From Cluster Assumption to Graph Convolution: Graph-Based Semi-Supervised Learning Revisited.

Graph-based semi-supervised learning (GSSL) has long been a research focus. Traditional methods are generally shallow learners, based on the cluster assumption. Recently, graph convolutional networks (GCNs) have become the predominant techniques for their promising performance. However, a critical question remains largely unanswered: why do deep GCNs encounter the oversmoothing problem, while traditional shallow GSSL methods do not, despite both progressing through the graph in a similar iterative manner? In this article, we theoretically discuss the relationship between these two types of methods in a unified optimization framework. One of the most intriguing findings is that, unlike traditional ones, typical GCNs may not effectively incorporate both graph structure and label information at each layer. Motivated by this, we propose three simple but powerful graph convolution methods. The first, optimized simple graph convolution (), is a supervised method, which guides the graph convolution process with labels. The others are two "no-learning" unsupervised methods: graph structure preserving graph convolution () and its multiscale version GGCM, both aiming to preserve the graph structure information during the convolution process. Finally, we conduct extensive experiments to show the effectiveness of our methods.

Dissociation of Nicotine from Acetylcholine-Binding Protein under Terahertz Waves Radiation.

The binding of nicotine (NCT) to acetylcholine-binding protein (AChBP) plays an important role in synaptic transmission and neurotransmitter regulation. However, effectively regulating their binding or dissociation processes remains a challenging problem. In this study, we employed all-atom molecular dynamics (MD) simulations to systematically investigate the impact of external terahertz (THz) waves on the binding kinetics between AChBP and NCT. We first identified the key residues (i.e., W143) and the key interactions (i.e., hydrogen bonding and cation-π interaction) in AChBP-NCT binding without THz waves. We then investigated the binding and dissociation of charged NCT with AChBP at three different frequencies (i.e., 13.02, 21.44, 42.55 THz). Importantly, the predominant vibrational modes at 13.02 THz can drive the rotation of the pentagonal ring on NCT. This leads to the disruption of hydrogen bonds between NCT and W143 and a reduced likelihood of forming cation-π interactions, resulting in the dissociation of NCT from AChBP. Additionally, we further investigated the influence of electric field intensities on the dissociation kinetics and found that when the electric field intensity exceeds a critical value (∼0.60 V/nm), the probability of ligand dissociation gradually rises as the intensity increases. In general, this study contributes to a better understanding of the effects of THz waves on protein-ligand interactions, which might also shed some light on potential applications in nicotine addiction treatment and therapeutic strategies for neurodegenerative diseases.

Effect of Nanoparticle Curvature on Its Interaction with Serum Proteins.

The size or the curvature of nanoparticles (NPs) plays an important role in regulating the composition of the protein corona. However, the molecular mechanisms of how curvature affects the interaction of NPs with serum proteins still remain elusive. In this study, we employ all-atom molecular dynamics simulations to investigate the interactions between two typical serum proteins and PEGylated Au NPs with three different surface curvatures (0, 0.1, and 0.5 nm-1, respectively). The results show that for proteins with a regular shape, the binding strength between the serum protein and Au NPs decreases with increasing curvature. For irregularly shaped proteins with noticeable grooves, the binding strength between the protein and Au NPs does not change obviously with increasing curvature in the cases of smaller curvature. However, as the curvature continues to increase, Au NPs may act as ligands firmly adsorbed in the protein grooves, significantly enhancing the binding strength. Overall, our findings suggest that the impact of NP curvature on protein adsorption may be nonmonotonic, which may provide useful guidelines for better design of functionalized NPs in biomedical applications.

Harnessing virus flexibility to selectively capture and profile rare circulating target cells for precise cancer subtyping.

