ETLD: an encoder-transformation layer-decoder architecture for protein contact and mutation effects prediction.

The latent features extracted from the multiple sequence alignments (MSAs) of homologous protein families are useful for identifying residue-residue contacts, predicting mutation effects, shaping protein evolution, etc. Over the past three decades, a growing body of supervised and unsupervised machine learning methods have been applied to this field, yielding fruitful results. Here, we propose a novel self-supervised model, called encoder-transformation layer-decoder (ETLD) architecture, capable of capturing protein sequence latent features directly from MSAs. Compared to the typical autoencoder model, ETLD introduces a transformation layer with the ability to learn inter-site couplings, which can be used to parse out the two-dimensional residue-residue contacts map after a simple mathematical derivation or an additional supervised neural network. ETLD retains the process of encoding and decoding sequences, and the predicted probabilities of amino acids at each site can be further used to construct the mutation landscapes for mutation effects prediction, outperforming advanced models such as GEMME, DeepSequence and EVmutation in general. Overall, ETLD is a highly interpretable unsupervised model with great potential for improvement and can be further combined with supervised methods for more extensive and accurate predictions.

Simultaneous Improvement in the Thermostability and Catalytic Activity of Epoxidase Lsd18 for the Synthesis of Lasalocid A.

Enzymes used in the synthesis of natural products are potent catalysts, capable of efficient and stereoselective chemical transformations. Lsd18 catalyzes two sequential epoxidations during the biosynthesis of lasalocid A, a polyether polyketide natural product. We performed protein engineering on Lsd18 to improve its thermostability and catalytic activity. Utilizing structure-guided methods of FoldX and Rosetta-ddG, we designed 15 mutants of Lsd18. Screening of these mutants using thermal shift assay identified stabilized variants Lsd18-T189M, Lsd18-S195M, and the double mutant Lsd18-T189M-S195M. Trypsin digestion, molecular dynamic simulation, circular dichroism (CD) spectroscopy, and X-ray crystallography provided insights into the molecular basis for the improved enzyme properties. Notably, enhanced hydrophobic interaction within the enzyme core and interaction of the protein with the FAD cofactor appear to be responsible for its better thermostability.

Identification of Potential Lead Compounds Targeting Novel Druggable Cavity of SARS-CoV-2 Spike Trimer by Molecular Dynamics Simulations.

The global pandemic of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) has become an urgent public health problem. Spike (S) protein mediates the fusion between the virus and the host cell membranes, consequently emerging as an important target of drug design. The lack of comparisons of in situ full-length S homotrimer structures in different states hinders understanding the structures and revealing the function, thereby limiting the discovery and development of therapeutic agents. Here, the steady-state structures of the in situ full-length S trimer in closed and open states (Sclosed and Sopen) were modeled with the constraints of density maps, associated with the analysis of the dynamic structural differences. Subsequently, we identified various regions with structure and property differences as potential binding pockets for ligands that promote the formation of inactive trimeric protein complexes. By using virtual screening strategy and a newly defined druggable cavity, five ligands were screened with potential bioactivities. Then molecular dynamic (MD) simulations were performed on apo protein structures and ligand bound complexes to reveal the conformational changes upon ligand binding. Our simulation results revealed that sulforaphane (SFN), which has the best binding affinity, could inhibit the conformational changes of S homotrimer that would occur during the viral membrane fusion. Our results could aid in the understanding of the regulation mechanism of S trimer aggregation and the structure-activity relationship, facilitating the development of potential antiviral agents.

Mechanism of Glycans Modulating Cholesteryl Ester Transfer Protein: Unveiled by Molecular Dynamics Simulation.

Inhibition of the cholesteryl ester transfer protein (CETP) has been considered as a promising way for the treatment of cardiovascular disease (CVD) for three decades. However, clinical trials of several CETP inhibitors with various potencies have been marginally successful at best, raising doubts on the target drugability of CETP. The in-depth understanding of the glycosylated CETP structure could be beneficial to more definitive descriptions of the CETP function and the underlying mechanism. In this work, large-scale molecular dynamics simulations were performed to thoroughly explore the mechanism of glycans modulating CETP. Here, the extensive simulation results intensely suggest that glycan88 tends to assist CETP in forming a continuous tunnel throughout interacting with the upper-right region of the N-barrel, while it also could prevent the formation of a continuous tunnel by swinging toward the right-rear of the N-barrel. Furthermore, glycan240 formed stable H-bonds with Helix-B and might further stabilize the central cavity of CETP. Furthermore, the nonspecific involvement of the hydroxyl groups from the various glycans with protein core interactions and the similar influence of different glycans trapped at similar regions on the protein structure suggest that physiological glycan may lead to a similar effect. This study would provide valuable insights into devising novel methods for CVD treatment targeting CETP and functional studies about glycosylation for other systems.

p38α Kinase Auto-Activation through Its Conformational Transition Induced by Tyr323 Phosphorylation.

