Cancer is one of the leading causes of death worldwide. Conventional cancer treatment relies on radiotherapy and chemotherapy, but both methods bring severe side effects to patients, as these therapies not only attack cancer cells but also damage normal cells. Anticancer peptides (ACPs) are a promising alternative as therapeutic agents that are efficient and selective against tumor cells. Here, we propose a deep learning method based on convolutional neural networks to predict biological activity (EC50, LC50, IC50, and LD50) against six tumor cells, including breast, colon, cervix, lung, skin, and prostate. We show that models derived with multitask learning achieve better performance than conventional single-task models. In repeated 5-fold cross validation using the CancerPPD data set, the best models with the applicability domain defined obtain an average mean squared error of 0.1758, Pearson's correlation coefficient of 0.8086, and Kendall's correlation coefficient of 0.6156. As a step toward model interpretability, we infer the contribution of each residue in the sequence to the predicted activity by means of feature importance weights derived from the convolutional layers of the model. The present method, referred to as xDeep-AcPEP, will help to identify effective ACPs in rational peptide design for therapeutic purposes. The data, script files for reproducing the experiments, and the final prediction models can be downloaded from http://github.com/chen709847237/xDeep-AcPEP. The web server to directly access this prediction method is at https://app.cbbio.online/acpep/home.
Lipoteichoic acid synthase (LtaS) is a key enzyme for the cell wall biosynthesis of Gram-positive bacteria. Gram-positive bacteria that lack lipoteichoic acid (LTA) exhibit impaired cell division and growth defects. Thus, LtaS appears to be an attractive antimicrobial target. The pharmacology around LtaS remains largely unexplored with only two small-molecule LtaS inhibitors reported, namely "compound 1771" and the Congo red dye. Structure-based drug discovery efforts against LtaS remain unattempted due to the lack of an inhibitor-bound structure of LtaS. To address this, we combined the use of a molecular docking technique with molecular dynamics (MD) simulations to model a plausible binding mode of compound 1771 to the extracellular catalytic domain of LtaS (eLtaS). The model was validated using alanine mutagenesis studies combined with isothermal titration calorimetry. Additionally, lead optimization driven by our computational model resulted in an improved version of compound 1771, namely, compound 4 which showed greater affinity for binding to eLtaS than compound 1771 in biophysical assays. Compound 4 reduced LTA production in S. aureus dose-dependently, induced aberrant morphology as seen for LTA-deficient bacteria, and significantly reduced bacteria titers in the lung of mice infected with S. aureus. Analysis of our MD simulation trajectories revealed the possible formation of a transient cryptic pocket in eLtaS. Virtual screening (VS) against the cryptic pocket led to the identification of a new class of inhibitors that could potentiate β-lactams against methicillin-resistant S. aureus. Our overall workflow and data should encourage further drug design campaign against LtaS. Finally, our work reinforces the importance of considering protein conformational flexibility to a successful VS endeavor.
ProBiS is a new method to identify the binding site of protein through local structural alignment against the nonredundant Protein Data Bank (PDB), which may result in unique findings compared to the energy-based, geometry-based, and sequence-based predictors. In this work, binding sites of Hemagglutinin (HA), which is an important target for drugs and vaccines in influenza treatment, have been revisited by ProBiS. For the first time, the identification of conserved binding sites by local structural alignment across all subtypes and strains of HA available in PDB is presented. ProBiS finds three distinctive conserved sites on HA's structure (named Site 1, Site 2, and Site 3). Compared to other predictors, ProBiS is the only one that accurately defines the receptor binding site (Site 1). Apart from that, Site 2, which is located slightly above the TBHQ binding site, is proposed as a potential novel conserved target for membrane fusion inhibitor. Lastly, Site 3, located around Helix A at the stem domain and recently targeted by cross-reactive antibodies, is predicted to be conserved in the latest H7N9 China 2013 strain as well. The further exploration of these three sites provides valuable insight in optimizing the influenza drug and vaccine development.
