Search

Accurately predicting molecular properties is a challenging but essential task in drug discovery. Recently, many mono-modal deep learning methods have been successfully applied to molecular property prediction. However, mono-modal learning is inherently limited as it relies solely on a single modality of molecular representation, which restricts a comprehensive understanding of drug molecules. To overcome the limitations, we propose a multimodal fused deep learning (MMFDL) model to leverage information from different molecular representations. Specifically, we construct a triple-modal learning model by employing Transformer-Encoder, Bidirectional Gated Recurrent Unit (BiGRU), and graph convolutional network (GCN) to process three modalities of information from chemical language and molecular graph: SMILES-encoded vectors, ECFP fingerprints, and molecular graphs, respectively. We evaluate the proposed triple-modal model using five fusion approaches on six molecule datasets, including Delaney, Llinas2020, Lipophilicity, SAMPL, BACE, and pKa from DataWarrior. The results show that the MMFDL model achieves the highest Pearson coefficients, and stable distribution of Pearson coefficients in the random splitting test, outperforming mono-modal models in accuracy and reliability. Furthermore, we validate the generalization ability of our model in the prediction of binding constants for protein-ligand complex molecules, and assess the resilience capability against noise. Through analysis of feature distributions in chemical space and the assigned contribution of each modal model, we demonstrate that the MMFDL model shows the ability to acquire complementary information by using proper models and suitable fusion approaches. By leveraging diverse sources of bioinformatics information, multimodal deep learning models hold the potential for successful drug discovery.

The lesion area of cancer presents complex physiological structures, posing significant challenges for accurate segmentation, which is required in radiotherapy or chemo-radiotherapy. The existing general-purpose segmentation models have made significant progress, but their segmentation performance remains unsatisfactory in specific cancers. To enhance the segmentation performance of nasopharyngeal carcinoma (NPC) MRI images, we develop a segmentation model by incorporating skip connections and receptive blocks within the deep convolutional neural network. We curate a collection of NPC images and evaluate the segmentation performance of the proposed model in comparison with the ground truth segmented images. The similarity between the two types of images is quantified by the Dice Similarity Coefficient (DSC) score, intersection over Union (IoU) score, recall value, and precision value. The mean DSC score was increased to 0.91, and the mean IoU score was improved to 0.84 in the segmentation of our curated NPC images. Our proposed method not only shows excellent performance in segmenting NPC, but also displays its generalization capacity on three public medical image datasets including skin cancer, lung, and chest image databases. The proposed model can leverage multi-layer feature extraction to improve the accuracy of tumor region segmentation. The improved generalization ability suggests that harnessing multi-level features in deep neural networks can improve segmentation performance.

Abstract For ligand binding prediction, it is crucial for molecular docking programs to integrate template-based modeling with a precise scoring function. Here, we proposed the CoDock-Ligand docking method that combines template-based modeling and the GNINA scoring function, a Convolutional Neural Network-based scoring function, for the ligand binding prediction in CASP15. Among the 21 targets, we obtained successful predictions in top 5 submissions for 14 targets and partially successful predictions for 4 targets. In particular, for the most complicated target, H1114, which contains 56 metal cofactors and small molecules, our docking method successfully predicted the binding of most ligands. Analysis of the failed systems showed that the predicted receptor protein presented conformational changes in the backbone and side chains of the binding site residues, which may cause large structural deviations in the ligand binding prediction. In summary, our hybrid docking scheme was efficiently adapted to the ligand binding prediction challenges in CASP15.

Drug discovery involves a crucial step of optimizing molecules with the desired structural groups. In the domain of computer-aided drug discovery, deep learning has emerged as a prominent technique in molecular modeling. Deep generative models, based on deep learning, play a crucial role in generating novel molecules when optimizing molecules. However, many existing molecular generative models have limitations as they solely process input information in a forward way. To overcome this limitation, we propose an improved generative model called BD-CycleGAN, which incorporates BiLSTM (bidirectional long short-term memory) and Mol-CycleGAN (molecular cycle generative adversarial network) to preserve the information of molecular input. To evaluate the proposed model, we assess its performance by analyzing the structural distribution and evaluation matrices of generated molecules in the process of structural transformation. The results demonstrate that the BD-CycleGAN model achieves a higher success rate and exhibits increased diversity in molecular generation. Furthermore, we demonstrate its application in molecular docking, where it successfully increases the docking score for the generated molecules. The proposed BD-CycleGAN architecture harnesses the power of deep learning to facilitate the generation of molecules with desired structural features, thus offering promising advancements in the field of drug discovery processes.