The effective isolation of rare target cells, such as circulating tumor cells, from whole blood is still challenging due to the lack of a capturing surface with strong target-binding affinity and non-target-cell resistance. Here we present a solution leveraging the flexibility of bacterial virus (phage) nanofibers with their sidewalls displaying target circulating tumor cell-specific aptamers and their ends tethered to magnetic beads. Such flexible phages, with low stiffness and Young's modulus, can twist and adapt to recognize the cell receptors, energetically enhancing target cell capturing and entropically discouraging non-target cells (white blood cells) adsorption. The magnetic beads with flexible phages can isolate and count target cells with significant increase in cell affinity and reduction in non-target cell absorption compared to magnetic beads having rigid phages. This differentiates breast cancer patients and healthy donors, with impressive area under the curve (0.991) at the optimal detection threshold (>4 target cells mL-1). Immunostaining of captured circulating tumor cells precisely determines breast cancer subtypes with a diagnostic accuracy of 91.07%. Our study reveals the power of viral mechanical attributes in designing surfaces with superior target binding and non-target anti-fouling.

Effect of terahertz waves on the aggregation behavior of neurotransmitters.

Understanding the dynamics of neurotransmitters is crucial for unraveling synaptic transmission mechanisms in neuroscience. In this study, we investigated the impact of terahertz (THz) waves on the aggregation of four common neurotransmitters through all-atom molecular dynamics (MD) simulations. The simulations revealed enhanced nicotine (NCT) aggregation under 11.05 and 21.44 THz, with a minimal effect at 42.55 THz. Structural analysis further indicated strengthened intermolecular interactions and weakened hydration effects under specific THz stimulation. In addition, enhanced aggregation was observed at stronger field strengths, particularly at 21.44 THz. Furthermore, similar investigations on epinephrine (EPI), 5-hydroxytryptamine (5-HT), and γ-aminobutyric acid (GABA) corroborated these findings. Notably, EPI showed increased aggregation at 19.05 THz, emphasizing the influence of vibrational modes on aggregation. However, 5-HT and GABA, with charged or hydrophilic functional groups, exhibited minimal aggregation under THz stimulation. The present study sheds some light on neurotransmitter responses to THz waves, offering implications for neuroscience and interdisciplinary applications.

Unveiling the multicomponent phase separation through molecular dynamics simulation and graph theory.

Biomolecular condensates formed by multicomponent phase separation play crucial roles in diverse cellular processes. Accurate assessment of individual-molecule contributions to condensate formation and precise characterization of their spatial organization within condensates are crucial for understanding the underlying mechanism of phase separation. Using molecular dynamics simulations and graph theoretical analysis, we demonstrated quantitatively the significant roles of cation-π and π-π interactions mediated by aromatic residues and arginine in the formation of condensates in polypeptide systems. Our findings reveal temperature and chain length-dependent alterations in condensate network parameters, such as the number of condensate network layers, and changes in aggregation and connectivity. Notably, we observe a transition between assortativity and disassortativity in the condensate network. Moreover, polypeptides W, Y, F, and R consistently promote condensate formation, while the contributions of other charged and two polar polypeptides (Q and N) to condensate formation depend on temperature and chain length. Furthermore, polyadenosine and polyguanosine can establish stable connections with aromatic and R polypeptides, resulting in the reduced involvement of K, E, D, Q, and N in phase separation. Overall, this study provides a distinctive, precise, and quantitative approach to characterize the multicomponent phase separation.

Uncovering the Importance of Ligand Mobility on Cellular Uptake of Nanoparticles: Insights from Experimental, Computational, and Theoretical Investigations.

The cellular uptake of nanoparticles (NPs) by biological cells is an important and fundamental process in drug delivery. Previous studies reveal that the physicochemical properties of nanoparticles as well as those of functionalized ligands can both critically affect the uptake behaviors. However, the effect of the conjugation strategy (i.e., the "bond" between the ligand and the NP) on the cellular uptake is overlooked and remains largely elusive. Here, by taking the broadly employed gold nanoparticle as an example, we comprehensively assessed the relationship between the conjugation strategy and uptake behaviors by introducing three ligands with the same functional terminal but different anchoring sites. As revealed by in vitro cell experiments and multiscale molecular simulations, the uptake efficiency of gold NPs was positively correlated with the strength of the "bond" and more specifically the ligand mobility on the NP surface. Moreover, we validated the results presented above by proposing a thermodynamic theory for the wrapping of NPs with mobile ligands. Further, we also showed that the endocytic pathway of NPs was highly dependent on ligand mobility. Overall, this study uncovered a vital role of conjugation strategy in the cellular uptake and may provide useful guidelines for tailoring the biobehaviors of nanoparticles.

