The European Physical Journal E最新文献

Artificial microswimmers, such as active colloids, have the potential to revolutionize targeted drug delivery, but controlling their motion under imposed flow conditions remains challenging. In this work, we implement reinforcement learning (RL) to control the navigation of a microswimmer in a plane Poiseuille flow, with applications in targeted drug delivery. With RL, the swimmer learns to efficiently reach its target by continuously adjusting its swinging or tumbling behavior depending upon its self-propulsion strength, chirality and the imposed flow strength. This RL-based approach enables precise control of the particle’s path, achieving reliable targeting even in stringent scenarios such as upstream motion in high bulk flow, thus advancing the design of intelligent in vivo medical microrobots.

This study investigates the impact of clay content and temperature variation on the electrical conductivity of three-dimensional fluid-filled porous rocks. The role of varying pore throat radii has been included in the course of clay fraction variation in the conducting channels of the rock samples. The research identifies a critical ratio of clay conductance to fluid conductance that dictates the regime of electrical conductance behaviour. A nonlinear increase in electrical conductance is observed when the clay-to-fluid conductance ratio exceeds the critical ratio, whereas a linear relationship is maintained below this critical ratio. A modified form of Archie’s law relating effective conductivity and porosity has been proposed for the clay coated channels. The intricate relationship between Peclet number, pore throat size, and temperature on the electrical conductivity of fluid-filled straight channels in three dimensions has also been investigated. Results revealed a quadratic increase in conductance with porosity under steady-state conditions across all Peclet number ranges examined. While the conductivity remained constant with porosity for each Peclet number, the rate of increase in conductivity diminished with it. Nonlinear increase in conductivity was observed with temperature in the transient flow regime with a threshold temperature marking the onset of conductivity. Conductivity was augmented with increase in observation time in the transient state for the entire temperature range considered. Close to the attainment of saturation in electrical conductivity, the conductivity changed linearly with temperature until a steady value was reached.

A novel field theoretical approach towards modelling dynamic networking in complex systems is presented. An equilibrium networking formalism which utilises Gaussian fields is adapted to model the dynamics of particles that can bind and unbind from one another. Here, networking refers to the introduction of instantaneous co-localisation constraints and does not necessitate the formation of a well-defined transient or persistent network. By combining this formalism with Martin–Siggia–Rose generating functionals, a weighted generating functional for the networked system is obtained. The networking formalism introduces spatial and temporal constraints into the Langevin dynamics, via statistical weights, thereby accounting for all possible configurations in which particles can be networked to one another. A simple example of Brownian particles which can bind and unbind from one another demonstrates the tool and that this leads to results for physical quantities in a collective description. Applying the networking formalism to model the dynamics of cross-linking polymers in a mixture, we can calculate the average number of networking instances. As expected, the dynamic structure factors for each type of polymer show that the system collapses once networking is introduced, but that the addition of a repulsive time-dependent potential above a minimum strength prevents this. The examples presented in this paper indicate that this novel approach towards modelling dynamic networking could be applied to a range of synthetic and biological systems to obtain theoretical predictions for experimentally verifiable quantities.

Barrier-crossing processes in nature are often non-Markovian and typically occur over an asymmetric double-well free-energy landscape. However, most theories and numerical studies on barrier-crossing rates assume symmetric free-energy profiles. Here, we use a one-dimensional generalized Langevin equation (GLE) to investigate non-Markovian reaction kinetics in asymmetric double-well potentials. We derive a general formula, confirmed by extensive simulations, that accurately predicts mean first-passage times from well to barrier top in an asymmetric double-well potential with arbitrary memory time and reaction coordinate mass. We extend our formalism to non-equilibrium non-Markovian systems, confirming its broad applicability to equilibrium and non-equilibrium systems in biology, chemistry, and physics.

