Essential proteins play a pivotal role in biological systems, and their identification is critical for elucidating complex life mechanisms and disease pathogenesis. While most existing identification methods rely on protein-protein interaction networks (PINs), the inherent noise in raw PINs frequently compromises identification efficiency. To address this limitation, researchers often employ network refinement strategies to improve PIN quality, with modularity-based approaches demonstrating particularly strong efficacy in enhancing the predictive performance of essential protein identification. Emerging evidence reveals that protein complexes exhibit distinct fine-grained modular architectures. This observation suggests that systematically incorporating modular granularity considerations during refined PIN construction could significantly improve the accuracy of essential protein identification. Building on this premise, we propose a novel granularity-aware network refinement framework that integrates hierarchical module division and multi-source biological evidence for essential protein discovery. Our methodology comprises three key phases: (1) Weighted PIN construction through integration of gene expression profiles and Gene Ontology (GO) biological process annotations; (2) Hierarchical module division using the Louvain algorithm to optimize granularity resolution; (3) Critical module detection via synthesis of evolutionary conservation scores and nuclear localization enrichment metrics, ultimately generating a Granularity-Modulated PIN (GM-PIN). To validate our approach, we performed comparative evaluations against four baseline networks (S-PIN, D-PIN, RD-PIN and CM-PIN) using ten state-of-the-art essential protein identification methods (DC, LAC, NC, EC, BC, CC, SC, PR, PeC and WDC) across two species (yeast and human). Experimental results demonstrate that GM-PIN achieves superior performance across multiple metrics, including essential proteins identified in top K, Jackknifing, PRAUC, sensitivity, F-measure and accuracy. These findings confirm that our granularity-modulated refinement method effectively constructs high-quality PINs that substantially enhance essential protein discovery.

Time-sensitive networking (TSN) is an emerging technology widely used in real-time systems for its high bandwidth and deterministic timing properties. To ensure the deterministic transmission of Time-triggered (TT) messages, a guard band mechanism is employed to prevent interference from other messages, such as Audio-Video Bridging (AVB) and Best-effort (BE) messages, before transmitting the TT messages in TSN. However, this mechanism introduces transmission delays for non-TT messages and bandwidth wastes for the physical links. Another challenge arises from the default First-in-first-out (FIFO) order of incoming messages, resulting in a relatively low acceptance rate for non-TT messages. To address these issues, a hybrid scheduling algorithm based on the min-heap structure (HSMH) is proposed. For AVB messages, HSMH sorts them in ascending style on the basis of deadlines, guaranteeing the earliest deadline message to be sent first. For BE messages, a threshold is designed to diverge them into two queues: a FIFO queue and a STF (shortest-time-first) queue. The former outputs the messages in a FIFO style, while the latter outputs messages in a STF style. All the output order of AVB messages and STF-queue messages are arranged in a min-heap structure. The algorithm can efficiently improve the transmission rate of AVB messages, the sending rate of BE messages, and the overall link utilization. Experimental results demonstrate that the proposed algorithm outperforms existing approaches in all these three aspects.

This paper introduces a novel Real Relative encoding Genetic Algorithm (R$^{2}$GA) to tackle the workflow scheduling problem in heterogeneous distributed computing systems (HDCS). R$^{2}$GA employs a unique encoding mechanism, using real numbers to represent the relative positions of tasks in the schedulable task set. Decoding is performed by interpreting these real numbers in relation to the directed acyclic graph (DAG) of the workflow. This approach ensures that any sequence of randomly generated real numbers, produced by cross-over and mutation operations, can always be decoded into a valid solution, as the precedence constraints between tasks are explicitly defined by the DAG. The proposed encoding and decoding mechanism simplifies genetic operations and facilitates efficient exploration of the solution space. This inherent flexibility also allows R$^{2}$GA to be easily adapted to various optimization scenarios in workflow scheduling within HDCS. Additionally, R$^{2}$GA overcomes several issues associated with traditional genetic algorithms (GAs) and existing real-number encoding GAs, such as the generation of chromosomes that violate task precedence constraints and the strict limitations on gene value ranges. Experimental results show that R$^{2}$GA consistently delivers superior performance in terms of solution quality and efficiency compared to existing techniques.

