Breaking the barriers of data scarcity in drug-target affinity prediction.

Accurate prediction of drug-target affinity (DTA) is of vital importance in early-stage drug discovery, facilitating the identification of drugs that can effectively interact with specific targets and regulate their activities. While wet experiments remain the most reliable method, they are time-consuming and resource-intensive, resulting in limited data availability that poses challenges for deep learning approaches. Existing methods have primarily focused on developing techniques based on the available DTA data, without adequately addressing the data scarcity issue. To overcome this challenge, we present the Semi-Supervised Multi-task training (SSM) framework for DTA prediction, which incorporates three simple yet highly effective strategies: (1) A multi-task training approach that combines DTA prediction with masked language modeling using paired drug-target data. (2) A semi-supervised training method that leverages large-scale unpaired molecules and proteins to enhance drug and target representations. This approach differs from previous methods that only employed molecules or proteins in pre-training. (3) The integration of a lightweight cross-attention module to improve the interaction between drugs and targets, further enhancing prediction accuracy. Through extensive experiments on benchmark datasets such as BindingDB, DAVIS and KIBA, we demonstrate the superior performance of our framework. Additionally, we conduct case studies on specific drug-target binding activities, virtual screening experiments, drug feature visualizations and real-world applications, all of which showcase the significant potential of our work. In conclusion, our proposed SSM-DTA framework addresses the data limitation challenge in DTA prediction and yields promising results, paving the way for more efficient and accurate drug discovery processes.

A scoring function for the prediction of protein complex interfaces based on the neighborhood preferences of amino acids.

Proteins often assemble into functional complexes, the structures of which are more difficult to obtain than those of the individual protein molecules. Given the structures of the subunits, it is possible to predict plausible complex models via computational methods such as molecular docking. Assessing the quality of the predicted models is crucial to obtain correct complex structures. Here, an energy-scoring function was developed based on the interfacial residues of structures in the Protein Data Bank. The statistically derived energy function (Nepre) imitates the neighborhood preferences of amino acids, including the types and relative positions of neighboring residues. Based on the preference statistics, a program iNepre was implemented and its performance was evaluated with several benchmarking decoy data sets. The results show that iNepre scores are powerful in model ranking to select the best protein complex structures.

Direct experimental observation of blue-light-induced conformational change and intermolecular interactions of cryptochrome.

Cryptochromes are blue light receptors that mediate circadian rhythm and magnetic sensing in various organisms. A typical cryptochrome consists of a conserved photolyase homology region domain and a varying carboxyl-terminal extension across species. The structure of the flexible carboxyl-terminal extension and how carboxyl-terminal extension participates in cryptochrome's signaling function remain mostly unknown. In this study, we uncover the potential missing link between carboxyl-terminal extension conformational changes and downstream signaling functions. Specifically, we discover that the blue-light induced opening of carboxyl-terminal extension in C. reinhardtii animal-like cryptochrome can structurally facilitate its interaction with Rhythm Of Chloroplast 15, a circadian-clock-related protein. Our finding is made possible by two technical advances. Using single-molecule Förster resonance energy transfer technique, we directly observe the displacement of carboxyl-terminal extension by about 15 Å upon blue light excitation. Combining structure prediction and solution X-ray scattering methods, we propose plausible structures of full-length cryptochrome under dark and lit conditions. The structures provide molecular basis for light active conformational changes of cryptochrome and downstream regulatory functions.

Elucidation of partial activation of cannabinoid receptor type 2 and identification of potential partial agonists: Molecular dynamics simulation and structure-based virtual screening.