Many of the conventional similarity methods assume that molecular fragments that do not relate to biological activity carry the same weight as the important ones. One possible approach to this problem is to use the Bayesian inference network (BIN), which models molecules and reference structures as probabilistic inference networks. The relationships between molecules and reference structures in the Bayesian network are encoded using a set of conditional probability distributions, which can be estimated by the fragment weighting function, a function of the frequencies of the fragments in the molecule or the reference structure as well as throughout the collection. The weighting function combines one or more fragment weighting schemes. In this paper, we have investigated five different weighting functions and present a new fragment weighting scheme. Later on, these functions were modified to combine the new weighting scheme. Simulated virtual screening experiments with the MDL Drug Data Report (23) and maximum unbiased validation data sets show that the use of new weighting scheme can provide significantly more effective screening when compared with the use of current weighting schemes.
Erythromycin A and roxithromycin are clinically important macrolide antibiotics that selectively act on the bacterial 50S large ribosomal subunit to inhibit bacteria's protein elongation process by blocking the exit tunnel for the nascent peptide away from ribosome. The detailed molecular mechanism of macrolide binding is yet to be elucidated as it is currently known to the most general idea only. In this study, molecular dynamics (MD) simulation was employed to study their interaction at the molecular level, and the binding free energies for both systems were calculated using the molecular mechanics Poisson-Boltzmann surface area (MM-PBSA) method. The calculated binding free energies for both systems were slightly overestimated compared to the experimental values, but individual energy terms enabled better understanding in the binding for both systems. Decomposition of results into residue basis was able to show the contribution of each residue at the binding pocket toward the binding affinity of macrolides and hence identified several key interacting residues that were in agreement with previous experimental and computational data. Results also indicated the contributions from van der Waals are more important and significant than electrostatic contribution in the binding of macrolides to the binding pocket. The findings from this study are expected to contribute to the understanding of a detailed mechanism of action in a quantitative matter and thus assisting in the development of a safer macrolide antibiotic.
T1 lipase is isolated from the palm Geobacillus zalihae strain T1 in Malaysia, functioning as a secreted protein responsible for the catalyzing hydrolysis of long-chain triglycerides into fatty acids and glycerol at high temperatures. In the current study, using 30 ns molecular dynamics simulations at different temperatures, an aqueous activation was detected for T1 lipase. This aqueous activation in T1 lipase was mainly caused by a double-flap movement mechanism. The double flaps were constituted by the hydrophobic helices 6 and 9. Helix 6 employed two major components with the hydrophilic part at the surface and the hydrophobic part inside. In the aqueous solution, the hydrophobic part could provide enough power for helix 6 to move away, driving the protein into an open configuration and exposing the catalytic triad. Our findings could provide structural evidence to support the double-flap movement, revealing the catalytic mechanism for T1 lipase.
Molecular dynamics (MD) simulations of membrane-embedded G-protein coupled receptors (GPCRs) have rapidly gained popularity among the molecular simulation community in recent years, a trend which has an obvious link to the tremendous pharmaceutical importance of this group of receptors and the increasing availability of crystal structures. In view of the widespread use of this technique, it is of fundamental importance to ensure the reliability and robustness of the methodologies so they yield valid results and enable sufficiently accurate predictions to be made. In this work, 200 ns simulations of the A2a adenosine receptor (A2a AR) have been produced and evaluated in the light of these requirements. The conformational dynamics of the target protein, as obtained from replicate simulations in both the presence and absence of an inverse agonist ligand (ZM241385), have been investigated and compared using principal component analysis (PCA). Results show that, on this time scale, convergence of the replicates is not readily evident and dependent on the types of the protein motions considered. Thus rates of inter- as opposed to intrahelical relaxation and sampling can be different. When studied individually, we find that helices III and IV have noticeably greater stability than helices I, II, V, VI, and VII in the apo form. The addition of the inverse agonist ligand greatly improves the stability of all helices.
