Deep learning is an important branch of artificial intelligence that has been successfully applied into medicine and two-dimensional ligand design. The three-dimensional (3D) ligand generation in the 3D pocket of protein target is an interesting and challenging issue for drug design by deep learning. Here, the MolAICal software is introduced to supply a way for generating 3D drugs in the 3D pocket of protein targets by combining with merits of deep learning model and classical algorithm. The MolAICal software mainly contains two modules for 3D drug design. In the first module of MolAICal, it employs the genetic algorithm, deep learning model trained by FDA-approved drug fragments and Vinardo score fitting on the basis of PDBbind database for drug design. In the second module, it uses deep learning generative model trained by drug-like molecules of ZINC database and molecular docking invoked by Autodock Vina automatically. Besides, the Lipinski’s rule of five, Pan-assay interference compounds (PAINS), synthetic accessibility (SA) and other user-defined rules are introduced for filtering out unwanted ligands in MolAICal. To show the drug design modules of MolAICal, the membrane protein glucagon receptor and non-membrane protein SARS-CoV-2 main protease are chosen as the investigative drug targets. The results show MolAICal can generate the various and novel ligands with good binding scores and appropriate XLOGP values. We believe that MolAICal can use the advantages of deep learning model and classical programming for designing 3D drugs in protein pocket. MolAICal is freely for any nonprofit purpose and accessible at https://molaical.github.io.

Epidermal growth factor receptor (EGFR) is a clinical therapeutic target to treat a subset of non-small cell lung cancer (NSCLC) harboring EGFR mutants. However, some patients with a similar kind of EGFR mutation show intrinsic resistance to tyrosine kinase inhibitors (TKI). It indicates that other key molecules are involved in the survival of these cancer cells. We showed here that 2-[(aminocarbonyl)amino]-5 -(4-fluorophenyl)-3-thiophenecarboxamide (TPCA-1), a previously reported inhibitor of IkB kinases (IKK), blocked STAT3 recruitment to upstream kinases by docking into SH2 domain of STAT3 and attenuated STAT3 activity induced by cytokines and cytoplasmic tyrosine kinases. TPCA-1 is an effective inhibitor of STAT3 phosphorylation, DNA binding, and transactivation in vivo. It selectively repressed proliferation of NSCLC cells with constitutive STAT3 activation. In addition, using pharmacologic and genetic approaches, we found that both NF-kB and STAT3 could regulate the transcripts of interleukin (IL)-6 and COX-2 in NSCLC harboring EGFR mutations. Moreover, gefitinib treatment only did not efficiently suppress NF-kB and STAT3 activity. In contrast, we found that treatment with TKIs increased phosho-STAT3 level in target cells. Inhibiting EGFR, STAT3, and NF-kB by combination of TKIs with TPCA-1 showed increased sensitivity and enhanced apoptosis induced by gefitinib. Collectively, in this work, we identified TPCA-1 as a direct dual inhibitor for both IKKs and STAT3, whereas treatment targeting EGFR only could not sufficiently repress NF-kB and STAT3 pathways for lung cancers harboring mutant EGFR. Therefore, synergistic treatment of TPCA-1 with TKIs has potential to be a more effective strategy for cancers. Mol Cancer Ther; 13(3); 617-29. Ó2014 AACR.

