Comput Biol Med. 2025 May 12;192(Pt B):110249. doi: 10.1016/j.compbiomed.2025.110249. Online ahead of print.

ABSTRACT

MOTIVATION: Accurately predicting drug-target interactions (DTIs) is critical for accelerating drug discovery, repositioning, and development. Traditional experimental methods are often expensive and time-consuming, emphasizing the need for efficient computational models to streamline these processes. To address this challenge, we developed MiRAGE-DTI, a novel computational framework that integrates diverse drug and target similarity measures with robust machine learning techniques to achieve superior predictive accuracy.

RESULTS: MiRAGE-DTI introduces a novel framework that integrates structural, functional, and interaction-based features into a unified model. Leveraging the strengths of Random Forest classifiers, MiRAGE-DTI ensures robust and interpretable predictions while addressing challenges such as class imbalance and data variability. Comprehensive evaluations across multiple benchmark datasets, including GPCR, IC, NR, and Enzyme, reveal that MiRAGE-DTI consistently outperforms state-of-the-art algorithms such as DTI-CNN, GCNMDA, MVGCN, MMGCN, GraphCDA, DTINet, and MIDTI. It achieves significant improvements in key metrics such as AUROC, AUPR, and accuracy. To validate its predictions, molecular docking studies were conducted, highlighting strong binding affinities for key drug-target interactions, including Oxandrolone-PGR and Metyrapone-CYP2E1, which demonstrate high therapeutic potential. Beyond predictive accuracy, MiRAGE-DTI has practical implications in drug repositioning and personalized medicine, as well as the potential to streamline preclinical workflows. Its ability to uncover novel drug-target relationships enhances the understanding of molecular mechanisms underlying diseases, paving the way for innovative therapeutic strategies. Furthermore, by predicting off-target interactions and potential side effects, MiRAGE-DTI contributes to improving drug safety profiles and regulatory evaluations.

CONCLUSION: MiRAGE-DTI represents a versatile and powerful tool for advancing drug discovery and development. Its ability to identify novel therapeutic opportunities, repurpose existing drugs, and enable precision medicine highlights its transformative potential in tackling unmet medical needs. The framework's robust performance, validated predictions, and wide-ranging practical applications position MiRAGE-DTI as a critical resource for modern drug discovery.

PMID:40359678 | DOI:10.1016/j.compbiomed.2025.110249

Nat Commun. 2025 May 13;16(1):4449. doi: 10.1038/s41467-025-59478-8.

ABSTRACT

Artificial intelligence applications in oncology imaging often struggle with diagnosing rare tumors. We identify significant gaps in detecting uncommon thyroid cancer types with ultrasound, where scarce data leads to frequent misdiagnosis. Traditional augmentation strategies do not capture the unique disease variations, hindering model training and performance. To overcome this, we propose a text-driven generative method that fuses clinical insights with image generation, producing synthetic samples that realistically reflect rare subtypes. In rigorous evaluations, our approach achieves substantial gains in diagnostic metrics, surpasses existing methods in authenticity and diversity measures, and generalizes effectively to other private and public datasets with various rare cancers. In this work, we demonstrate that text-guided image augmentation substantially enhances model accuracy and robustness for rare tumor detection, offering a promising avenue for more reliable and widespread clinical adoption.

PMID:40360460 | DOI:10.1038/s41467-025-59478-8

Nat Genet. 2025 May 13. doi: 10.1038/s41588-025-02189-z. Online ahead of print.

ABSTRACT

Obsessive-compulsive disorder (OCD) affects ~1% of children and adults and is partly caused by genetic factors. We conducted a genome-wide association study (GWAS) meta-analysis combining 53,660 OCD cases and 2,044,417 controls and identified 30 independent genome-wide significant loci. Gene-based approaches identified 249 potential effector genes for OCD, with 25 of these classified as the most likely causal candidates, including WDR6, DALRD3 and CTNND1 and multiple genes in the major histocompatibility complex (MHC) region. We estimated that ~11,500 genetic variants explained 90% of OCD genetic heritability. OCD genetic risk was associated with excitatory neurons in the hippocampus and the cortex, along with D1 and D2 type dopamine receptor-containing medium spiny neurons. OCD genetic risk was shared with 65 of 112 additional phenotypes, including all the psychiatric disorders we examined. In particular, OCD shared genetic risk with anxiety, depression, anorexia nervosa and Tourette syndrome and was negatively associated with inflammatory bowel diseases, educational attainment and body mass index.