Analysis of 300 ns (ns) molecular dynamics (MD) simulations of an adenosine A2a receptor (A2a AR) model, conducted in triplicate, in 1-palmitoyl-2-oleoylphosphatidylcholine (POPC) and 1-palmitoyl-2-oleoylphosphatidylethanolamine (POPE) bilayers reveals significantly different protein dynamical behavior. Principal component analysis (PCA) shows that the dissimilarities stem from interhelical rather than intrahelical motions. The difference in the hydrophobic thicknesses of these simulated lipid bilayers is potentially a significant reason for the observed difference in results. The distinct lipid headgroups might also lead to different molecular interactions and hence different protein loop motions. Overall, the A2a AR shows higher mobility and flexibility in POPC as compared to POPE.
A Bayesian inference network (BIN) provides an interesting alternative to existing tools for similarity-based virtual screening. The BIN is particularly effective when the active molecules being sought have a high degree of structural homogeneity but has been found to perform less well with structurally heterogeneous sets of actives. In this paper, we introduce an alternative network model, called a Bayesian belief network (BBN), that seeks to overcome this limitation of the BIN approach. Simulated virtual screening experiments with the MDDR, WOMBAT and MUV data sets show that the BIN and BBN methods allow effective screening searches to be carried out. However, the results obtained are not obviously superior to those obtained using a much simpler approach that is based on the use of the Tanimoto coefficient and of the square roots of fragment occurrence frequencies.
Ensemble docking can be a successful virtual screening technique that addresses the innate conformational heterogeneity of macromolecular drug targets. Yet, lacking a method to identify a subset of conformational states that effectively segregates active and inactive small molecules, ensemble docking may result in the recommendation of a large number of false positives. Here, three knowledge-based methods that construct structural ensembles for virtual screening are presented. Each method selects ensembles by optimizing an objective function calculated using the receiver operating characteristic (ROC) curve: either the area under the ROC curve (AUC) or a ROC enrichment factor (EF). As the number of receptor conformations, N, becomes large, the methods differ in their asymptotic scaling. Given a set of small molecules with known activities and a collection of target conformations, the most resource intense method is guaranteed to find the optimal ensemble but scales as O(2(N)). A recursive approximation to the optimal solution scales as O(N(2)), and a more severe approximation leads to a faster method that scales linearly, O(N). The techniques are generally applicable to any system, and we demonstrate their effectiveness on the androgen nuclear hormone receptor (AR), cyclin-dependent kinase 2 (CDK2), and the peroxisome proliferator-activated receptor δ (PPAR-δ) drug targets. Conformations that consisted of a crystal structure and molecular dynamics simulation cluster centroids were used to form AR and CDK2 ensembles. Multiple available crystal structures were used to form PPAR-δ ensembles. For each target, we show that the three methods perform similarly to one another on both the training and test sets.
Melting dynamics of hafnium clusters are investigated using a novel approach based on the idea of the chemical similarity index. Ground state configurations of small hafnium clusters are first derived using Basin-Hopping and Genetic Algorithm in the parallel tempering mode, employing the COMB potential in the energy calculator. These assumed ground state structures are verified by using the Low Lying Structures (LLS) method. The melting process is carried out either by using the direct heating method or prolonged simulated annealing. The melting point is identified by a caloric curve. However, it is found that the global similarity index is much more superior in locating premelting and total melting points of hafnium clusters.
