Computational biology and chemistry最新文献

Binding affinity prediction has been considered as a fundamental task in drug discovery. Despite much effort to improve accuracy of binding affinity prediction, the prior work considered only macro-level features that can represent the characteristics of the whole architecture of a drug and a target protein, and the features from local structure of the drug and the protein tend to be lost. In this paper, we propose a deep learning model that can comprehensively extract the local features of both a drug and a target protein for accurate binding affinity prediction. The proposed model consists of two components named as Multi-Stream CNN and Multi-Stream GCN, each of which is responsible for capturing micro-level characteristics or local features from subsequences of a target protein sequence and subgraph of a drug molecule, respectively. Having multiple streams consisting of different numbers of layers, both the components can compute and preserve the local features with a stream consisting of a single layer. Our evaluation with two popular datasets, Davis and KIBA, demonstrates that the proposed model outperforms all the baseline models using the global features, implying that local features play significant roles of binding affinity prediction.

Thyroid cancer includes papillary thyroid carcinoma (PTC) and anaplastic thyroid carcinoma (ATC). While PTC has an excellent prognosis, ATC has a dismal prognosis, necessitating the identification of novel targets in ATC to aid in ATC diagnosis and treatment. Therefore, we analyzed ATC single-cell RNA sequencing (scRNA-seq) and bulk RNA sequencing (bulk RNA-seq) data from the Gene Expression Omnibus (GEO), retrieved immune-related genes from the ImmPort database, and identified differentially expressed immune genes within single-cell subgroups. The AUCell package in R was used to calculate activity scores for single-cell subgroups and identify active cell populations. Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analysis were performed on differentially expressed genes (DEGs) in active cell populations. Then, we integrated thyroid-cancer scRNA-seq and bulk RNA-seq data to identify overlapping DEGs. Relevant transcription factors (TFs) were retrieved from the TRRUST database. A protein-protein interaction (PPI) network for key TFs was created using the STRING database. Simultaneously, drugs associated with key TFs were obtained from DGIdb. ScRNA-seq cluster analysis showed that T/natural killer (NK) cells were more distributed in ATC and that thyrocytes cells were more distributed in PTC. We obtained 264 differential immune genes (DIGs) from the IMMPORT database and integrated scRNA-seq cluster analysis to identify the active cell T/NK cells and myeloid cells. Integrated bulk RNA-seq analysis obtained common immune genes (CIGs) such as TMSB4X, NFKB1, TNFRSF1B, and B2M. The nine CIG-related TFs (CEBPB, SPI1, NFKB1, RUNX1, NFE2L2, REL, CIITA, KLF6, and CEBPD) in myeloid cells and three TFs (NFKB1, FOXO1, and NR3C1) in T/NK cells were obtained from the TRRUST database. The key genes we identified represent potential targets for treating ATC.

Drug Repurposing gives us facility to find the new uses of previously developed drugs rather than developing new drugs from start. Particularly during pandemic, drug repurposing caught much attention to provide new applications of the previously approved drugs. In our research, we provide a novel method for drug repurposing based on feature learning process from drug-disease-gene network. In our research, we aimed at finding drug candidates which can be repurposed under neurodegenerative diseases and glioma. We collected association data between drugs, diseases and genes from public resources and primarily examined the data related to Alzheimer's, Parkinson's and Glioma diseases. We created a Knowledge Graph using neo4j by integrating all these datasets and applied scalable feature learning algorithm known as node2vec to create node embeddings. These embeddings were later used to predict the unknown associations between disease and their candidate drugs by finding cosine similarity between disease and drug nodes embedding. We obtained a definitive set of candidate drugs for repurposing. These results were validated from the literature and CodReS online tool to rank the candidate drugs. Additionally, we verified the status of candidate drugs from pharmaceutical knowledge databases to confirm their significance.