Effect of surface functionalization on DNA sequencing using MXene-based nanopores.

As one of the most promising types of label-free nanopores has great potential for DNA sequencing via fast detection of different DNA bases. As one of the most promising types of label-free nanopores, two-dimensional nanopore materials have been developed over the past two decades. However, how to detect different DNA bases efficiently and accurately is still a challenging problem. In the present work, the translocation of four homogeneous DNA strands (i.e., poly(A)20, poly(C)20, poly(G)20, and poly(T)20) through two-dimensional transition-metal carbide (MXene) membrane nanopores with different surface terminal groups is investigated via all-atom molecular dynamics simulations. Interestingly, it is found that the four types of bases can be distinguished by different ion currents and dwell times when they are transported through the Ti3C2(OH)2 nanopore. This is mainly attributed to the different orientation and position distributions of the bases, the hydrogen bonding inside the MXene nanopore, and the interaction of the ssDNA with the nanopore. The present study enhances the understanding of the interaction between DNA strands and MXene nanopores with different functional groups, which may provide useful guidelines for the design of MXene-based devices for DNA sequencing in the future.

Ligand Phase Separation-Promoted, "Squeezing-Out" Mode Explaining the Mechanism and Implications of Neutral Nanoparticles That Escaped from Lysosomes.

Neutral nanomaterials functionalized with PEG or similar molecules have been popularly employed as nanomedicines. Compared to positive counterparts that are capable of harnessing the well-known proton sponge effect to facilitate their escape from lysosomes, it is yet unclear how neutral substances got their entry into the cytosol. In this study, by taking PEGylated, neutral Au nanospheres as an example, we systematically investigated their time-dependent translocation postuptake. Specifically, we harnessed dissipative particle dynamics simulations to uncover how nanospheres bypass lysosomal entrapment, wherein a mechanism termed as "squeezing-out" mode was discovered. We next conducted a comprehensive investigation on how nanomaterials implicate lysosomes in terms of integrity and functionality. By using single-molecule imaging, specific preservation of PEG-terminated with targeting moieties in lysosomes supports the "squeezing-out" mode as the mechanism underlying the lysosomal escape of nanomaterials. All evidence points out that such a process is benign to lysosomes, wherein the escape of nanomaterials proceeds at the expense of targeting moieties loss. Furthermore, we proved that by fine-tuning of the efficacy of nanomaterials escaping from lysosomes, modulation of distinct pathways and metabolic machinery can be achieved readily, thereby offering us a simple and robust tool to implicate cells.

Molecular Modeling of the Fluorination Effect on the Penetration of Nanoparticles across Lipid Bilayers.

The fluorinated decorations have recently been widely used in many biomedical applications. However, the potential mechanism of the fluorination effect on the cellular delivery of nanoparticles (NPs) still remains elusive. In this work, we systemically explore the penetration of a perfluoro-octanethiol-coated gold NP (PF-Au NP) and, for comparison, an octanethiol-coated gold NP (OT-Au NP) across lipid bilayers. We also investigated the effect of these two types of NPs on the properties of lipid bilayers. Our findings indicate that the lipid type and the surface tension of the lipid bilayer significantly impact the penetration capabilities of the fluorinated gold NP. By examining the distribution of ligands on the surface of the two types of NPs in water and during the penetration process, we unveil their distinct penetration characteristics. Specifically, the PF-Au NP exhibits amphiphobic behavior (both hydrophobic and lipophobic), while the OT-Au NP exhibits solely hydrophobic characteristics. Finally, we observe that the penetration capabilities can be increased by adjusting the degree of fluorination of the ligands on the NP surface. Overall, this study provides useful physical insights into the unique properties of the fluorinated decorations in NP permeation.