ABSTRACT The capsid protein (CA), an essential component of human immunodeficiency virus type 1 (HIV-1), represents an appealing target for antivirals. Small molecules targeting the CAI-binding cavity in the C-terminal domain of HIV-1 CA (CA CTD) confer potent antiviral activities. In this study, we report that a small molecule, protoporphyrin IX (PPIX), targets the HIV-1 CA by binding to this pocket. PPIX was identified via in vitro drug screening, using a homogeneous and time-resolved fluorescence-based assay. CA multimerization and a biolayer interferometry (BLI) assay showed that PPIX promoted CA multimerization and bound directly to CA. The binding model of PPIX to CA CTD revealed that PPIX forms hydrogen bonds with the L211and E212 residues in the CA CTD. Moreover, the BLI assay demonstrated that this compound preferentially binds to the CA hexamer versus the monomer. The superposition of the CAI CTD-PPIX complex and the hexameric CA structure suggests that PPIX binds to the interface formed by the NTD and the CTD between adjacent protomers in the CA hexamer via the T72 and E212 residues, serving as a glue to enhance the multimerization of CA. Taken together, our studies demonstrate that PPIX, a hexamer-targeted CA assembly enhancer, should be a new chemical probe for the discovery of modulators of the HIV-1 capsid assembly. IMPORTANCE CA and its assembled viral core play essential roles in distinct steps during HIV-1 replication, including reverse transcription, integration, nuclear entry, virus assembly, and maturation through CA–CA or CA–host factor interactions. These functions of CA are fundamental for HIV-1 pathogenesis, making it an appealing target for antiviral therapy. In the present study, we identified protoporphyrin IX (PPIX) as a candidate CA modulator that can promote CA assembly and prefers binding the CA hexamer versus the monomer. PPIX, like a glue, bound at the interfaces between CA subunits to accelerate CA multimerization. Therefore, PPIX could be used as a new lead for a CA modulator, and it holds potential research applications.

Persistent organic pollutants (POPs) are ubiquitous and bioaccumulative, posing potential and long-term threats to human health and the ecological environment. Quantitative structure–activity relationship (QSAR) studies play a guiding role in analyzing the toxicity and environmental fate of different organic pollutants. In the current work, five molecular descriptors are utilized to construct QSAR models for predicting the mean and maximum air half-lives of POPs, including specifically the energy of the highest occupied molecular orbital (HOMO_Energy_DMol3), a component of the dipole moment along the z-axis (Dipole_Z), fragment contribution to SAscore (SAscore_Fragments), subgraph counts (SC_3_P), and structural information content (SIC). The QSAR models were achieved through the application of three machine learning methods: partial least squares (PLS), multiple linear regression (MLR), and genetic function approximation (GFA). The determination coefficients (R2) and relative errors (RE) for the mean air half-life of each model are 0.916 and 3.489% (PLS), 0.939 and 5.048% (MLR), 0.938 and 5.131% (GFA), respectively. Similarly, the determination coefficients (R2) and RE for the maximum air half-life of each model are 0.915 and 5.629% (PLS), 0.940 and 10.090% (MLR), 0.939 and 11.172% (GFA), respectively. Furthermore, the mechanisms that elucidate the significant factors impacting the air half-lives of POPs have been explored. The three regression models show good predictive and extrapolation abilities for POPs within the application domain.

Optimizing the concentration of different functional peptides on a surface can be a complex process. Here, the authors report on the use of a click immobilization strategy to create gradients of two different functional peptides on a surface to screen different density functions for rapid optimization.