Cannabinoid receptor type 2 (CB2R) is a member of the class A G protein-coupled receptor (GPCRs) family and a component of the endocannabinoid system that is modulated by the psychoactive chemical from Cannabis sativa, partial agonist Δ9-tetrahydrocannabinol (Δ9-THC). Selective activation of CB2R allows for the treatment of inflammatory and immune-related conditions without the psychotropic effects of CB1R. While CB2R-selective agonists are available, CB2R partial agonists are scarce. Hence, the pharmacological difference between CB2R full agonists and partial agonists remains to be deciphered, prompting the search for novel partial agonists. Here, using an induced-fit docking approach, we built a partial agonist Δ9-THC bound CB2R system from the inactive CB2R structure (PDB ID: 5ZTY) and performed microsecond molecular dynamics (MD) simulations. The simulations reveal an upward shift of the "toggle switch" W6.48(258) and minor outward movement of the transmembrane helix 6 (TM6). Dynamic network model identifies a possible communication path between the ligand and the toggle switch" W6.48(258). Furthermore, to identify potential CB2R partial agonists, we conducted structure-based virtual screening of ZINC15 "Druglike" library containing 17,900742 compounds against 3 conformations derived from MD simulation of CB2R complexed with partial agonist Δ9-THC using Glide virtual screening protocol comprising various filters with increasing accuracy. Nine diverse compounds predicted to have high MM-GBSA binding energy scores and good ADMET properties (including high gastrointestinal absorption and low toxicity) are proposed as potential CB2R partial agonists.

Identification of flexible Pif1-DNA interactions and their impacts on enzymatic activities.

Flexible regions in biomolecular complexes, although crucial to understanding structure-function relationships, are often unclear in high-resolution crystal structures. In this study, we showed that single-molecule techniques, in combination with computational modeling, can characterize dynamic conformations not resolved by high-resolution structure determination methods. Taking two Pif1 helicases (ScPif1 and BsPif1) as model systems, we found that, besides a few tightly bound nucleotides, adjacent solvent-exposed nucleotides interact dynamically with the helicase surfaces. The whole nucleotide segment possessed curved conformations and covered the two RecA-like domains of the helicases, which are essential for the inch-worm mechanism. The synergetic approach reveals that the interactions between the exposed nucleotides and the helicases could be reduced by large stretching forces or electrostatically shielded with high-concentration salt, subsequently resulting in reduced translocation rates of the helicases. The dynamic interactions between the exposed nucleotides and the helicases underlay the force- and salt-dependences of their enzymatic activities. The present single-molecule based approach complements high-resolution structural methods in deciphering the molecular mechanisms of the helicases.

MD Simulations Revealing Special Activation Mechanism of Cannabinoid Receptor 1.

Cannabinoid receptor 1 (CB1) is a G protein-coupled receptor (GPCR) that is gaining much interest for its regulating role in the central nervous system and its value as a drug target. Structures of CB1 in inactive and active states have revealed conformational change details that are not common in other GPCRs. Here, we performed molecular dynamics simulations of CB1 in different ligand binding states and with mutations to reveal its activation mechanism. The conformational change of the "twin toggle switch" residues F2003.36 and W3566.48 that correlates with ligand efficacy is identified as a key barrier step in CB1 activation. Similar conformational change of residues 3.36/6.48 is also observed in melanocortin receptor 4, showing this "twin toggle switch" residue pair is crucial for the activation of multiple GPCR members.

Mutations on RBD of SARS-CoV-2 Omicron variant result in stronger binding to human ACE2 receptor.

The COVID-19 pandemic caused by the SARS-CoV-2 virus has led to more than 270 million infections and 5.3 million of deaths worldwide. Several major variants of SARS-CoV-2 have emerged and posed challenges in controlling the pandemic. The recently occurred Omicron variant raised serious concerns about reducing the efficacy of vaccines and neutralization antibodies due to its vast mutations. We have modelled the complex structure of the human ACE2 protein and the receptor binding domain (RBD) of Omicron Spike protein (S-protein), and conducted atomistic molecular dynamics simulations to study the binding interactions. The analysis shows that the Omicron RBD binds more strongly to the human ACE2 protein than the original strain. The mutations at the ACE2-RBD interface enhance the tight binding by increasing hydrogen bonding interaction and enlarging buried solvent accessible surface area.

Computational insights into differential interaction of mammalian angiotensin-converting enzyme 2 with the SARS-CoV-2 spike receptor binding domain.

The severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) is the causative agent of the COVID-19 pandemic. Angiotensin-converting enzyme 2 (ACE2) has been identified as the host cell receptor that binds to the receptor-binding domain (RBD) of the SARS-COV-2 spike protein and mediates cell entry. Because the ACE2 proteins are widely available in mammals, it is important to investigate the interactions between the RBD and the ACE2 of other mammals. Here we analyzed the sequences of ACE2 proteins from 16 mammals, predicted the structures of ACE2-RBD complexes by homology modeling, and refined the complexes using molecular dynamics simulation. Analyses on sequence, structure, and dynamics synergistically provide valuable insights into the interactions between ACE2 and RBD. The analysis outcomes suggest that the ACE2 of bovine, cat, and panda form strong binding interactions with RBD, while in the cases of rat, least horseshoe bat, horse, pig, mouse, and civet, the ACE2 proteins interact weakly with RBD.

Drug repurposing against SARS-CoV-2 receptor binding domain using ensemble-based virtual screening and molecular dynamics simulations.

Severe acute respiratory syndrome coronavirus-2 (SARS-CoV-2) has caused worldwide pandemic and is responsible for millions of worldwide deaths due to -a respiratory disease known as COVID-19. In the search for a cure of COVID-19, drug repurposing is a fast and cost-effective approach to identify anti-COVID-19 drugs from existing drugs. The receptor binding domain (RBD) of the SARS-CoV-2 spike protein has been a main target for drug designs to block spike protein binding to ACE2 proteins. In this study, we probed the conformational plasticity of the RBD using long molecular dynamics (MD) simulations, from which, representative conformations were identified using clustering analysis. Three simulated conformations and the original crystal structure were used to screen FDA approved drugs (2466 drugs) against the predicted binding site at the ACE2-RBD interface, leading to 18 drugs with top docking scores. Notably, 16 out of the 18 drugs were obtained from the simulated conformations, while the crystal structure suggests poor binding. The binding stability of the 18 drugs were further investigated using MD simulations. Encouragingly, 6 drugs exhibited stable binding with RBD at the ACE2-RBD interface and 3 of them (gonadorelin, fondaparinux and atorvastatin) showed significantly enhanced binding after the MD simulations. Our study shows that flexibility modeling of SARS-CoV-2 RBD using MD simulation is of great help in identifying novel agents which might block the interaction between human ACE2 and the SARS-CoV-2 RBD for inhibiting the virus infection.

Activation Mechanism of Corticotrophin Releasing Factor Receptor Type 1 Elucidated Using Molecular Dynamics Simulations.

The corticotropin-releasing factor receptor type 1 (CRF1R), a member of class B G-protein-coupled receptors (GPCRs), is a good drug target for treating depression, anxiety, and other stress-related neurodisorders. However, there is no approved drug targeting the CRF1R to date, partly due to inadequate structural information and its elusive activation mechanism. Here, by use of the crystal structures of its transmembrane domain (TMD) and the N-terminal extracellular domain (ECD) as a template, a full-length homology model of CRF1R was built and its complexes with peptide agonist urocortin 1 or small molecule antagonist CP-376395 were subjected to all-atom molecular dynamics simulations. We observed well preserved helical contents in the TMD through simulations, while the transmembrane (TM) helices showed clear rearrangements. The TM rearrangement is especially pronounced for the TM6 in the agonist-bound CRF1R system. The observed conformational changes are likely due to breakage of interhelical/inter-regional hydrogen bonds in the TMD. Dynamical network analysis identifies communities with high connections to TM6. Simulations reveal three key residues, Y3566.53, Q3847.49, and L3957.60, which corroborate experimental mutagenesis data, implying the important roles in the receptor activation. The observed large-scale conformational changes are related to CRF1R activation by agonist binding, providing guidance for ligand design.

Development and Validation of a Nomogram for Preoperative Prediction of Central Compartment Lymph Node Metastasis in Patients with Papillary Thyroid Carcinoma and Type 2 Diabetes Mellitus.