PMID:40360802 | DOI:10.1038/s41588-025-02189-z

Nat Metab. 2025 May 13. doi: 10.1038/s42255-025-01308-8. Online ahead of print.

NO ABSTRACT

PMID:40360757 | DOI:10.1038/s42255-025-01308-8

Korean J Intern Med. 2025 May;40(3):458-467. doi: 10.3904/kjim.2024.244. Epub 2025 Apr 30.

ABSTRACT

BACKGROUND/AIMS: Patients with acute exacerbation of idiopathic pulmonary fibrosis (AE-IPF) typically have a poor prognosis; however, no effective treatment is available. In recent years, several retrospective studies have suggested the clinical benefits of direct hemoperfusion with a polymyxin B-immobilized fiber column (PMX-DHP) in patients with AE-IPF. Herein, we aimed to investigate the efficacy and safety of PMX-DHP treatment in patients with AE-IPF.

METHODS: Patients diagnosed with AE-IPF (n = 10) with a partial pressure of oxygen to fraction of inspiratory oxygen ratio (P/F ratio) > 100 were prospectively enrolled at a single center. PMX-DHP was performed twice for 6 hours (at 24-h intervals) at a flow rate of 80-100 mL/min, and steroid pulse therapy was concurrently administered (500 mg of methylprednisolone for 3 d).

RESULTS: The mean patient age was 67 years, and 80.0% were male. During the follow-up (median, 42.5 d; interquartile range, 16.0-174.0 d), seven (70.0%) patients died (including two who underwent transplantation); the in-hospital mortality rate was 70%, while the 30- and 90-day mortality rates were 50.0% and 70.0%, respectively. After 48 hours of PMX-DHP treatment, the P/F ratio improved (mean, 160.0 vs. 229.0; p = 0.054) and C-reactive protein level decreased (mean, 8.3 mg/dL vs. 3.5 mg/dL; p = 0.012). During hospitalization, no PMX-DHP-associated adverse events were observed.

CONCLUSION: Our results suggest that PMX-DHP treatment may be useful at improving oxygenation and reducing inflammation in patients with AE-IPF with acceptable safety profiles, however without affecting their prognosis.

PMID:40360222 | DOI:10.3904/kjim.2024.244

Korean J Intern Med. 2025 May;40(3):345-346. doi: 10.3904/kjim.2025.111. Epub 2025 Apr 30.

NO ABSTRACT

PMID:40360217 | DOI:10.3904/kjim.2025.111

J Pharmacol Exp Ther. 2025 Apr 21;392(6):103588. doi: 10.1016/j.jpet.2025.103588. Online ahead of print.

ABSTRACT

Idiopathic pulmonary fibrosis (IPF) is characterized by the pathological replacement of alveolar structures with thickened, inelastic fibrous tissue, which significantly hinders gas exchange in the lungs. Disulfiram (DSF), a Food and Drug Administration-approved drug for alcohol dependence, has shown potential in various diseases. This study investigates the effects of DSF on IPF and its mechanisms, focusing on the cyclooxygenase-2 (COX-2)/prostaglandin E2 (PGE2) pathway. Utilizing primary diseased human lung fibroblast-IPF cells and A549 cells induced with transforming growth factor-beta 1 to model epithelial-mesenchymal transition (EMT), we employed a battery of in vitro assays to assess cellular viability, migratory capacity, and the expression of fibrosis-related genes and proteins. To further substantiate our in vitro findings, a bleomycin-induced mouse model of IPF was treated with DSF, and subjected to a comprehensive evaluation of pulmonary function, histological examination, hydroxyproline assay, and western blot analysis to quantify the extent of fibrosis. DSF reduced cell viability and migration in fibrotic cell models. It increased COX-2 and PGE2 levels, regulated EMT, and extracellular matrix collagen deposition. In vivo, DSF improved pulmonary function and reduced EMT and extracellular matrix accumulation in mice. The COX-2/PGE2 axis was identified as a critical mediator of DSF's effects. DSF exhibits antifibrotic properties in IPF by modulating the COX-2/PGE2 signaling pathway. This study provides a novel therapeutic strategy for IPF and highlights the potential of repurposing DSF for clinical use in this context. SIGNIFICANCE STATEMENT: Disulfiram shows promise in treating idiopathic pulmonary fibrosis by targeting the cyclooxygenase-2/prostaglandin E2 pathway, offering a new therapeutic strategy and highlighting its potential for repurposing in this context.