Nasopharyngeal carcinoma (NPC) is one of the most common head and neck cancers. Early diagnosis plays a critical role in the treatment of NPC. To aid diagnosis, deep learning methods can provide interpretable clues for identifying NPC from magnetic resonance images (MRI). To identify the optimal models, we compared the discrimination performance of hierarchical and simple layered convolutional neural networks (CNN). Retrospectively, we collected the MRI images of patients and manually built the tailored NPC image dataset. We examined the performance of the representative CNN models including shallow CNN, ResNet50, ResNet101, and EfficientNet-B7. By fine-tuning, shallow CNN, ResNet50, ResNet101, and EfficientNet-B7 achieved the precision of 72.2%, 94.4%, 92.6%, and 88.4%, displaying the superiority of deep hierarchical neural networks. Among the examined models, ResNet50 with pre-trained weights demonstrated the best classification performance over other types of CNN with accuracy, precision, and an F1-score of 0.93, 0.94, and 0.93, respectively. The fine-tuned ResNet50 achieved the highest prediction performance and can be used as a potential tool for aiding the diagnosis of NPC tumors.

COVID-19 can cause different neurological symptoms in some people, including smell, inability to taste, dizziness, confusion, delirium, seizures, stroke, etc. Owing to the issue of vaccine effectiveness, update and coverage, we still need one or more diversified strategies as the backstop to manage illness. Characterizing the structural basis of ligand recognition in the main protease (Mpro) of SARS-CoV-2 will facilitate its rational design and development of potential drug candidates with high affinity and selectivity against COVID-19. Up to date, covalent-, non-covalent inhibitors and allosteric modulators have been reported to bind to different active sites of Mpro. In the present work, we applied the molecular dynamics (MD) simulations to systematically characterize the potential binding features of catalytic active site and allosteric binding sites in Mpro using a dataset of 163 3D structures of Mpro-inhibitor complexes, in which our results are consistent with the current studies. In addition, umbrella sampling (US) simulations were used to explore the dissociation processes of substrate pathway and allosteric pathway. All the information provided new insights into the protein features of Mpro and will facilitate its rational drug design for COVID-19.

The application of deep learning in the field of drug discovery brings the development and expansion of molecular generative models along with new challenges in this field. One of challenges in de novo molecular generation is how to produce new reasonable molecules with desired pharmacological, physical, and chemical properties. To improve the similarity between the generated molecule and the starting molecule, we propose a new molecule generation model by embedding Long Short-Term Memory (LSTM) and Attention mechanism in CycleGAN architecture, LA-CycleGAN. The network layer of the generator in CycleGAN is fused head and tail to improve the similarity of the generated structure. The embedded LSTM and Attention mechanism can overcome long-term dependency problems in treating the normally used SMILES input. From our quantitative evaluation, we present that LA-CycleGAN expands the chemical space of the molecules and improves the ability of structure conversion. The generated molecules are highly similar to the starting compound structures while obtaining expected molecular properties during cycle generative adversarial network learning, which comprehensively improves the performance of the generative model.

The accurate predicting of physical properties and bioactivity of drug molecules in deep learning depends on how molecules are represented. Many types of molecular descriptors have been developed for quantitative structure-activity/property relationships quantitative structure-activity relationships (QSPR). However, each molecular descriptor is optimized for a specific application with encoding preference. Considering that standalone featurization methods may only cover parts of information of the chemical molecules, we proposed to build the conjoint fingerprint by combining two supplementary fingerprints. The impact of conjoint fingerprint and each standalone fingerprint on predicting performance was systematically evaluated in predicting the logarithm of the partition coefficient (logP) and binding affinity of protein-ligand by using machine learning/deep learning (ML/DL) methods, including random forest (RF), support vector regression (SVR), extreme gradient boosting (XGBoost), long short-term memory network (LSTM), and deep neural network (DNN). The results demonstrated that the conjoint fingerprint yielded improved predictive performance, even outperforming the consensus model using two standalone fingerprints among four out of five examined methods. Given that the conjoint fingerprint scheme shows easy extensibility and high applicability, we expect that the proposed conjoint scheme would create new opportunities for continuously improving predictive performance of deep learning by harnessing the complementarity of various types of fingerprints.