This paper discusses the weighting of two-dimensional fingerprints for similarity-based virtual screening, specifically the use of weights that assign greatest importance to the substructural fragments that occur least frequently in the database that is being screened. Virtual screening experiments using the MDL Drug Data Report and World of Molecular Bioactivity databases show that the use of such inverse frequency weighting schemes can result, in some circumstances, in marked increases in screening effectiveness when compared with the use of conventional, unweighted fingerprints. Analysis of the characteristics of the various schemes demonstrates that such weights are best used to weight the fingerprint of the reference structure in a similarity search, with the database structures' fingerprints unweighted. However, the increases in performance resulting from such weights are only observed with structurally homogeneous sets of active molecules; when the actives are diverse, the best results are obtained using conventional, unweighted fingerprints for both the reference structure and the database structures.
The allosteric modulation of the homodimeric H10-03-6 protein to glycan ligands L1 and L2, and the STAB19 protein to glycan ligands L3 and L4, respectively, has been studied by molecular dynamics simulations and free energy calculations. The results revealed that the STAB19 protein has a significantly higher affinity for L3 (-11.38 ± 2.32 kcal/mol) than that for L4 (-5.51 ± 1.92 kcal/mol). However, the combination of the H10-03-6 protein with glycan L2 (1.23 ± 6.19 kcal/mol) is energetically unfavorable compared with that of L1 (-13.96 ± 0.35 kcal/mol). Further, the binding of glycan ligands L3 and L4 to STAB19 would result in the significant closure of the two CH2 domains of the STAB19 conformation with the decrease of the centroid distances between the two CH2 domains compared with the H10-03-6/L1/L2 complex. The CH2 domain closure of STAB19 relates directly to the formation of new hydrogen bonds and hydrophobic interactions between the residues Ser239, Val240, Asp265, Glu293, Asn297, Thr299, Ser337, Asp376, Thr393, Pro395, and Pro396 in STAB19 and glycan ligands L3 and L4, which suggests that these key residues would contribute to the specific regulation of STAB19 to L3 and L4. In addition, the distance analysis revealed that the EF loop in the H10-03-6/L1/L2 model presents a high flexibility and partial disorder compared with the stabilized STAB19/L3/L4 complex. These results will be helpful in understanding the specific regulation through the asymmetric structural characteristics in the CH2 and CH3 domains of the H10-03-6 and STAB19 proteins.
14-3-3σ plays an important role in controlling tumor metabolic reprogramming and cancer cell growth. However, its function is often compromised in many cancers due to its downregulation. Previous studies found that homodimerization of 14-3-3σ is critical for its activity. However, to date, it is not known if stabilization of 14-3-3σ homodimers can improve its activity or prevent its degradation. In our previous work, we have showed that GCP-Lys-OMe is a potential 14-3-3σ homodimer stabilizer. However, its stabilizing effect was not experimentally validated. Therefore, in this study, we have attempted to predict few potential peptides that can stabilize the dimeric form of 14-3-3σ using similar in silico techniques as described previously for GCP-Lys-OMe. Subsequent [1H]-CPMG NMR experiments confirmed the binding of the peptides (peptides 3, 5, 9, and 16) on 14-3-3σ, with peptide 3 showing the strongest binding. Competitive [1H]-CPMG assays further revealed that while peptide 3 does not compete with a 14-3-3σ binding peptide (ExoS) for the protein's amphipathic groove, it was found to improve ExoS binding on 14-3-3σ. When 14-3-3σ was subjected to dynamic light scattering experiments, the 14-3-3σ homodimer was found to undergo dissociation into monomers prior to aggregation. Intriguingly, the presence of peptide 3 increased 14-3-3σ stability against aggregation. Overall, our findings suggest that (1) docking accompanied by MD simulations can be used to identify potential homodimer stabilizing compounds of 14-3-3σ and (2) peptide 3 can slow down 14-3-3σ aggregation (presumably by preventing its dissociation into monomers), as well as improving the binding of 14-3-3σ to ExoS protein.
In this Viewpoint, we provide a commentary on the impact of the Journal of Chemical Information and Modeling Special Issue on Women in Computational Chemistry published in May 2019 and the feedback we received.