p38α is a key serine/threonine kinase that can enable atypical auto-activation through Zap70 phosphorylation and initiate T cell receptor signaling. The auto-activation plays an important role in autoimmune diseases. Although the classical activation mechanism of p38α has been studied in-depth, the atypical activation mechanism of Y323 phosphorylation-induced p38α auto-activation remains largely unexplained, especially the regulatory effects of phosphorylation on different sites (Y323 vs T180). From the X-ray experimental data, we identified the inactive and active states of p38α using principal component analysis. To understand the auto-activation process and the internal driving mechanism, a computational paradigm that couples the targeted molecular dynamics simulations, the String Method, and the umbrella sampling strategy were employed to generate the conformational landscape of p38α, including p38α T180-Y323, p38α T180-pY323, and p38α pT180-pY323 systems (pT180/pY323: phosphorylated T180/Y323). We explored that pY323 could change the conformational distribution and promote the conformational transition of p38α from the inactive state to the active state. Auto-activation of p38α is regulated by pY323 through destabilization of the hydrophobic core structure and aided by R173. This study will further explain the conformational transition of p38α induced by Y323 phosphorylation and provide insights into the universal molecular auto-activation mechanism of the p38 subfamily at the atomic level.

TAB1 binding induced p38α conformation change: an accelerated molecular dynamics simulation study.

p38α mitogen-activated protein kinase (MAPK) undergoes autophosphorylation induced by the binding of TGFß-activated kinase 1 binding protein 1 (TAB1) in myocardial ischemia. Investigation of the conformational transformations in p38α triggered by TAB1 binding is motivated by the need to find selective p38α activation inhibitors to treat myocardial ischemia. Herein, the conformational transformations of p38α were studied via all-atom accelerated molecular dynamics simulations and principal component analysis. With the binding of TAB1, the conformational changes of p38α auto-activation were characterized by the movement of the activation loop (A-loop) away from the αG helix toward the αF, αE helixes and L16-loop. In addition, a diverse intermediate state with an extensional and phosphorylated A-loop different from the transition intermediate state was explored. The conformational changes, including the A-loop alpha-structure breaking and the stronger hydrogen bond network formation, are accompanied by the extension of the A-loop and more intramolecular interactions in p38α. TAB1 correlates with other regions of p38α that are distal from the TAB1-binding site, including the A-loop, αC helix, and L16-loop, which regulates the intramolecular correlation of p38α. And, the phosphorylation further enhances the correlations between the A-loop and the other regions of p38α. The correlation results imply the regulation process of p38α conformational transformations. These findings will improve our understanding of the autophosphorylation of kinase and facilitate the development of selective inhibitors for the treatment of ischemic injury.

Recognition between CD147 and cyclophilin A deciphered by accelerated molecular dynamics simulations.

CD147 functions as the receptor of extracellular cyclophilin A (CypA) in various diseases, and CD147-CypA binding ulteriorly underlies the pathological process of various viral infections including HIV-1, SARS, and SARS-CoV-2. Although CyPA has been identified as a key intermediate pro-inflammatory factor, the mechanism by which CD147 cooperates with CypA in the development of the cytokine storm remains largely unknown, and the binding profile of CD147 with CypA remains to be elucidated as well. Here, we prepared three binding models of the CD147-CypA complex, including the active site of CypA severally binding to the groove bound by the Ig1 and Ig2 domains (model-0), P180-G181 (model-1), and P211 (model-2) of CD147, as well as introducing mutations P180A-G181A and P211A individually in each model. All systems were studied using accelerated molecular dynamics simulations and the molecular mechanics generalized Born surface area (MM/GBSA) method. For model-0, CypA bound to the ectodomain of CD147 with the highest binding affinity. Moreover, mutations P180A-G181A of CD147 in model-0 decreased the binding affinity and weakened the dynamic correlation between CD147 and CypA, which resulted in CypA shifting from the initial binding location. Other residue mutations of CD147 did not significantly affect the CD147-CypA binding, as reflected by the energy and structural analyses. Compared with surface plasmon resonance results and nuclear magnetic resonance shift signals, CypA should tend to reciprocally bind to the groove of CD147, and the binding process might be modulated by P180-G181 rather than P211. Besides, residue R201 of CD147 is critical for CD147-CypA binding and needs further experimental verification. These findings further our understanding of the recruitment between CD147 and CypA and its potential role in the development of inflammation and viral infection.

Molecular insights into the binding variance of the SARS-CoV-2 spike with human, cat and dog ACE2 proteins.

SARS-CoV-2 has recently caused an epidemic in humans and poses a huge threat to global public health. As a primary receptor of SARS-CoV-2, angiotensin-converting enzyme 2 (ACE2) exists in different hosts that are in close contact with humans, especially cats and dogs. However, the underlying mechanism of how the spike receptor binding domain (RBD) of SARS-CoV-2 cooperates with human ACE2 (hACE2), cat ACE2 (cACE2) and dog ACE2 (dACE2) and the variation in binding remains largely unsolved. Therefore, we explored the binding behavior of the spike RBD with cACE2, dACE2 and hACE2 via all-atom molecular dynamics simulations. In accordance with the binding free energies and residue interactions, the spike RBD has respective binding specificities with cACE2, dACE2 and hACE2, and the binding affinities decrease in the order of hACE2, cACE2, dACE2, mainly due to changes in the amino acids Q24L, H34Y, and M82T in cACE2 or dACE2. Furthermore, alanine scanning analysis results validated some key residues of the spike RBD interact with ACE2 and provided clues to the variation of amino acid that could influence the transmissibility or immune responses of SARS-CoV-2. Decreasing dynamic correlations strengths of ACE2 with the RBD were found in all hACE2-RBD, cACE2-RBD and dACE2-RBD systems. The ACE2 protein shows variable motion modes across the zinc metallopeptidase domain, which induces different interactions between ACE2 and the RBD. Our studies reveal that the motion pattern of the zinc metallopeptidase domain is critical to the binding behavior of RBD with ACE2. These findings could aid our understanding of selective recognition involving various ACE2 with the SARS-CoV-2 spike and shed further light on the binding mechanisms.