C-reactive protein (CRP) 3 is a major human acute phase reactant composed of five identical subunits (1, 2). Accumulating evidence demonstrates that the actions of CRP depend on conformation and localization (3-5). CRP is primarily produced by the liver and circulates in the blood as a pentamer but is induced to dissociate into subunits (monomeric CRP, mCRP) upon interaction with the microenvironment at inflammatory loci (6 -18). mCRP exhibits markedly enhanced activities and recognizes an expanded list of binding partners. Following reduction of the intra-subunit disulfide bond, mCRP can be further activated (19). The degradation of mCRP, on the other hand, would generate bioactive fragments showing anti-inflammatory actions (20 -23). The differential contributions of these CRP conformations at distinct locations may therefore account for the intense controversies regarding the function of CRP in animal models and in clinical studies (3-5). mCRP appears to be the major conformation of CRP that regulates local inflammation (3-5), yet how it acts remains incompletely understood. In particular, though the markedly enhanced binding capability of mCRP underlies its actions, little is known through which sites mCRP recognizes diverse ligands. Consequently, no means is available to specifically modulate the actions of mCRP, which is necessary for clarifying the exact contributions of different CRP conformations in vitro and in vivo. The current study was designed to identify the recognition site of mCRP for ligand binding. The results unexpectedly demonstrated cholesterol binding sequence (CBS) as a versatile motif that mediates the interactions of mCRP with different types of ligands, including two newly identified herein. We further showed that synthetic CBS peptide was able to inhibit the proinflammatory actions of mCRP both in vitro and in vivo. Hence, optimized CBS peptide may be developed as a potential inhibitor of mCRP. Experimental ProceduresReagents-Human native CRP (purity Ͼ 99%) purified from ascites was purchased from the BindingSite (Birmingham, UK; catalogue number: BP300.X). mCRP and acylated Cys-mutated mCRP was prepared as described (19,24). Proteins were dialyzed to remove NaN 3 , and passed through Detoxi-Gel Columns (Thermo Fisher Scientific, Rockford, IL; catalogue number: 20344) to remove endotoxin when necessary. CRP peptides (purity Ͼ 98%) were synthesized by Science Peptide Biological Technology (Shanghai, China). Lyophilized peptides were reconstituted aseptically with DMSO at 40 mg/ml and stored at Ϫ20°C in aliquots or kept at 4°C for a maximum of 1 week.

The reported crystal structures of β2 adrenergic receptor (β2AR) reveal that the open and closed states of the water channel are correlated with the inactive and active conformations of β2AR. However, more details about the process by which the water channel states are affected by the active to inactive conformational change of β2AR remain illusive. In this work, molecular dynamics simulations are performed to study the dynamical inactive and active conformational change of β2AR induced by inverse agonist ICI 118,551. Markov state model analysis and free energy calculation are employed to explore the open and close states of the water channel. The simulation results show that inverse agonist ICI 118,551 can induce water channel opening during the conformational transition of β2AR. Markov state model (MSM) analysis proves that the energy contour can be divided into seven states. States S1, S2 and S5, which represent the active conformation of β2AR, show that the water channel is in the closed state, while states S4 and S6, which correspond to the intermediate state conformation of β2AR, indicate the water channel opens gradually. State S7, which represents the inactive structure of β2AR, corresponds to the full open state of the water channel. The opening mechanism of the water channel is involved in the ligand-induced conformational change of β2AR. These results can provide useful information for understanding the opening mechanism of the water channel and will be useful for the rational design of potent inverse agonists of β2AR.

Dynamics of the actin cytoskeleton are essential for pollen germination and pollen tube growth. ACTIN-DEPOLYMERIZING FACTORs (ADFs) typically contribute to actin turnover by severing/depolymerizing actin filaments. Recently, we demonstrated that Arabidopsis subclass III ADFs (ADF5 and ADF9) evolved F-actin-bundling function from conserved F-actin-depolymerizing function. However, little is known about the physiological function, the evolutional significance, and the actin-bundling mechanism of these neofunctionalized ADFs. Here, we report that loss of ADF5 function caused delayed pollen germination, retarded pollen tube growth, and increased sensitive to latrunculin B (LatB) treatment by affecting the generation and maintenance of actin bundles. Examination of actin filament dynamics in living cells revealed that the bundling frequency was significantly decreased in adf5 pollen tubes, consistent with its biochemical functions. Further biochemical and genetic complementation analyses demonstrated that both the N- and C-terminal actin-binding domains of ADF5 are required for its physiological and biochemical functions. Interestingly, while both are atypical actin-bundling ADFs, ADF5, but not ADF9, plays an important role in mature pollen physiological activities. Taken together, our results suggest that ADF5 has evolved the function of bundling actin filaments and plays an important role in the formation, organization, and maintenance of actin bundles during pollen germination and pollen tube growth.