Abstract Hospital‐acquired infection causes many deaths worldwide and calls for the urgent need for antibacterial biomaterials used in clinic that can selectively kill harmful bacteria. The present study rationally designs fusion peptides capable of undergoing 2D self‐assembly on the poly(methyl methacrylate) surface to form a smart surface, which can maintain a desirable orientation via electrostatic interactions. The in vitro assay shows that the smart surface can recognize bacteria to exert antibacterial activity and is nontoxic toward mouse bone mesenchymal stem cells. Excitingly, the smart surface can distinguish different bacterial strains. This selective feature, from being broad‐spectrum to being highly selective against S. aureus, can be altered by varying the number of amino acids in the recognition sequences. By all‐atom molecular dynamics simulations, it is also found that the recognition sequence in the peptide is critical for the selectivity toward specific bacterial strains, in which a less accessible surface area for the bacteria in the antimicrobial peptide sequence is responsible for such selectivity. Finally, the smart surface can inhibit S. aureus infection in vivo with much more rapid tissue‐healing compared to the control.

Although energy barriers can be efficiently crossed in the reaction coordinate (RC) guided sampling, this type of method suffers from identification of the correct RCs or requirements of high dimensionality of the defined RCs for a given system. If only the approximate RCs with significant barriers are used in the simulations, hidden energy barriers with small to medium height would exist in other degrees of freedom (DOFs) relevant to the target process and consequently cause the problem of insufficient sampling. To address the sampling in this so-called hidden barrier situation, here we propose an effective approach to combine temperature accelerated molecular dynamics (TAMD), an efficient RC-guided sampling method, with the integrated tempering sampling (ITS), a generalized ensemble sampling method. In this combined ITS-TAMD method, the sampling along the major RCs with high energy barriers is guided by TAMD and the sampling of the rest of the DOFs with lower but not negligible barriers is enhanced by ITS. The performance of ITS-TAMD to three systems in the processes with hidden barriers has been examined. In comparison to the standalone TAMD or ITS approach, the present hybrid method shows three main improvements. (1) Sampling efficiency can be improved at least five times even if in the presence of hidden energy barriers. (2) The canonical distribution can be more accurately recovered, from which the thermodynamic properties along other collective variables can be computed correctly. (3) The robustness of the selection of major RCs suggests that the dimensionality of necessary RCs can be reduced. Our work shows more potential applications of the ITS-TAMD method as the efficient and powerful tool for the investigation of a broad range of interesting cases. [ABSTRACT FROM AUTHOR]

Canonical molecular dynamics simulations are performed to investigate the behavior of single-chain and multiple-chain poly(ethylene glycol) (PEG) contained within a cubic framework spanned by polyethylene (PE) chains. This simple model is the first of its kind to study the chemical physics of polymer-threaded organic frameworks, which are materials with potential applications in catalysis and separation processes. For a single-chain 9-mer, 14-mer, and 18-mer in a small framework, the PEG will interact strongly with the framework and assume a more linear geometry chain with an increased radius of gyration Rg compared to that of a large framework. The interaction between PEG and the framework decreases with increasing mesh size in both vacuum and water. In the limit of a framework with an infinitely large cavity (infinitely long linkers), PEG behavior approaches simulation results without a framework. The solvation of PEG is simulated by adding explicit TIP3P water molecules to a 6-chain PEG 14-mer aggregate confined in a framework. The 14-mer chains are readily solvated and leach out of a large 2.6 nm mesh framework. There are fewer water-PEG interactions in a small 1.0 nm mesh framework, as indicated by a smaller number of hydrogen bonds. The PEG aggregate, however, still partially dissolves but is retained within the 1.0 nm framework. The preliminary results illustrate the effectiveness of the simple model in studying polymer-threaded framework materials and in optimizing polymer or framework parameters for high performance. [ABSTRACT FROM AUTHOR]