Traditional drug discovery is a resource-intensive process with high attrition rates and the huge difficulty of working with a chemical space that is thought to include [Formula: see text] molecules. Even though computational chemistry has come a long way, traditional generative models still use string-based representations like SMILES, which have trouble capturing intricate three-dimensional spatial interactions and often make structures that aren't real. Moreover, current reinforcement learning methodologies frequently do not achieve an equilibrium between molecular diversity and high-affinity biological activity. To overcome these constraints, this research introduces an innovative integrated framework that merges Geometric Multi-Discrete Soft Actor-Critic (Geom-SAC) and Multi-stage Variational Autoencoders (MS-VAE) to improve de novo molecular creation and activity optimisation. The main new idea is the combination of geometric deep learning, which enforces physical atomic restrictions, and a hierarchical VAE architecture, which organises the latent space into manageable structural steps from scaffold formation to functional group optimisation. We also use a Non-Covalent Interaction-Aware (NCIA) graph neural network in our method to improve protein-ligand affinity predictions by simulating complex intermolecular forces. Experimental results on benchmark datasets, such as ZINC250k and PDBbind, show that the proposed framework improves binding affinity scores by 15% and the Valid-Unique-Novel (VUN) molecule ratio by 20% compared to the best existing methods. Also, adding a security layer based on blockchain technology makes sure that data is secure and can be tracked. This all-encompassing method provides a strong, highly accurate answer for next-generation AI-driven pharmacology. It greatly narrows the gap between computational design and experimental validation.

Lysophosphatidic acid receptor 2 (LPAR2), a G protein-coupled receptor, has been implicated in the progression of fibrosis and is therefore a promising novel drug target for the treatment of fibrosis and related diseases. In this paper, a reliable homology model of LPAR2 was obtained by using three templates (PDB IDs: 4Z34, 7TD0, and 7VIE) and evaluations. A new binding site for a series of selective LPAR2 inhibitors were identified through molecular docking with the reference compound 50. Subsequently, a three-dimensional quantitative structure-activity relationship (3D QSAR) analysis was conducted on a series of N-sulfonyl heterocyclic antagonists of LPAR2. The derived optimal CoMFA model (q2 = 0.792, r2 = 0.999, [Formula: see text] = 0.998, [Formula: see text] = 0.978) and CoMSIA model (q2 = 0.713, r2 = 0.996, [Formula: see text] = 0.978, [Formula: see text] = 0.958) demonstrated strong statistical robustness and high external predictability. The 3D contour maps generated from these models were analyzed and compared with the binding mode of the reference compound. This provided insights into the structural requirements of these LPAR2-selective inhibitors. Furthermore, the predictive capability of these models was validated by accurately predicting the antagonistic activities of other types of LPAR2-selective inhibitors (CoMFA-SE, [Formula: see text] = 0.862; CoMSIA-SEHDA, [Formula: see text] = 0.934), confirming the robustness of the optimal 3D QSAR models. The new binding site and the optimal 3D QSAR models will be helpful to design novel molecules and predict their inhibitory activity against LPAR2.

Pseudomonas aeruginosa is increasingly becoming resistant to multiple drugs and is held responsible for high rate of mortality and morbidity across the globe. This study aims to examine proteins from outer membrane such as OprE, OprF, OprC, and OprG, for the multi-epitope vaccine design. In this article, the prediction of epitopes for helper T-cells, cytotoxic T-cells, and B-cell epitopes was carried out by the application of various immune-informatics tools. All the predicted epitopes were aligned and assembled into a peptide sequence along with linkers and adjuvant sequences. Further, secondary structure and three-dimensional structure were predicted for the multiepitope construct. The vaccine construct designs were evaluated and validated for allergenicity, toxicity, and antigenicity. The validation of the predicted structure was carried out by determining its physiochemical properties by the application of ProtParam. Protein docking and molecular-dynamic simulation confirmed strong and stable interaction between the vaccine construct and Toll-like receptor-4. The vaccine construct was cloned into pET-29b vector and expressed in Escherichia coli. The designed multiepitope vaccine construct was overexpressed and purified by the application of Ni-NTA affinity chromatography and subjected to SDS-PAGE analysis. The circular dichroism spectroscopy analysis revealed it to be stable and structured. The hemolysis assay demonstrated minimal RBC toxicity suggesting that it was safe to use. The designed vaccine construct could activate both humoral and cellular immune responses as demonstrated by the advanced immunoinformatic approach making it a promising vaccine construct for protection against P. aeruginosa.