Development of novel inhibitors is necessary to counteract the rising prevalence of breast cancer (BC) in women in recent years, as evidenced by the side-effect profiles of a few clinically approved inhibitors. In this study, the usnic acid derivative (UA1) was synthesized due to the effectiveness of usnic acid (UA) against BC cell line. Furthermore, the structure of synthesized compound was determined using FT-IR, ¹H NMR, ¹³C NMR, HSQC, and HMBC spectroscopic techniques. The anticancer potential of UA1 was assessed using the MTT assay on two different cell lines of BC including MCF7 and T47D. To ascertain the binding affinity and stability of the docking complex, further procedures included the in silico molecular docking, molecular dynamic simulation, principal component analysis, and binding free energy experiments. The cytotoxicity results show that the UA1 exhibits strong antitumor activities and comparable effects against BC cell lines with the IC₅₀ values of 9.21 µM for MCF7 cell and 14.8 µM for T47D cell, respectively, where the positive control cisplatin showed the IC₅₀ values of 8.95 µM for MCF7 cell and 10.9 µM for T47D cell. Additionally, the molecular docking results of UA1 showed that it interacts strongly into the active site of target protein. Molecular dynamics simulation results also revealed that the docking complex was formed stability with the RMSD and RMSF values of 0.50 nm and 0.19 nm, respectively. According to the PCA analysis, the target protein displays good conformational space behaviour when bound with UA1. Furthermore, the UA1 showed the free binding energy value of -18.52 kcal/mol with the target protein, which indicating that UA1 may prevent BC.

This study explores the use of machine learning algorithms, including traditional approaches and graph neural networks (GNNs), to predict certain diseases by analyzing protein-protein interactions. Protein-protein interactions (PPIs) are complex, multifaceted, and sometimes ever-changing. Therefore, analyzing PPIs and making predictions based on them present significant challenges to traditional computational techniques. However, machine learning, particularly GNNs, with their powerful ability to identify complex patterns within large, convoluted datasets, emerge as compelling and revolutionary tools for unraveling these intricate biological networks. We apply machine learning, aided by SHAP explainability and GNNs, on three networks of distinct sizes, ranging from small to large. While the ML results highlight the higher importance of network features in prediction, GNNs exhibit superior accuracy.

The ten papers in this special issue were presented at the 21th Asia Pacific Bioinformatics Conference (APBC), which was held in Changsha, Hunan, PR China, Apr. 14-16, 2023.

In this study 1-vinyl-3-alkyl imidazolium-based ionic liquid monomers (ILs) with different alkyl chain lengths {R = hexyl (A), octyl (B) and decyl (C)} have been synthesized for antibacterial applications. The prepared ILs have been characterized using UV, FT-IR and NMR spectroscopy. The antibacterial activities of the synthesized ILs against Staphylococcus aureus (S. aureus) and Escherichia coli (E. coli) have been examined by measuring their minimal inhibitory concentrations (MICs) and minimum bactericidal concentrations (MBCs). The results exhibit that these ILs have admirable antibacterial activities with MIC values range from < 1.2 to 12.2 μM for S. aureus and < 2.4 to 12.2 μM for E. coli. A notable dependence of antibacterial and antibiofilm efficacy on the alkyl chain length (ILC> ILB > ILA) has been observed. From in-silico evaluation, the binding energies of β-lactamase protein of S. aureus (PDB ID: 1GHP) are found to be -4.4, -4.6, -4.7 kcal/mol for IL A, IL B, and IL C. For dihydrofolate reductase (DHFR) of S. aureus and E. coli the binding energies -4.6, -4.5, -5.3 kcal/mol and -5.3, -5.4, -5.6 kcal/mol have been noted for IL A, IL B, and IL C respectively. MD simulations (100 ns) have been performed to predict the stability and understand the binding mechanism of the docked complexes.