PMID:40359874 | DOI:10.1016/j.jpet.2025.103588

Nat Methods. 2025 May 13. doi: 10.1038/s41592-025-02696-1. Online ahead of print.

ABSTRACT

The subcellular localization of a protein is important for its function, and its mislocalization is linked to numerous diseases. Existing datasets capture limited pairs of proteins and cell lines, and existing protein localization prediction models either miss cell-type specificity or cannot generalize to unseen proteins. Here we present a method for Prediction of Unseen Proteins' Subcellular localization (PUPS). PUPS combines a protein language model and an image inpainting model to utilize both protein sequence and cellular images. We demonstrate that the protein sequence input enables generalization to unseen proteins, and the cellular image input captures single-cell variability, enabling cell-type-specific predictions. Experimental validation shows that PUPS can predict protein localization in newly performed experiments outside the Human Protein Atlas used for training. Collectively, PUPS provides a framework for predicting differential protein localization across cell lines and single cells within a cell line, including changes in protein localization driven by mutations.

PMID:40360932 | DOI:10.1038/s41592-025-02696-1

Nat Metab. 2025 May 13. doi: 10.1038/s42255-025-01296-9. Online ahead of print.

ABSTRACT

Adipose tissue (AT) is a complex connective tissue with a high relative proportion of adipocytes, which are specialized cells with the ability to store lipids in large droplets. AT is found in multiple discrete depots throughout the body, where it serves as the primary repository for excess calories. In addition, AT has an important role in functions as diverse as insulation, immunity and regulation of metabolic homeostasis. The Human Cell Atlas Adipose Bionetwork was established to support the generation of single-cell atlases of human AT as well as the development of unified approaches and consensus for cell annotation. Here, we provide a first roadmap from this bionetwork, including our suggested cell annotations for humans and mice, with the aim of describing the state of the field and providing guidelines for the production, analysis, interpretation and presentation of AT single-cell data.

PMID:40360756 | DOI:10.1038/s42255-025-01296-9

Nat Metab. 2025 May 13. doi: 10.1038/s42255-025-01276-z. Online ahead of print.

ABSTRACT

Metabolic reprogramming determines γδ T cell fate during thymic development; however, the metabolic requirements of interleukin (IL)-17A-producing γδ T cells (γδT17 cells) under psoriatic conditions are unclear. Combining high-throughput techniques, including RNA sequencing, SCENITH, proteomics and stable isotope tracing, we demonstrated that psoriatic inflammation caused γδT17 cells to switch toward aerobic glycolysis. Under psoriatic conditions, γδT17 cells upregulated ATP-citrate synthase to convert citrate to acetyl-CoA, linking carbohydrate metabolism and fatty acid synthesis (FAS). Accordingly, we used a pharmacological inhibitor, Soraphen A, which blocks acetyl-CoA carboxylase (ACC), to impair FAS in γδT17 cells, reducing their intracellular lipid stores and ability to produce IL-17A under psoriatic conditions in vitro. We pinpointed the pathogenic role of ACC1 in γδT17 cells in vivo by genetic ablation, ameliorating inflammation in a psoriatic mouse model. Furthermore, ACC inhibition limited human IL-17A-producing γδT17 cells. Targeting ACC1 to attenuate pathogenic γδT17 cell function has important implications for psoriasis management.

PMID:40360755 | DOI:10.1038/s42255-025-01276-z

Cell Death Discov. 2025 May 13;11(1):234. doi: 10.1038/s41420-025-02405-z.