PURPOSE: To develop and validate a nomogram to predict central compartment lymph node metastasis in PTC patients with Type 2 Diabetes. PATIENTS AND METHODS: The total number of enrolled patients was 456. The optimal cut-off values of continuous variables were obtained by ROC curve analysis. Significant risk factors in univariate analysis were further identified to be independent variables in multivariable logistic regression analysis, which were then incorporated and presented in a nomogram. The ROC curve analysis was performed to evaluate the discrimination of the nomogram, calibration curves and Hosmer-Lemeshow test were used to visualize and quantify the consistency. Decision curve analysis (DCA) was performed to evaluate the net clinical benefit patients could get by applying this nomogram. RESULTS: ROC curve analysis showed the optimal cutoff values of NLR, PLR, and tumor size were 2.9204, 154.7003, and 0.95 (cm), respectively. Multivariate logistic regression analysis indicated that age, multifocality, largest tumor size, and neutrophil-to-lymphocyte ratio were independent prognostic factors of CLNM. The C-index of this nomogram in the training data set was 0.728, and 0.618 in the external validation data set. When we defined the predicted possibility (>0.5273) as high-risk of CLNM, we could get a sensitivity of 0.535, a specificity of 0.797, a PPV(%) of 67.7, and an NPV(%) of 68.7. Great consistencies were represented in the calibration curves. DCA showed that applying this nomogram will help patients get more clinical net benefit than having all of the patients or none of the patients treated with central compartment lymph node dissection (CLND). CONCLUSION: A high level of preoperative NLR was an independent predictor for CLNM in PTC patients with T2DM. And the verified optimal cutoff value of NLR in this study was 2.9204. Applying this nomogram will help stratify high-risk CLNM patients, consequently enabling these patients to be treated with appropriate measures. What is more, we hope to find more sensitive indicators in the near future to further improve the sensitivity and specificity of our nomogram.

Early-stage dynamics of chloride ion-pumping rhodopsin revealed by a femtosecond X-ray laser.

Chloride ion-pumping rhodopsin (ClR) in some marine bacteria utilizes light energy to actively transport Cl- into cells. How the ClR initiates the transport is elusive. Here, we show the dynamics of ion transport observed with time-resolved serial femtosecond (fs) crystallography using the Linac Coherent Light Source. X-ray pulses captured structural changes in ClR upon flash illumination with a 550 nm fs-pumping laser. High-resolution structures for five time points (dark to 100 ps after flashing) reveal complex and coordinated dynamics comprising retinal isomerization, water molecule rearrangement, and conformational changes of various residues. Combining data from time-resolved spectroscopy experiments and molecular dynamics simulations, this study reveals that the chloride ion close to the Schiff base undergoes a dissociation-diffusion process upon light-triggered retinal isomerization.

Harnessing peak transmission around symptom onset for non-pharmaceutical intervention and containment of the COVID-19 pandemic.

Within a short period of time, COVID-19 grew into a world-wide pandemic. Transmission by pre-symptomatic and asymptomatic viral carriers rendered intervention and containment of the disease extremely challenging. Based on reported infection case studies, we construct an epidemiological model that focuses on transmission around the symptom onset. The model is calibrated against incubation period and pairwise transmission statistics during the initial outbreaks of the pandemic outside Wuhan with minimal non-pharmaceutical interventions. Mathematical treatment of the model yields explicit expressions for the size of latent and pre-symptomatic subpopulations during the exponential growth phase, with the local epidemic growth rate as input. We then explore reduction of the basic reproduction number R0 through specific transmission control measures such as contact tracing, testing, social distancing, wearing masks and sheltering in place. When these measures are implemented in combination, their effects on R0 multiply. We also compare our model behaviour to the first wave of the COVID-19 spreading in various affected regions and highlight generic and less generic features of the pandemic development.

GDP Release from the Open Conformation of Gα Requires Allosteric Signaling from the Agonist-Bound Human ß2 Adrenergic Receptor.