Topology- and size-dependent binding of DNA nanostructures to the DNase I.

The susceptibility of DNA nanomaterials to enzymatic degradation in biological environments is a significant obstacle limiting their broad applications in biomedicine. While DNA nanostructures exhibit some resistance to nuclease degradation, the underlying mechanism of this resistance remains elusive. In this study, the interaction of tetrahedral DNA nanostructures (TDNs) and double-stranded DNA (dsDNA) with DNase I is investigated using all-atom molecular dynamics simulations. Our results indicate that DNase I can effectively bind to all dsDNA molecules, and certain key residues strongly interact with the nucleic bases of DNA. However, the binding of DNase I to TDNs exhibits a non-monotonic behavior based on size; TDN15 and TDN26 interact weakly with DNase I (∼ - 75 kcal/mol), whereas TDN21 forms a strong binding with DNase I (∼ - 110 kcal/mol). Furthermore, the topological properties of the DNA nanostructures are analyzed, and an under-twisting (∼32°) of the DNA helix is observed in TDN15 and TDN26. Importantly, this under-twisting results in an increased width of the minor groove in TDN15 and TDN26, which primarily explains their reduced binding affinity to DNase I comparing to the dsDNA. Overall, this study demonstrated a novel mechanism for local structural control of DNA at the nanoscale by adjusting the twisting induced by length.

Effect of Terahertz Waves on the Structure of the Aß42 Monomer, Dimer, and Protofibril: Insights from Molecular Dynamics Simulations.

Amyloid-ß (Aß) and its assemblies play important roles in the pathogenesis of Alzheimer's disease (AD). Recent studies conducted by experimental and computational researchers have extensively explored the structure, assembly, and influence of biomolecules and cell membranes on Aß. However, the impact of terahertz waves on the structures of Aß monomers and aggregates remains largely unexplored. In this study, we systematically investigate the molecular mechanisms by which terahertz waves affect the structure of the Aß42 monomer, dimer, and tetramer through all-atom molecular dynamics (MD) simulations. Our findings indicate that terahertz waves at a specific frequency (42.55 THz) can enhance intramolecular and intermolecular interactions in the Aß42 monomer and dimer, respectively, by resonating with the symmetric stretching mode of the -COO- groups and the symmetric bending/stretching mode of -CH3 groups. Consequently, the ß-structure content of the Aß42 monomer is greatly increased, and the binding energy between the monomers in the Aß42 dimer is significantly enhanced. Additionally, our observations suggest that terahertz waves can mildly stabilize the structure of tetrameric protofibrils by enhancing the interactions among peripheral peptides. Furthermore, we also investigated the effect of the frequency of terahertz waves on the structure of Aß42. The present study contributes to a better understanding of the impact of external fields on the biobehavior of Aß42 peptides and may shed some light on the potential risks associated with electromagnetic field radiation.

Exploring the non-monotonic DNA capture behavior in a charged graphene nanopore.

Nanopore-based biomolecule detection has emerged as a promising and sought-after innovation, offering high throughput, rapidity, label-free analysis, and cost-effectiveness, with potential applications in personalized medicine. However, achieving efficient and tunable biomolecule capture into the nanopore remains a significant challenge. In this study, we employ all-atom molecular dynamics simulations to investigate the capture of double-stranded DNA (dsDNA) molecules into graphene nanopores with varying positive charges. We discover a non-monotonic relationship between the DNA capture rate and the charge of the graphene nanopore. Specifically, the capture rate initially decreases and then increases with an increase in nanopore charge. This behavior is primarily attributed to differences in the electrophoretic force, rather than the influence of electroosmosis or counterions. Furthermore, we also observe this non-monotonic trend in various ionic solutions, but not in ionless solutions. Our findings shed light on the design of novel DNA sequencing devices, offering valuable insights into enhancing biomolecule capture rates in nanopore-based sensing platforms.

