RNA-based technologies have demonstrated significant potential for diverse applications, ranging from vaccination to gene editing. However, their widespread adoption is limited by the critical challenge of efficient delivery. Lipid nanoparticles (LNPs) have emerged as a widely utilized RNA delivery system, yet their formulation design and optimization primarily rely on empirical trial-and-error, which is labor-intensive, time-consuming, and cost-prohibitive, thus hindering the rapid development of RNA therapeutics. To facilitate the early-stage design and optimization of LNPs for enhanced delivery efficiency, in this study, we construct LNPs-TE, a benchmark dataset comprising over 10 000 experimentally measured transfection efficiency (TE) values, and introduce LNPs integrated feature fusion Transformer (LIFT), a deep learning framework for LNPs TE prediction. Comprehensive experiments demonstrate that LIFT effectively integrates multidimensional molecular representations of ionizable lipids, the key component in LNPs formulation, achieving superior predictive performance, with an average Pearson correlation coefficient of 0.845 for regression and an area under the receiver operating characteristic curve (AUC-ROC) of 0.818 for multi-class classification across multiple datasets. Through scaffold-based splitting and activity cliff tasks, we further validated the exceptional generalization ability and robustness of LIFT, which achieved over a 10% improvement in the coefficient of determination (R2) compared with state-of-the-art baseline models, highlighting its potential as a practical and stable approach for the virtual screening of efficient LNPs formulation. The relevant data, model and code are made publicly available at https://github.com/U12458/LIFT.

Link prediction is widely used in various fields to predict the missing or potential future links between node pairs in the network. Knowledge distillation (KD) methods have been introduced for the link prediction task, demonstrating exceptional performance in inference acceleration. However, these methods primarily focus on supervised learning settings that rely on high-quality labels, and they fail to capture the rich structural features inherent in graph data within the teacher model effectively, resulting in suboptimal performance. To address the problem of reducing label dependency, we propose a link prediction method via contrastive learning with relational distillation (LCLRD) that aims to improve model performance under label scarcity. LCLRD incorporates contrastive learning into the teacher model for pretraining, enabling the teacher model to obtain high-quality node representations more effectively. Then, graph structural information is extracted from the teacher model via relational distillation and transferred to the student model. In addition, we leverage the Pearson correlation coefficient (PCC) as a new matching strategy to replace the Kullback–Leibler (KL) divergence, aiming to alleviate the prediction discrepancy between the student and the stronger teacher model. The experimental results show that LCLRD significantly outperforms other baseline methods on 12 datasets, demonstrating the superiority of this approach.

The fruit fly optimization algorithm has drawn various attention to researchers and engineers, due to the simple theory and flexible frame. For solving the complex continuous optimization problems, an improved fruit fly algorithm based on vision scanning search and extensive learning mechanism is proposed in this paper. The vision scanning search strategy is used to scan the potential area by changing the search angle of swarm center. This strategy is utilized to guide the population to jump out the local trap. The extensive learning is using the knowledge of neighboring structure to increase the diversity of population for solving the non-separable issues. Furthermore, a new mutation strategy based on difference vector is proposed to improve the search efficiency of VLFOA. Testified in CEC 2017 benchmark problems, the results show that the VLFOA has a superior performance compared with the original FOA and the state of art variants of the FOA.

In the past decades, various attention has been paid to the global optimization problems. The Monarch Butterfly Optimization (MBO) algorithm is an effective meta-heuristic algorithm for the global optimization problems. However, in the MBO, the diversity of the population is lost in the late iteration. The MBO is easy to trap into the local optima. In this study, an algorithm based on MBO with learning mechanism and topological structure, named LTMBO, is proposed to enhance the ability of exploration and exploitation on the global optimization problems. The learning mechanism is present for the migration operator to increase the speed of the iteration. The topological structure is proposed for the butterfly adjusting operator to improve the diversity of the population. The experimental results demonstrated that the efficiency and significance of the proposed LTMBO algorithm.