ABSTRACT

Non-POU domain-containing octamer-binding protein (NONO) is a multifunctional member of the Drosophila behavior/human splicing (DBHS) protein family with DNA- and RNA-binding activity. NONO is highly expressed in various types of cancer, and excessive O-GlcNAcylation has also been implicated in tumorigenesis. Although recent studies revealed that NONO is O-GlcNAcylated and that this modification is involved in DNA damage repair, it remains unknown whether O-GlcNAcylation of NONO regulates cancer cell proliferation. Additionally, little is known about the effect of O-GlcNAcylation on other biological properties of NONO. In this study, we identify Thr440 as the primary NONO O-GlcNAcylation site and demonstrates its crucial role in the assembly of paraspeckles, an important subnuclear compartment that facilitates NONO-dependent transcriptional regulation in mammalian cells. Moreover, we found that O-GlcNAcylation of NONO is required to maintain the expression of genes related to microtubule cytoskeleton organization involved in mitosis and to suppress the expression of genes related to cellular response to type I interferon. Regarding the regulation of these genes, depletion of NONO O-GlcNAcylation at Thr440 significantly inhibited the proliferation of colon cancer cells. Collectively, our findings highlight NONO O-GlcNAcylation as a key regulator modulating paraspeckle formation and as a candidate therapeutic target in colon cancer.

PMID:40360465 | DOI:10.1038/s41420-025-02405-z

RNA. 2025 May 13:rna.080561.125. doi: 10.1261/rna.080561.125. Online ahead of print.

ABSTRACT

Processing bodies (P-bodies) are membrane-less organelles in eukaryotic cells that play a central role in mRNA metabolism, including mRNA decay, storage, and translational repression. However, the molecular mechanisms governing their assembly remain incompletely understood. Here, we identify the C-terminal domain of EDC4 as the minimal region required for P-body formation, with residues 1266-1401 driving phase separation and EDC4 condensation. To investigate the functional relevance of P-body integrity, we employed the microprotein Nobody (NBDY) as a selective perturbation tool. Our results revealed that the NBDY 22-41 peptide directly binds the EDC4 C-terminal domain and inhibits its self-association, thereby selectively dissolving P-bodies without affecting the canonical mRNA decay pathway. Using this tool, we further examined the impact of P-body disruption on gene expression. Transcriptome profiling combined with quantitative validation revealed that P-body loss activates the p53 pathway and enhances the stability of associated transcripts. Consistent with these findings, clinical data show that NBDY overexpression is associated with p53 pathway activation in various cancers, and the NBDY 22-41 fragment reduces tumor cell proliferation and invasion, suggesting a potentially complex role of P-body dynamics in cancer biology. Together, our study defines the EDC4 C-terminal domain as a core scaffold for P-body assembly and uncovers a regulatory role of P-body dynamics in p53-mediated gene expression, with potential implications for cancer biology.

PMID:40360209 | DOI:10.1261/rna.080561.125

Genome Res. 2025 May 13. doi: 10.1101/gr.279398.124. Online ahead of print.

ABSTRACT

The relationship between TP53 and transposable elements (TEs) has been obscure. Given the important role of TEs in oncogenesis, a comprehensive profiling of TE expression dynamics under the regulation of TP53 provides valuable resources for more clarity in TP53's roles in cancer. In this study, we characterized the TE transcriptomic landscape using long-read RNA-seq and short-read RNA-seq in three cancer cell lines varying only in TP53 genetic status. To identify transcripts that use TEs as potential promoters, we developed a computational pipeline, TEProf3, and identified in total 1942 transcripts with high confidence. Among these TE-derived transcripts, 239 are activated by TP53 and 221 are repressed by TP53. These TP53-responsive TE-derived transcripts are mainly driven by members of the ERV and LINE families. Following knockdown of wild-type (WT) TP53 expression, rescuing WT TP53 expression allows for partial recovery of the TE expression profile observed in the context of chronic TP53 expression. TP53 mutations R175H and R273H manifest their oncogenic characteristic partially through activating TE promoters in a cell type-specific manner. Lastly, we identified important sequence motifs that help govern the interactions between TEs and TP53, where TP53 activates TEs with TP53 binding motifs through direct binding and represses TEs indirectly via other pathways. Overall, we present a comprehensive profiling of the impact of TP53 on the activity of TE-derived promoters in isogenic cancer cell lines and provide a high-confidence TE expression atlas of TE promoters that are direct and indirect targets of TP53.

PMID:40360186 | DOI:10.1101/gr.279398.124

Hum Gene Ther. 2025 May 13. doi: 10.1089/hum.2024.260. Online ahead of print.