Effect of maleylation and denaturation of human serum albumin on its interaction with scavenger receptors.

The specific recognition of serum proteins by scavenger receptors is critical and fundamental in many biological processes. However, the underlying mechanism of scavenger receptor-serum protein interaction remains elusive. In this work, taking scavenger receptors class A1 (SR-A1) as an example, we systematically investigate its interaction with human serum albumin (HSA) at different states through a combination of molecular docking and all-atom molecular dynamics simulations. It is found that native HSA can moderately bind to collagen-like (CL) region or scavenger receptor cysteine-rich (SRCR) region, with both electrostatic (ELE) and van der Waals (VDW) interactions, playing important roles. After maleylation, the binding energy, particularly the ELE energy, between HSA and CL region is significantly enhanced, while the binding energy between HSA and SRCR region remains nearly unchanged. Additionally, we also observe that unfolding of the secondary structures in HSA leads to a larger contact surface area between denatured HSA and CL region, but has little impact on the HSA-SRCR region interaction. Therefore, similar to maleylated HSA, denatured HSA is also more likely to bind to the CL region of SR-A1.

A Fluid Multivalent Magnetic Interface for High-Performance Isolation and Proteomic Profiling of Tumor-Derived Extracellular Vesicles.

Isolation and analysis of tumor-derived extracellular vesicles (T-EVs) are important for clinical cancer management. Here, we develop a fluid multivalent magnetic interface (FluidmagFace) in a microfluidic chip for high-performance isolation, release, and protein profiling of T-EVs. The FluidmagFace increases affinity by 105-fold with fluidity-enhanced multivalent binding to improve isolation efficiency by 13.9 % compared with a non-fluid interface. Its anti-adsorption property and microfluidic hydrodynamic shear minimize contamination, increasing detection sensitivity by two orders of magnitude. Moreover, its reversibility and expandability allow high-throughput recovery of T-EVs for mass spectrometric protein analysis. With the chip, T-EVs were detected in all tested cancer samples with identification of differentially expressed proteins compared with healthy controls. The FluidmagFace opens a new avenue to isolation and release of targets for cancer diagnosis and biomarker discovery.

Effect of the Graphene Nanosheet on Functions of the Spike Protein in Open and Closed States: Comparison between SARS-CoV-2 Wild Type and the Omicron Variant.

The spread of coronavirus disease 2019 caused by SARS-CoV-2 and its variants has become a global health crisis. Although there were many attempts to use nanomaterials-based devices to fight against SARS-CoV-2, it still remains elusive as to how the nanomaterials interact with SARS-CoV-2 and affect its biofunctions. Here, taking the graphene nanosheet (GN) as the model nanomaterial, we investigate its interaction with the spike protein in both WT and Omicron by molecular simulations. In the closed state, the GN can insert into the region between the receptor binding domain (RBD) and the N-terminal domain (NTD) in both wild type (WT) and Omicron, which keeps the RBD in the down conformation. In the open state, the GN can hamper the binding of up RBD to ACE2 in WT, but it has little impact on up RBD and, even worse, stimulates the down-to-up transition of down RBDs in Omicron. Moreover, the GN can insert in the vicinity of the fusion peptide in both WT and Omicron and prevents the detachment of S1 from the whole spike protein. The present study reveals the effect of the SARS-CoV-2 variant on the nanomaterial-spike protein interaction, which informs prospective efforts to design functional nanomaterials against SARS-CoV-2.

Synergetic collision and space separation in microfluidic chip for efficient affinity-discriminated molecular selection.