A new metaheuristic algorithm, named surrogate-guided algorithm(S-Jaya), is proposed to solve the single objective continuous optimization problems in this paper. A novel mutation strategy for the non-separable single objective continuous optimization problems is introduced to alter the search engine of the Jaya algorithm. The surrogate is embedded to accelerate the convergence of the population and avoid the proposed algorithm falling into the local optimal during the evolutionary process. The suggested S-Jaya algorithm to address the CEC 2017 benchmark problems is effective and validated. On the quality of solution and execution time, the experimental results reveal that the effectiveness of the S-Jaya algorithm is superior compare with the Jaya algorithm and its variants.

BACKGROUND: Long noncoding RNAs (lncRNAs) have dense linkages with a plethora of important cellular activities. lncRNAs exert functions by linking with corresponding RNA-binding proteins. Since experimental techniques to detect lncRNA-protein interactions (LPIs) are laborious and time-consuming, a few computational methods have been reported for LPI prediction. However, computation-based LPI identification methods have the following limitations: (1) Most methods were evaluated on a single dataset, and researchers may thus fail to measure their generalization ability. (2) The majority of methods were validated under cross validation on lncRNA-protein pairs, did not investigate the performance under other cross validations, especially for cross validation on independent lncRNAs and independent proteins. (3) lncRNAs and proteins have abundant biological information, how to select informative features need to further investigate. RESULTS: Under a hybrid framework (LPI-HyADBS) integrating feature selection based on AdaBoost, and classification models including deep neural network (DNN), extreme gradient Boost (XGBoost), and SVM with a penalty Coefficient of misclassification (C-SVM), this work focuses on finding new LPIs. First, five datasets are arranged. Each dataset contains lncRNA sequences, protein sequences, and an LPI network. Second, biological features of lncRNAs and proteins are acquired based on Pyfeat. Third, the obtained features of lncRNAs and proteins are selected based on AdaBoost and concatenated to depict each LPI sample. Fourth, DNN, XGBoost, and C-SVM are used to classify lncRNA-protein pairs based on the concatenated features. Finally, a hybrid framework is developed to integrate the classification results from the above three classifiers. LPI-HyADBS is compared to six classical LPI prediction approaches (LPI-SKF, LPI-NRLMF, Capsule-LPI, LPI-CNNCP, LPLNP, and LPBNI) on five datasets under 5-fold cross validations on lncRNAs, proteins, lncRNA-protein pairs, and independent lncRNAs and independent proteins. The results show LPI-HyADBS has the best LPI prediction performance under four different cross validations. In particular, LPI-HyADBS obtains better classification ability than other six approaches under the constructed independent dataset. Case analyses suggest that there is relevance between ZNF667-AS1 and Q15717. CONCLUSIONS: Integrating feature selection approach based on AdaBoost, three classification techniques including DNN, XGBoost, and C-SVM, this work develops a hybrid framework to identify new linkages between lncRNAs and proteins.

Bag-of-features (BOF) derived from local visual features has recently been widely used in content based image classification and scene detection owing to their simplicity and good performance. However, the hyper-dimension of the BOF vector has limited its implementation in large scale datasets because of its high computation complexity. In this paper, we present a new strategy based on the multi-resolution structure of BOF vectors to gain a speed-up of matching. We construct the new structure in two different ways: the uniform quantization method and the non-uniform quantization method. The main idea is to build low level histograms according to the BOF vector. We also introduce the VA-file method in our approach to give an approximation limit in order to accelerate the searching speed of multi-resolution BOF candidate vectors. Experiments results show that our approach has made a great improvement in both efficiency and computational complexity than traditional BOF methods.

A new approach is presented for solving long-horizon, constrained optimal control problems by using the mixed coordination method. The method was originally developed for unconstrained optimal control problems, with key ideas including time decomposition, mixed coordination and parallel processing. In extending the method to constrained problems, constraints on state and control variables are relaxed by using the multiplier method. For a given set of Lagrange multipliers, the problem is unconstrained and is solved by using the mixed coordination method. The Lagrange multipliers are then updated in a simple and efficient way. Three problems, including one with nonlinear system dynamics and constraints, and a hydroelectric scheduling problem with ten reservoirs, are tested. Results shows that the new approach is numerically stable, and significant speedups are obtained in a simulated parallel-processing environment. The method can be easily extended to handle unpredicted changes during the online operation phase of a system.>