The viruses has spread globally and have been impacted lives of people socially and economically, which causes immense suffering throughout the world. Thousands of people died and millions of illnesses were brought, by the outbreak worldwide. In order to control the coronavirus pandemic, mathematical modeling proved to be an invaluable tool for analyzing and determining the potential and severity of the illness. This work proposed and assessed a deterministic six-compartment model with a novel stochastic neural network. The significance of the proposed model was demonstrated by numerical simulation in which the results are agreed with sensitivity analysis. Furthermore, the efficacy of stochastic neural network has been proven with the help of numerical simulations. Some investigations have been conducted through graphs and tables that how the vaccination process is helpful to minimize stress in society. The numerical simulations also focused on preventing the community-wide spread of the disease. The lowest residual errors have been achieved by our proposed stochastic neural network and compared with numerical solvers to assess the accuracy and robustness of the proposed approach.

Influenza A virus is a leading cause of acute respiratory tract infections, posing a significant global health threat. Current treatment options are limited and increasingly ineffective due to viral mutations. This study aimed to identify potential drug candidates targeting the nucleoprotein of the H3N2 subtype of Influenza A virus. We focused on epicatechin derivatives and employed a series of computational approaches, including ADMET profiling, drug-likeness evaluation, PASS predictions, molecular docking, molecular dynamics simulations, Principal Component Analysis (PCA), dynamic cross-correlation matrix (DCCM) analyses, and free energy landscape assessments. Molecular docking and dynamics simulations revealed strong and stable binding interactions between the derivatives and the target protein, with complexes 01 and 81 exhibiting the highest binding affinities. Additionally, ADMET profiling indicated favorable pharmacokinetic properties for these compounds, supporting their potential as effective antiviral agents. Compound 81 demonstrated exceptional quantum chemical descriptors, including a small HOMO-LUMO energy gap, high electronegativity, and significant softness, suggesting high chemical reactivity and strong electron-accepting capabilities. These properties enhance Compound 81's potential to interact effectively with the H3N2 nucleoprotein. Experimental validation is strongly recommended to advance these compounds toward the development of novel antiviral therapies to address the global threat of influenza.

LIMK2 is crucial in regulating actin cytoskeleton dynamics, significantly contributing to cancer cell proliferation, invasion, and metastasis. Inhibitors like LIMKi3 effectively suppress LIMK2 kinase activity by directly affecting actin polymerization and preventing the formation of structures like filopodia and lamellipodia, which are typical of motile cancer cells. By modulating these actin dynamics, LIMKi3 inhibits cancer cell migration and invasion, reducing the potential for metastasis. Thus, this study aims to explore potential anti-cancer therapeutic LIMK2 inhibitors with properties resembling LIMKi3. Henceforth, molecular docking was utilized in this study to comprehend the ATP mimetic binding mode of LIMKi3, followed by Pharmacophore-based virtual screening to identify small molecules resembling LIMKi3. In addition, molecular dynamics simulations were performed to explore the dynamic behavior of LIMK2 and potential inhibitors. Further, network analysis and binding free energy calculations were implemented to comprehensively assess the interactions between the compounds and LIMK2. In molecular docking, LIMKi3 demonstrated an ATP mimetic hinge binding mode with hydrogen bonds at Ile408. Among the screened compounds (NCI300395, ChemDiv-8020-2508, and ChemDiv-7997-0024), three displayed "ADRH" pharmacophoric features like LIMKi3, with favorable ADMET properties, higher binding affinity, and significant hydrogen bond interactions at Ile408. LIMK2-inhibitor complexes showed lower RMSD than LIMK2-LIMKi3, indicating higher equilibrium by identified compounds. Protein-drug Complexes exhibited significant inter-domain correlation in N-lobe residues of LIMK2, including conserved β3, αC, and Hinge residues. Binding free energy analysis ranked LIMK2-NCI300395 highest, followed by LIMK2-ChemDiv-7997-0024 and LIMK2-ChemDiv-8020-2508, highlighting their potential as effective LIMK2-targeting compounds. Hence, this study emphasizes LIMKi3's significance and identifies potential candidates (NCI300395, ChemDiv-7997-0024, and ChemDiv-8020-2508) for developing cancer therapeutics targeting LIMK2. These findings open avenues for further investigations into the complex interplay between cytoskeletal dynamics and cancer progression.