Markov State Models Underlying the N-Terminal Premodel of TOPK/PBK.

Lymphokine-activated killer T-cell-originated protein kinase (TOPK) is a potential target for cancer therapy. To explore the micromechanism, we proposed the N-terminal premodel (NTPM) of the TOPK monomer via homology modeling and molecular dynamic simulations and analyzed the conformational dynamics by Markov state model analysis. The electronegative insert (ENI) motif of the NTPM can be opened with a small probability under wild type, regulated by the so-called "N-C" interaction zone consisting of the N-terminal head, the coil between ß3-strand and αC-helix, and the ENI motif. Glutamate substitution at threonine residue 9 or tyrosine residue 74 promotes the closed-open transition, revealing the details of phosphorylation. Allosteric effects induce functionally relevant structural changes, such as increased structural flexibility and active sites, which are thought to be necessary for further activation or binding. These findings provide rational structural templates for designing state-dependent inhibitors and give insight into the molecular regulatory mechanisms of TOPK monomers.

Sequence Matching between Hemagglutinin and Neuraminidase through Sequence Analysis Using Machine Learning.

To date, many experiments have revealed that the functional balance between hemagglutinin (HA) and neuraminidase (NA) plays a crucial role in viral mobility, production, and transmission. However, whether and how HA and NA maintain balance at the sequence level needs further investigation. Here, we applied principal component analysis and hierarchical clustering analysis on thousands of HA and NA sequences of A/H1N1 and A/H3N2. We discovered significant coevolution between HA and NA at the sequence level, which is closely related to the type of host species and virus epidemic years. Furthermore, we propose a sequence-to-sequence transformer model (S2STM), which mainly consists of an encoder and a decoder that adopts a multi-head attention mechanism for establishing the mapping relationship between HA and NA sequences. The training results reveal that the S2STM can effectively realize the "translation" from HA to NA or vice versa, thereby building a relationship network between them. Our work combines unsupervised and supervised machine learning methods to identify the sequence matching between HA and NA, which will advance our understanding of IAVs' evolution and also provide a novel idea for sequence analysis methods.

Receptor Dynamics in Molecular Recognition by Cryo-EM and Molecular Simulation.

The appropriate selection of initial receptor structure has been the "cornerstone" or foundation of successful structure-based virtual screening (SBVS), and plagued the structure-based design with a significant practical problem to determine the major physiological states or important transition states of receptors (e.g. proteins with multiple low-energy conformations and liganddependent conformational dynamics). It is well known that current SBVS methods lack the capacity to capture and characterize the intrinsic receptor flexibility with ideal cost-effectiveness. In recent years, cryoelectron microscopy (cryo-EM) has been routinely applied in the determination of biomolecular assemblies within the physiological state. In this work, we review the roles of cryo-EM and ensemble docking methods to present the intrinsically dynamic behavior of biomacromolecules, as well as the ever-improving estimation of ligand binding affinities and receptor-ligand thermodynamics. Finally, we also provide a viewpoint for further research works on modeling receptor dynamics.

Structure-based methoxyflavone derivatives with potent inhibitory activity against various influenza neuraminidases.

Communicated by Ramaswamy H. Sarma.

Rapid Structure-Based Screening Informs Potential Agents for Coronavirus Disease (COVID-19) Outbreak.

Coronavirus Disease 2019 (COVID-19), caused by the novel coronavirus, has spread rapidly across China. Consequently, there is an urgent need to sort and develop novel agents for the prevention and treatment of viral infections. A rapid structure-based virtual screening is used for the evaluation of current commercial drugs, with structures of human angiotensin converting enzyme II (ACE2), and viral main protease, spike, envelope, membrane and nucleocapsid proteins. Our results reveal that the reported drugs Arbidol, Chloroquine and Remdesivir may hinder the entry and release of virions through the bindings with ACE2, spike and envelope proteins. Due to the similar binding patterns, NHC (ß-d-N4-hydroxycytidine) and Triazavirin are also in prospects for clinical use. Main protease (3CLpro) is likely to be a feasible target of drug design. The screening results to target 3CL-pro reveal that Mitoguazone, Metformin, Biguanide Hydrochloride, Gallic acid, Caffeic acid, Sulfaguanidine and Acetylcysteine seem be possible inhibitors and have potential application in the clinical therapy of COVID-19.