, which aims to predict the binding mode and binding affinity of a complex formed by two or more constituent molecules with known structures. An important type of molecular docking is protein-ligand docking because of its therapeutic applications in modern structure-based drug design. Here, we review the recent advances of protein flexibility, ligand sampling, and scoring functions-the three important aspects in protein-ligand docking. Challenges and possible future directions are discussed in the Conclusion.     

BP神经网络在加氢裂化装置航煤性质软测量中的应用

应用Matlab软件构建单隐层BP神经网络,并对中压加氢裂化装置航煤性质进行软测量应用。以700组样本数据作为训练集,对预测航煤闪点、终馏点模型进行训练。结果表明,在152组验证数据集上模型对闪点、终馏点预测分别取得1.57 ℃和2.74 ℃的均方误差(RMSE),随之在80组测试数据集上模型取得的泛化RMSE分别为1.87 ℃和1.98 ℃。以300组样本数据作为训练集,对预测航煤密度的模型进行训练。结果表明,在100组验证集上模型RMSE为2.18 kg·m^-3,随之在70组测试数据集上的泛化RMSE为2.72 kg·m^-3。BP神经网络的泛化RMSE表明,通过合理选择特征变量和设计网络结构,单隐层BP神经网络能够满足航煤性质的工业软测量要求。     

Improved Artificial Neural Network for Data Analysis and Property Prediction in Slag Glass–Ceramic

The development of slag glass–ceramics has environmental and commercial value. However, new types of these materials are usually developed using the "trial and error" method because of little understanding of the relationship between the composition, processing, microstructure, and properties. In this paper, artificial neural network (ANN) technology was applied to investigate the relationship between the composition content and the properties of slag glass–ceramic. The investigation showed that the ANN model had an outstanding learning ability and was effective in complex data analysis. If the data set reflects the relationship of the composition and property, the trained network will learn the relationship and then give relatively accurate and stable prediction. A new "virtual sample" technology has also been created which improves the prediction performance of the network by providing greater accuracy and reliability. With this virtual sample technology, the ANN model can establish the exact relationship from a small-size-data set, and gives accurate predictions. This improved ANN model is a powerful and reliable tool for data analysis and property prediction, and will facilitate the material design and development of slag glass–ceramics.     

Inferential Estimation of Polymer Melt Index Using Sequentially Trained Bootstrap Aggregated Neural Networks

The inferential estimation of a polymer melt index in an industrial polymerization process using aggregated neural networks is presented in this paper. The difficult‐to‐measure polymer melt index is estimated from easy‐to‐measure process variables, and their relationship is estimated using aggregated neural networks. The individual networks are trained on bootstrap re‐samples of the original training data by a sequential training algorithm. In this training method, individual networks, within a bootstrap aggregated neural network model, are trained sequentially. The first network is trained to minimize its prediction error on the training data. In the training of subsequent networks, the training objective is not only to minimize the individual networks' prediction errors but also to minimize the correlation among the individual networks. Training is terminated when the aggregated network prediction performance on the training and testing data cannot be further improved. Application to real industrial data demonstrates that the polymer melt index can be successfully estimated using an aggregated neural network.     

Automated Selection of Gate Location Based on Desired Quality of Injection-Molded Part

The objective is to develop a methodology that automatically predicts the “optimal” gate location(s) of injection molds based on injection-molding simulation. User-defined design evaluating criteria for three important parameters–-warpage, weld and meld lines in a constrained area, and Izod impact strength at the specific regions of the injection-molded part–-are introduced to determine the optimal gate location. Among the three parameters, the Izod impact strength is obtained using a previously trained neural network. The difficulty in predicting accurate values of engineering property like Izod impact strength is that they vary throughout a part with respect to the thermomechanical history. Upon evaluating each gate location, the trained neural network computation predicts, regardless of part geometry, Izod impact strength by a nonparameteric modeling of the complex relation with thermomechanical processing histories. The methodology comprises a two-stage process: (1) choosing the best among a set of gate locations generated based on a human designer's intuition, and (2) locally searching for the better gate location.     

Biomolecular Simulations with the Three-Dimensional Reference Interaction Site Model with the Kovalenko-Hirata Closure Molecular Solvation Theory

The statistical mechanics-based 3-dimensional reference interaction site model with the Kovalenko-Hirata closure (3D-RISM-KH) molecular solvation theory has proven to be an essential part of a multiscale modeling framework, covering a vast region of molecular simulation techniques. The successful application ranges from the small molecule solvation energy to the bulk phase behavior of polymers, macromolecules, etc. The 3D-RISM-KH successfully predicts and explains the molecular mechanisms of self-assembly and aggregation of proteins and peptides related to neurodegeneration, protein-ligand binding, and structure-function related solvation properties. Upon coupling the 3D-RISM-KH theory with a novel multiple time-step molecular dynamic (MD) of the solute biomolecule stabilized by the optimized isokinetic Nosé–Hoover chain thermostat driven by effective solvation forces obtained from 3D-RISM-KH and extrapolated forward by generalized solvation force extrapolation (GSFE), gigantic outer time-steps up to picoseconds to accurately calculate equilibrium properties were obtained in this new quasidynamics protocol. The multiscale OIN/GSFE/3D-RISM-KH algorithm was implemented in the Amber package and well documented for fully flexible model of alanine dipeptide, miniprotein 1L2Y, and protein G in aqueous solution, with a solvent sampling rate ~150 times faster than a standard MD simulation in explicit water. Further acceleration in computation can be achieved by modifying the extent of solvation layers considered in the calculation, as well as by modifying existing closure relations. This enhanced simulation technique has proven applications in protein-ligand binding energy calculations, ligand/solvent binding site prediction, molecular solvation energy calculations, etc. Applications of the RISM-KH theory in molecular simulation are discussed in this work.