G-protein-coupled receptors (GPCRs) transmit signals into the cell in response to ligand binding at its extracellular domain, which is characterized by the coupling of agonist-induced receptor conformational change to guanine nucleotide (GDP) exchange with guanosine triphosphate on a heterotrimeric (αßγ) guanine nucleotide-binding protein (G-protein), leading to the activation of the G-protein. The signal transduction mechanisms have been widely researched in vivo and in silico. However, coordinated communication from stimulating ligands to the bound GDP still remains elusive. In the present study, we used microsecond (µS) molecular dynamic (MD) simulations to directly probe the communication from the ß2 adrenergic receptor (ß2AR) with an agonist or an antagonist or no ligand to GDP bound to the open conformation of the Gα protein. Molecular mechanism-general Born surface area calculation results indicated either the agonist or the antagonist destabilized the binding between the receptor and the G-protein but the agonist caused a higher level of destabilization than the antagonist. This is consistent with the role of agonist in the activation of the G-protein. Interestingly, while GDP remained bound with the Gα-protein for the two inactive systems (antagonist-bound and apo form), GDP dissociated from the open conformation of the Gα protein for the agonist activated system. Data obtained from MD simulations indicated that the receptor and the Gα subunit play a big role in coordinated communication and nucleotide exchange. Based on residue interaction network analysis, we observed that engagement of agonist-bound ß2AR with an α5 helix of Gα is essential for the GDP release and the residues in the phosphate-binding loop, α1 helix, and α5 helix play very important roles in the GDP release. The insights on GPCR-G-protein communication will facilitate the rational design of agonists and antagonists that target both active and inactive GPCR binding pockets, leading to more precise drugs.

Structure-Based Functional Modification Study of a Cyanobacterial Chloride Pump for Transporting Multiple Anions.

Understanding the structure and functional mechanisms of cyanobacterial halorhodopsin has become increasingly important, given the report that Synechocystis halorhodopsin (SyHR), a homolog of the cyanobacterial halorhodopsin from Mastigocladopsis repens (MrHR), can take up divalent ions, such as SO42-, as well as chloride ions. Here, the crystal structure of MrHR, containing a unique "TSD" chloride ion conduction motif, was determined as a homotrimer at a resolution of 1.9 Å. The detailed structure of MrHR revealed a unique trimeric topology of the light-driven chloride pump, with peculiar coordination of two water molecules and hydrogen-mediated bonds near the TSD motif, as well as a short B-C loop. Structural and functional analyses of MrHR revealed key residues responsible for the anion selectivity of cyanobacterial halorhodopsin and the involvement of two chloride ion-binding sites in the ion conduction pathway. Alanine mutant of Asn63, Pro118, and Glu182 locating in the anion inlet induce multifunctional uptake of chloride, nitrate, and sulfate ions. Moreover, the structure of N63A/P118A provides information on how SyHR promotes divalent ion transport. Our findings significantly advance the structural understanding of microbial rhodopsins with different motifs. They also provide insight into the general structural framework underlying the molecular mechanisms of the cyanobacterial chloride pump containing SyHR, the only molecule known to transport both sulfate and chloride ions.

Author Correction: Prediction of disulfide bond engineering sites using a machine learning method.

An amendment to this paper has been published and can be accessed via a link at the top of the paper.

Author Correction: Neighborhood Preference of Amino Acids in Protein Structures and its Applications in Protein Structure Assessment.

An amendment to this paper has been published and can be accessed via a link at the top of the paper.

Prediction of disulfide bond engineering sites using a machine learning method.

Disulfide bonds are covalently bonded sulfur atoms from cysteine pairs in protein structures. Due to the importance of disulfide bonds in protein folding and structural stability, artificial disulfide bonds are often engineered by cysteine mutation to enhance protein structural stability. To facilitate the experimental design, we implemented a method based on neural networks to predict amino acid pairs for cysteine mutations to form engineered disulfide bonds. The designed neural network was trained with high-resolution structures curated from the Protein Data Bank. The testing results reveal that the proposed method recognizes 99% of natural disulfide bonds. In the test with engineered disulfide bonds, the algorithm achieves similar accuracy levels with other state-of-the-art algorithms in published dataset and better performance for two comprehensively studied proteins with 70% accuracy, demonstrating potential applications in protein engineering. The neural network framework allows exploiting the full features in distance space, and therefore improves accuracy of the disulfide bond engineering site prediction. The source code and a web server are available at http://liulab.csrc.ac.cn/ssbondpre.