Increased reports of oseltamivir (OTV)-resistant strains of the influenza virus, such as the H274Y mutation on its neuraminidase (NA), have created some cause for concern. Many studies have been conducted in the attempt to uncover the mechanism of OTV resistance in H274Y NA. However, most of the reported studies on H274Y focused only on the drug-bound system, so the direct effects of the mutation on NA itself prior to drug binding still remain unclear. Therefore, molecular dynamics simulations of NA in apo form, followed by principal component analysis and interaction energy calculations, were performed to investigate the structural changes of the NA binding site as a result of the H274Y mutation. It was observed that the disruption of the NA binding site due to the H274Y mutation was initiated by the repulsive effect of Y274 on the 250-loop, which in turn altered the hydrogen-bonding network around residue 274. The rotated W295 side chain caused the upward movement of the 340-loop. Consequently, sliding box docking results suggested that the binding pathway of OTV was compromised because of the disruption of this binding site. This study also highlighted the importance of the functional group at C6 of the sialic acid mimicry. It is hoped that these results will improve the understanding of OTV resistance and shed some light on the design of a novel anti-influenza drug.
The continuing rise in tuberculosis incidence and the problem of drug resistance strains have prompted the research on new drug candidates and the mechanism of drug resistance. Molecular docking and molecular dynamics simulation (MD) were performed to study the binding of isoniazid onto the active site of Mycobacterium tuberculosis enoyl-acyl carrier protein reductase (InhA) in an attempt to address the mycobacterial resistance against isoniazid. Results show that isonicotinic acyl-NADH (INADH) has an extremely high binding affinity toward the wild type InhA by forming stronger interactions compared to the parent drug (isoniazid) (INH). Due to the increase of hydrophobicity and reduction in the side chain's volume of A94 of mutant type InhA, both INADH and the mutated protein become more mobile. Due to this reason, the molecular interactions of INADH with mutant type are weaker than that observed with the wild type. However, the reduced interaction caused by the fluctuation of INADH and the mutant protein only inflected minor resistance in the mutant strain as inferred from free energy calculation. MD results also showed there exists a water-mediated hydrogen bond between INADH and InhA. However, the bridged water molecule is only present in the INADH-wild type complex, reflecting the putative role of the water molecule in the binding of INADH to the wild type protein. The results support the assumption that the conversion of prodrug isoniazid into its active form INADH is mediated by KatG as a necessary step prior to target binding on InhA. Our findings also contribute to a better understanding of INH resistance in mutant type.
A group of flavanones and their chalcones, isolated from Boesenbergia rotunda L., were previously reported to show varying degrees of noncompetitive inhibitory activities toward Dengue virus type 2 (Den2) protease. Results obtained from automated docking studies are in agreement with experimental data in which the ligands were shown to bind to sites other than the active site of the protease. The calculated K(i) values are very small, indicating that the ligands bind quite well to the allosteric binding site. Greater inhibition by pinostrobin, compared to the other compounds, can be explained by H-bonding interaction with the backbone carbonyl of Lys74, which is bonded to Asp75 (one of the catalytic triad residues). In addition, structure-activity relationship analysis yields structural information that may be useful for designing more effective therapeutic drugs against dengue virus infections.
The goal of consensus clustering methods is to find a consensus partition that optimally summarizes an ensemble and improves the quality of clustering compared with single clustering algorithms. In this paper, an enhanced voting-based consensus method was introduced and compared with other consensus clustering methods, including co-association-based, graph-based, and voting-based consensus methods. The MDDR and MUV data sets were used for the experiments and were represented by three 2D fingerprints: ALOGP, ECFP_4, and ECFC_4. The results were evaluated based on the ability of the clustering method to separate active from inactive molecules in each cluster using four criteria: F-measure, Quality Partition Index (QPI), Rand Index (RI), and Fowlkes-Mallows Index (FMI). The experiments suggest that the consensus methods can deliver significant improvements for the effectiveness of chemical structures clustering.