ABSTRACT

Cystic fibrosis (CF) is caused by mutations in the cystic fibrosis transmembrane conductance regulator (CFTR) gene. While CRISPR-based CFTR editing approaches have shown proof-of-concept for functional rescue in primary airway basal cells, induced pluripotent stem cells, and organoid cultures derived from patients with CF, their efficacy remains suboptimal. Here, we developed the CuFiCas9(Y66S)eGFP reporter system by integrating spCas9 and a non-fluorescent Y66S eGFP mutant into CuFi-8 cells, an immortalized human airway epithelial cell line derived from a patient with CF with homozygous F508del mutations. These cells retain the basal cell phenotype in proliferating cultures and can differentiate into polarized airway epithelium at an air-liquid interface (ALI), enabling both visualized detection of gene editing and electrophysiological assessment of CFTR functional restoration. Using this system, recombinant adeno-associated virus (rAAV)-mediated homology-directed repair (HDR) was evaluated in proliferating cultures. A correction rate of 13.5 ± 0.8% was achieved in a population where 82.3 ± 5.6% of cells were productively transduced by AAV.eGFP630g2-CMVmCh, an rAAV editing vector with an mCherry reporter. Dual-editing of F508del CFTR and Y66S eGFP was explored using AAV.HR-eGFP630-F508(g03) to deliver two templates and single guide RNAs. eGFP+ (Y66S-corrected) cells and eGFP- (non-corrected) cells were sorted via fluorescence-activated cell sorting and differentiated at an ALI to assess the recovery of CFTR function. Despite a low F508 correction rate of 2.8%, ALI cultures derived from the eGFP- population exhibited 25.2% of the CFTR-specific transepithelial Cl- transport observed in CuFi-ALI cultures treated with CFTR modulators. Next-generation sequencing revealed frequent co-editing at both genomic loci, with sixfold higher F508 correction rate in the eGFP+ cells than eGFP- cells. In both populations, non-homology end joining predominated over HDR. This reporter system provides a valuable platform for optimizing editing efficiencies in proliferating airway basal cells, particularly for development of strategies to enhance HDR through modulation of DNA repair pathways.

PMID:40359132 | DOI:10.1089/hum.2024.260

Virology. 2025 Apr 29;609:110559. doi: 10.1016/j.virol.2025.110559. Online ahead of print.

ABSTRACT

Bacteriophages, or bacterial viruses, play diverse ecological roles by shaping bacterial populations and also hold significant biotechnological and medical potential, including the treatment of infections caused by multidrug-resistant bacteria. The discovery of novel bacteriophages using large-scale metagenomic data has been accelerated by the accessibility of deep learning (Artificial Intelligence), the increased computing power of graphical processing units (GPUs), and new bioinformatics tools. This review addresses the recent revolution in bacteriophage research, ranging from the adoption of neural network algorithms applied to metagenomic data to the use of pre-trained language models, such as BERT, which have improved the reconstruction of viral metagenome-assembled genomes (vMAGs). This article also discusses the main aspects of bacteriophage biology using deep learning, highlighting the advances and limitations of this approach. Finally, prospects of deep-learning-based metagenomic algorithms and recommendations for future investigations are described.

PMID:40359589 | DOI:10.1016/j.virol.2025.110559

J Invasive Cardiol. 2025 May 13. doi: 10.25270/jic/25.00089. Online ahead of print.

ABSTRACT

Artificial intelligence (AI) has become pivotal in advancing medical care, particularly in interventional cardiology. Recent AI developments have proven effective in guiding advanced procedures and complex decisions. The authors review the latest AI-based innovations in the diagnosis of chronic total occlusions (CTO) and in determining the probability of success of CTO percutaneous coronary intervention (PCI). Neural networks and deep learning strategies were the most commonly used algorithms, and the models were trained and deployed using a variety of data types, such as clinical parameters and imaging. AI holds great promise in facilitating CTO PCI.

PMID:40359582 | DOI:10.25270/jic/25.00089

J Chem Theory Comput. 2025 May 13. doi: 10.1021/acs.jctc.5c00256. Online ahead of print.