Inhibition of carbohydrate digesting enzymes like pancreatic α-amylase has proven to be an effective strategy in countering postprandial hyperglycaemia. However, unwanted adverse effects associated with current therapies, such as ACB, voglibose, and miglitol, have necessitated the development of safer and more effective alternatives. These challenges have prompted growing interest in natural product scaffolds, particularly phenolic compounds, which exhibit favourable safety profiles and multifaceted enzyme inhibition. In this context, the strategic design of molecular conjugates integrating bioactive phenolics offers a promising route to enhance inhibitory potency, binding specificity, and therapeutic relevance. Reportedly, molecular hybrids with chalcone, phenolic acid, coumarin and polyphenol backbone exhibited impressive antioxidant and potent inhibition against digestive enzymes. In the quest of developing novel anti-hyperglycaemic agents, we report in vitro and in silico evaluation of the novel TPAC as potential pancreatic α-amylase inhibitors in the present work. Among the ten conjugates, T5 (IC50 = 50.65 ± 0.76 µM) exhibited strong inhibition against PPA which was comparable to positive control ACB (24.81 ± 0.98 µM). Computational binding analysis revealed binding of T5 to HPA (PDB ID: 2QV4) by interacting with the amino acid residues and distorting the receptor's catalytic site conformation. Furthermore, the conformational dynamic studies and electron density driven simulations established the stability and high reactivity of T5 within the ligand-receptor complex. The in silico studies corroborated the in vitro enzyme inhibition results, reinforcing the mechanistic insights into ligand-receptor interactions. Taken together, the experimental and computational results indicate that T5 merits further investigation as a candidate molecule targeting pancreatic α-amylase for the management of T2D .

Calcific aortic valve disease (CAVD) is driven by immune–metabolic dysregulation and osteogenic remodeling, yet actionable molecular targets remain limited. This study integrated multimodal computational analyses with experimental validation to identify functional biomarkers and therapeutic candidates. Single-nucleus RNA sequencing of human aortic valves was combined with CellChat, Monocle 3 and consensus NMF to define immune–stromal interactions and gene programs. A 10 gene classifier was developed using multi model machine learning and interpreted with SHAP. Cross cohort validation, immune deconvolution and network enrichment were applied. Network pharmacology, docking and 50 ns MD simulations were used to identify ligand candidates. Experimental validation was performed in human VICs using osteogenic stimulation, siRNA knockdown and small molecule treatments, followed by RT qPCR, Western blot, Alizarin Red S staining and ALP and proliferation assays. Computational analyses consistently prioritized HLA DRA and FTL as stable, high AUC biomarkers linked to macrophage rich and MHC II enriched microenvironments. Network analyses anchored HLA DRA to adaptive immunity and FTL to iron metabolic pathways. Docking and MD identified Chrysin, α Tocopherol, Colchicine and Diallyl trisulfide as structurally compatible ligands. In vitro, osteogenic stimulation markedly upregulated HLA DRA, FTL, RUNX2 and ALPL. siRNA knockdown of HLA DRA or FTL reduced mineral deposition, ALP activity and VIC proliferation. Predicted small molecule ligands decreased HLA DRA and FTL expression and attenuated calcification. HLA DRA and FTL function as immune–metabolic drivers of VIC osteogenic remodeling and represent viable therapeutic targets. This integrated systems to experimental framework highlights natural compound candidates with potential relevance for CAVD therapy. The study employed a stepwise multiomic workflow beginning with snRNA-seq of human calcified aortic valve tissue for cell-type clustering and ligand–receptor signaling inference. Monocle 3 trajectory reconstruction and cNMF gene-program extraction delineated dynamic transcriptional states and identified 12 pseudotime-associated genes. A machine-learning framework integrating 12 algorithms and SHAP interpretability analysis prioritized FTL, HLA-DRA, CXCL8, C1QA, and FTH1 as the top predictive markers. Functional enrichment, immune infiltration, and protein–protein interaction analyses revealed immune–metabolic cross-talk within the valve microenvironment. Subsequent network pharmacology and molecular docking integrated TCMBank screening, identifying Chrysin, α-Tocopherol, Colchicine, and Diallyl Trisulfide as high-affinity ligands targeting HLA-DRA and FTL. Docking and 50 ns MD simulations confirmed the dynamic stability of these interactions. Experimental validation was performed in human VICs using osteogenic stimulation, siRNA knockdown and small molecule treatments followed by RT qPCR, Western blot, Alizarin Red S staining and ALP and proliferation assays.

Massively-multitask regression models (MMRMs) have revolutionized activity prediction for drug discovery. MMRMs trained on millions of compounds and many thousands of assays can predict bioactivity with accuracy comparable to 4-concentration IC50 experiments. This report compares six MMRMs: pQSAR, Alchemite, MT-DNN, MetaNN, Macau and IMC. Models were trained by experts in each method, on identical sets of 159 kinase and 4276 diverse ChEMBL assays, employing realistically novel training/test set splits. Results were compared both qualitatively and with statistical rigor. Our use-case is imputing full bioactivity profiles for the very sparse compound collections on which the models were trained. MMRMs performed much better than the single-task random forest regression (ST-RFR) model. Five MMRMs train all models simultaneously, so must leave out test-set measurements from all assays to avoid leakage (here 25% of data), whereas one method trains models one-at-a-time, so only holds out test data for that assay (< 1% of data). Thus, all algorithms were compared both using 75/25 splits, and when possible, 99 + / < 1 splits. Many MMRM evaluations achieved similar accuracy when tested on the same split. However, when evaluated on 75/25 splits, all MMRMs performed much worse than when evaluated on 99 + / < 1% splits. Thus, while many MMRMs produce comparable final production models (trained on all the data), models that require 75/25 splits greatly underestimate the accuracy of the final models. While outstanding for imputations, MMRMs proved little better than ST-RFR for compounds very unlike the training collection. Thus, MMRMs are best for hit-finding, off-target, promiscuity, MoA, polypharmacology or drug-repurposing within the training collection. Since accuracy is not a deciding factor, other pros and cons of each method are also described.