This research investigates the anticipated physicochemical and topological properties of compounds such as drug complexity (C), molecular weight (MW), and topological polar surface area (TPSA) using quantitative structure–activity relationship (QSAR) analysis. Several machine learning models, including Linear Regression, Ridge Regression, Lasso Regression, Random Forest Regression, and Gradient Boosting, were developed to improve prediction accuracy using topological indices. The datasets were combined with appropriate topological indices for individual compounds. Model performance was evaluated using Mean Squared Error (MSE) and (R^2) score after hyperparameter tuning via GridSearchCV. Ridge and Lasso Regression models stood out due to their lowest Test MSE averages (3617.74 and 3540.23, respectively) and highest (R^2) scores (0.9322 and 0.9374, respectively), demonstrating their effectiveness in handling multicollinearity and preventing overfitting. Linear Regression also performed robustly, achieving an MSE of 5249.97 and an (R^2) of 0.8563, highlighting the suitability of simpler models for datasets with inherent linear relationships. While Random Forest and Gradient Boosting Regression are capable of capturing nonlinear relationships, their performance varied. Random Forest Regression achieved an MSE of 6485.45 and an (R^2) of 0.6643, while Gradient Boosting initially performed poorly with an MSE of 4488.04 and an (R^2) of 0.5659. After fine-tuning Gradient Boosting with an expanded hyperparameter grid, its performance improved significantly, achieving a Test MSE of 1494.74 and an (R^2) of 0.9171. However, it still ranked fourth, suggesting that simpler models like Linear, Ridge, and Lasso Regression may be better suited for this dataset. This work emphasizes the significance of accurate model selection and optimization in QSAR analysis, demonstrating how these approaches can be used to develop dependable predictive models in computational drug discovery and cheminformatics.

A machine learning pipeline for predicting physicochemical and topological properties of chemical compounds using QSAR analysis. The process begins with compound data collection from PubChem, followed by data preprocessing, feature engineering, and feature selection. The selected features are used to train various regression models-including Linear, Ridge, Lasso, Random Forest, and Gradient Boosting Regression-evaluated using MSE and (R^2) metrics for performance comparison.caption for the graphical abstract: Caption for Graphical Abstract: A machine learning pipeline for predicting physicochemical and topological properties of chemical compounds using QSAR analysis. The process begins with compound data collection from PubChem, followed by data preprocessing, feature engineering, and feature selection. The selected features are used to train various regression models-incl

Suspension of particles in a fluid solvent are ubiquitous in nature, for example water mixed with sugar or bacteria self-propelling through mucus. Particles create local flow perturbations that can modify drastically the effective (homogenized) bulk properties of the fluid. Understanding the link between the properties of particles and the fluid solvent, and the effective properties of the medium is a classical problem in fluid mechanics. Here we study a special case of a two-dimensional model of a suspension of undeformable particles in a liquid crystal solvent. In the dilute regime, we calculate asymptotic solutions of the perturbations of the velocity and director fields and derive an explicit formula for an effective shear viscosity of the liquid crystal medium. Such effective shear viscosity increases linearly with the area fraction of particles, similar to Einstein formula but with a different prefactor. We provide explicit asymptotic formulas for the dependence of this prefactor on the material parameters of the solvent. Finally, we identify a case of decreasing the effective viscosity by increasing the magnitude of the shear-flow alignment coefficient of the liquid crystal solvent.

Collective behavior is all around us, from flocks of birds to schools of fish. These systems are immensely complex, which makes it pertinent to study their behavior through minimal models. We introduce such a minimal model for cohesive and aligning self-propelled particles in which group cohesion is established through additive, non-reciprocal torques. These torques cause a particle’s orientation vector to turn toward its neighbor so that it aligns with the separation vector. We additionally incorporate an alignment torque, which competes with the cohesive torque in the same spatial range. By changing the strength and range of these torque interactions, we uncover six states which we distinguish via their static and dynamic properties: a disperse state, a multiple worm state, a line state, a persistent worm state, a rotary worm state, and an aster state. Their occurrence strongly depends on initial conditions and stochasticity, so the model exhibits multistabilities. A number of the states exhibit collective dynamics which are reminiscent of those seen in nature.

Supervised machine learning methods like random forests and extreme gradient boosting plays an important role in drug development for predicting bioactivity and resolving structure-activity correlations. These approaches use topological descriptors in the study of polycyclic aromatic hydrocarbons that represent molecular structural characteristics to enhance the prediction capacity of quantitative structure–property relationships (QSPR). The objective is to identify the physoichemical properties such as density, boiling point, flash point, enthalpy, polarizability, surface tension, molar volume, molecular weight and complexity that significantly impact physicochemical attributes. The combination of machine learning and QSPR also demonstrates the potential of computational techniques in drug development. Then effective algorithms are constructed to express the link between the eccentricity-based topological indices and the physicochemical characteristics of each of the polycyclic aromatic hydrocarbons, which grows our understanding of their behavior and paves the way for future development of environmental forecasting techniques and toxicological evaluations of polycyclic aromatic hydrocarbons.