ABSTRACT

Proteins are inherently dynamic molecules, and their conformational transitions among various states are essential for numerous biological processes, which are often modulated by their interactions with surrounding environments. Although molecular dynamics (MD) simulations are widely used to investigate these transitions, all-atom (AA) methods are often limited by short time scales and high computational costs, and coarse-grained (CG) implicit-solvent Go̅-like models are usually incapable of studying the interactions between proteins and their environments. Here, we present an approach called Multiple-basin Go̅-Martini, which combines the recent Go̅-Martini model with an exponential mixing scheme to facilitate the simulation of spontaneous protein conformational transitions in explicit environments. We demonstrate the versatility of our method through five diverse case studies: GlnBP, Arc, Hinge, SemiSWEET, and TRAAK, representing ligand-binding proteins, fold-switching proteins, de novo designed proteins, transporters, and mechanosensitive ion channels, respectively. Multiple-basin Go̅-Martini offers a new computational tool for investigating protein conformational transitions, identifying key intermediate states, and elucidating essential interactions between proteins and their environments, particularly protein-membrane interactions. In addition, this approach can efficiently generate thermodynamically meaningful data sets of protein conformational space, which may enhance deep learning-based models for predicting protein conformation distributions.

PMID:40359486 | DOI:10.1021/acs.jctc.5c00256

PLoS One. 2025 May 13;20(5):e0321530. doi: 10.1371/journal.pone.0321530. eCollection 2025.

ABSTRACT

The Consumer Price Index (CPI) is a key economic indicator used by policymakers worldwide to monitor inflation and guide monetary policy decisions. In Korea, the CPI significantly impacts decisions on interest rates, fiscal policy frameworks, and the Bank of Korea's strategies for economic stability. Given its importance, accurately forecasting the Total CPI is crucial for informed decision-making. Achieving accurate estimation, however, presents several challenges. First, the Korean Total CPI is calculated as a weighted sum of 462 items grouped into 12 categories of goods and services. This heterogeneity makes it difficult to account for all variations in consumer behavior and price dynamics. Second, the monthly frequency of CPI data results in a relatively sparse time series, limiting the performance of the analysis. Furthermore, external factors such as policy changes and pandemics add further volatility to the CPI. To address these challenges, we propose a novel framework consisting of four key components: (1) a hybrid Convolutional Neural Network-Long Short-Term Memory mechanism designed to capture complex patterns in CPI data, enhancing estimation accuracy; (2) multivariate inputs that incorporate CPI component indices alongside auxiliary variables for richer contextual information; (3) data augmentation through linear interpolation to convert monthly data into daily data, optimizing it for highly parametrized deep learning models; and (4) sentiment index derived from Korean CPI-related news articles, providing insights into external factors influencing CPI fluctuations. Experimental results demonstrate that the proposed model outperforms existing approaches in CPI prediction, as evidenced by lower RMSE values. This improved accuracy has the potential to support the development of more timely and effective economic policies.

PMID:40359407 | DOI:10.1371/journal.pone.0321530

Retina. 2025 May 6. doi: 10.1097/IAE.0000000000004511. Online ahead of print.

ABSTRACT

PURPOSE: To compare the prognosis of neovascularization of the disc (NVD) after panretinal photocoagulation (PRP) and/or ranibizumab treatment, based on OCT angiography (OCTA) patterns.

METHODS: In this prospective study, treatment-naive patients with stage IV diabetic retinopathy (DR) and NVD were imaged with 6x6 mm2 OCTA scans. NVD was classified according to OCTA morphological features: different sources (retinal arteries and veins), different activities (exuberant vascular proliferation (EVP)+ and EVP-) and different configurations (type I&II, III and IV). All patients were treated with PRP or in combination with ranibizumab. Patients were monitored monthly to detect the occurrence of vitreous haemorrhage and/or retinal detachment (VH&RD), as well as changes in best corrected visual acuity (BCVA) and NVD.

RESULTS: Among 114 eyes, 35 developed VH&RD (mean onset 6.1 months). Different configurations and EVP status (+/-) significantly affected VH&RD occurrence (p<0.05). NVD regression occurred in 52 eyes, with EVP status significantly influencing resolution (p=0.022). No significant effect was observed on visual acuity (p>0.05).

CONCLUSION: NVD can be classified into different patterns based on morphological features in OCTA, which play a crucial role in the prognosis of NVD patients after PRP and/or ranibizumab treatment.

PMID:40359330 | DOI:10.1097/IAE.0000000000004511

J Phys Chem B. 2025 May 13. doi: 10.1021/acs.jpcb.5c03007. Online ahead of print.

NO ABSTRACT

PMID:40359295 | DOI:10.1021/acs.jpcb.5c03007