De novo drug design is a stationary way to build novel ligands in the confined pocket of receptor by assembling the atoms or fragments, while molecular dynamics (MD) simulation is a dynamical way to study the interaction mechanism between the ligands and receptors based on the molecular force field. De novo drug design and MD simulation are effective tools for novel drug discovery. With the development of technology, deep learning methods, and interpretable machine learning (IML) have emerged in the research area of drug design. Deep learning methods and IML can be used further to improve the efficiency and accuracy of de novo drug design and MD simulations. The application summary of deep learning methods for de novo drug design, MD simulations, and IML can further promote the technical development of drug discovery. In this article, two major workflow methods and the related components of classical algorithm and deep learning are described for de novo drug design from a new perspective. The application progress of deep learning is also summarized for MD simulations. Furthermore, IML is introduced for the deep learning model interpretability of de novo drug design and MD simulations. Our paper deals with an interesting topic about deep learning applications of de novo drug design and MD simulations for the scientific community. This article is categorized under: Data Science > Chemoinformatics Data Science > Artificial Intelligence/Machine Learning

Recently, small-molecule compounds have been reported to block the PD-1/PD-L1 interaction by inducing the dimerization of PD-L1. All these inhibitors had a common scaffold and interacted with the cavity formed by two PD-L1 monomers. This special interactive mode provided clues for the structure-based drug design, however, also showed limitations for the discovery of small-molecule inhibitors with new scaffolds. In this study, we revealed the structure-activity relationship of the current small-molecule inhibitors targeting dimerization of PD-L1 by predicting their binding and unbinding mechanism via conventional molecular dynamics and metadynamics simulation. During the binding process, the representative inhibitors (BMS-8 and BMS-1166) tended to have a more stable binding mode with one PD-L1 monomer than the other and the small-molecule inducing PD-L1 dimerization was further stabilized by the non-polar interaction of Ile54, Tyr56, Met115, Ala121, and Tyr123 on both monomers and the water bridges involved in ALys124. The unbinding process prediction showed that the PD-L1 dimerization kept stable upon the dissociation of ligands. It's indicated that the formation and stability of the small-molecule inducing PD-L1 dimerization was the key factor for the inhibitory activities of these ligands. The contact analysis, R-group based quantitative structure-activity relationship (QSAR) analysis and molecular docking further suggested that each attachment point on the core scaffold of ligands had a specific preference for pharmacophore elements when improving the inhibitory activities by structural modifications. Taken together, the results in this study could guide the structural optimization and the further discovery of novel small-molecule inhibitors targeting PD-L1.

As co-chaperones of the 90-kDa heat shock protein(HSP90), FK506 binding protein 51 (FKBP51) and FK506 binding protein 52 (FKBP52) modulate the maturation of steroid hormone receptor through their specific FK1 domains (FKBP12-like domain 1). The inhibitors targeting FK1 domains are potential therapies for endocrine-related physiological disorders. However, the structural conservation of the FK1 domains between FKBP51 and FKBP52 make it difficult to obtain satisfactory selectivity in FK506-based drug design. Fortunately, a series of iFit ligands synthesized by Hausch et al exhibited excellent selectivity for FKBP51, providing new opportunity for design selective inhibitors. We performed molecular dynamics simulation, binding free energy calculation and unbinding pathway analysis to reveal selective mechanism for the inhibitor iFit4 binding with FKBP51 and FKBP52. The conformational stability evaluation of the "Phe67-in" and "Phe67-out" states implies that FKBP51 and FKBP52 have different preferences for "Phe67-in" and "Phe67-out" states, which we suggest as the determinant factor for the selectivity for FKBP51. The binding free energy calculations demonstrate that nonpolar interaction is favorable for the inhibitors binding, while the polar interaction and entropy contribution are adverse for the inhibitors binding.According to the results from binding free energy decomposition, the electrostatic difference of residue 85 causes the most significant thermodynamics effects on the binding of iFit4 to FKBP51 and FKBP52. Furthermore, the importance of substructure units on iFit4 were further evaluated by unbinding pathway analysis and residue-residue contact analysis between iFit4 and the proteins. The results will provide new clues for the design of selective inhibitors for FKBP51.