Efficient molecular selection is a prerequisite for generating molecular tools used in diagnosis, pathology, vaccinology, and therapeutics. Selection efficiency is thermodynamically highly dependent on the dissociation equilibrium that can be reached in a single round. Extreme shifting of equilibrium towards dissociation favors the retention of high-affinity ligands over those with lower affinity, thus improving the selection efficiency. We propose to synergize dual effects by deterministic lateral-displacement microfluidics, including the collision-based force effect and the two-dimensional (2D) separation-based concentration effect, to greatly shift the equilibrium. Compared with previous approaches, this system can remove more low- or moderate-affinity ligands and maintain most high-affinity ligands, thereby improving affinity discrimination in selection. This strategy is demonstrated on phage display in both experiment and simulation, and two peptides against tumor markers ephrin type-A receptor 2 (EphA2) and CD71 were obtained with high affinity and specificity within a single round of selection, which offers a promising direction for discovery of robust binding ligands for a wide range of biomedical applications.

Loading of DOX into a tetrahedral DNA nanostructure: the corner does matter.

With the rapid development of nanotechnology, various DNA nanostructures have been synthesized and widely used in drug delivery. However, the underlying mechanisms of drug molecule loading into the DNA nanostructure are still elusive. In this work, we systematically investigate the interactions of a tetrahedral DNA nanostructure (TDN) with the anti-cancer drug doxorubicin (DOX) by combining molecular docking and all-atom molecular dynamics simulations. It is found that there are five possible binding modes in the single TDN-DOX interactions, namely the outside-corner mode, the inside-corner mode, the major-groove mode, the minor-groove mode, and the intercalation mode, where the van der Waals (VDW) interaction and the electrostatic (ELE) interaction dominate in the case of unionized DOX and ionized DOX, respectively. Moreover, with the increase of the DOX number, some of the interaction modes may disappear and the inside-corner mode is the most energy-favorable mode. The present study enhances the molecular understanding of the role of TDN as the drug carrier, which may provide a useful guideline for the future design of DNA nanostructures.

Uncovering the molecular mechanism for dual effect of ATP on phase separation in FUS solution.

Recent studies reported that adenosine triphosphate (ATP) could inhibit and enhance the phase separation in prion-like proteins. The molecular mechanism underlying such a puzzling phenomenon remains elusive. Here, taking the fused in sarcoma (FUS) solution as an example, we comprehensively reveal the underlying mechanism by which ATP regulates phase separation by combining the semiempirical quantum mechanical method, mean-field theory, and molecular simulation. At the microscopic level, ATP acts as a bivalent or trivalent binder; at the macroscopic level, the reentrant phase separation occurs in dilute FUS solutions, resulting from the ATP concentration-dependent binding ability under different conditions. The ATP concentration for dissolving the protein condensates is about 10 mM, agreeing with experimental results. Furthermore, from a dynamic point of view, the effect of ATP on phase separation is also nonmonotonic. This work provides a clear physical description of the microscopic interaction and macroscopic phase diagram of the ATP-modulated phase separation.

Efficient calculation of protein-ligand binding free energy using GFN methods: the power of the cluster model.

Protein-ligand interactions are crucial in many biochemical processes and biomedical applications, yet accurately calculating the binding free energy of the interactions still remains challenging. In this work, we systematically investigate the performance of a generic force field GFN-FF and some semi-empirical quantum mechanical (SQM) methods (GFNn, n = 0, 1, 2) in terms of the accuracy of the calculated binding free energy. It is found that the performance of the GFN-FF method is quite good in a neutral-ligand system since the Pearson correlation coefficient (rp) is 0.70 and the mean absolute error (MAE) is 5.49 kcal mol-1. However, it may fail in a charged-ligand system (the MAE is 18.98 kcal mol-1). Moreover, we also propose a cluster model (i.e., truncating the protein at a given cutoff) along with the SQM method in the GFN family. Importantly, the GFN2-xTB shows the best performance among the SQM methods (the MAE is 4.91 kcal mol-1 and 10.25 kcal mol-1 in the neutral-ligand and charged-ligand systems, respectively), much better than GFN-FF in the charged-ligand system. Notably, the computing cost of the GFN2-xTB in the appropriate cluster model is even lower than that of the GFN-FF (in the entire complex). The present study sheds some light on the potential power of the GFN family in the efficient calculation of the binding free energy in bio-systems.