BACKGROUND: The identification of essential proteins can help in understanding the minimum requirements for cell survival and development to discover drug targets and prevent disease. Nowadays, node ranking methods are a common way to identify essential proteins, but the poor data quality of the underlying PIN has somewhat hindered the identification accuracy of essential proteins for these methods in the PIN. Therefore, researchers constructed refinement networks by considering certain biological properties of interacting protein pairs to improve the performance of node ranking methods in the PIN. Studies show that proteins in a complex are more likely to be essential than proteins not present in the complex. However, the modularity is usually ignored for the refinement methods of the PINs. METHODS: Based on this, we proposed a network refinement method based on module discovery and biological information. The idea is, first, to extract the maximal connected subgraph in the PIN, and to divide it into different modules by using Fast-unfolding algorithm; then, to detect critical modules according to the orthologous information, subcellular localization information and topology information within each module; finally, to construct a more refined network (CM-PIN) by using the identified critical modules. RESULTS: To evaluate the effectiveness of the proposed method, we used 12 typical node ranking methods (LAC, DC, DMNC, NC, TP, LID, CC, BC, PR, LR, PeC, WDC) to compare the overall performance of the CM-PIN with those on the S-PIN, D-PIN and RD-PIN. The experimental results showed that the CM-PIN was optimal in terms of the identification number of essential proteins, precision-recall curve, Jackknifing method and other criteria, and can help to identify essential proteins more accurately.

Essential proteins play an important role in various life activities and are considered to be a vital part of the organism. Gene expression data are an important dataset to construct dynamic protein-protein interaction networks (DPIN). The existing methods for the construction of DPINs generally utilize all features (or the features in a cycle) of the gene expression data. However, the features observed from successive time points tend to be highly correlated, and thus there are some redundant and irrelevant features in the gene expression data, which will influence the quality of the constructed network and the predictive performance of essential proteins. To address this problem, we propose a construction method of DPINs by using selected relevant features rather than continuous and periodic features. We adopt an improved unsupervised feature selection method based on Laplacian algorithm to remove irrelevant and redundant features from gene expression data, then integrate the chosen relevant features into the static protein-protein interaction network (SPIN) to construct a more concise and effective DPIN (FS-DPIN). To evaluate the effectiveness of the FS-DPIN, we apply 15 network-based centrality methods on the FS-DPIN and compare the results with those on the SPIN and the existing DPINs. Then the predictive performance of the 15 centrality methods is validated in terms of sensitivity, specificity, positive predictive value, negative predictive value, F-measure, accuracy, Jackknife and AUPRC. The experimental results show that the FS-DPIN is superior to the existing DPINs in the identification accuracy of essential proteins.

The Internet of Things (IoT) platform has played a significant role in improving road transport safety and efficiency by ubiquitously connecting intelligent vehicles through wireless communications. Such an IoT paradigm however, brings in considerable strain on limited spectrum resources due to the need of continuous communication and monitoring. Cognitive radio (CR) is a potential approach to alleviate the spectrum scarcity problem through opportunistic exploitation of the underutilized spectrum. However, highly dynamic topology and time-varying spectrum states in CR-based vehicular networks introduce quite a few challenges to be addressed. Moreover, a variety of vehicular communication modes, such as vehicle-to-infrastructure and vehicle-to-vehicle, as well as data QoS requirements pose critical issues on efficient transmission scheduling. Based on this motivation, in this paper, we adopt a deep Q -learning approach for designing an optimal data transmission scheduling scheme in cognitive vehicular networks to minimize transmission costs while also fully utilizing various communication modes and resources. Furthermore, we investigate the characteristics of communication modes and spectrum resources chosen by vehicles in different network states, and propose an efficient learning algorithm for obtaining the optimal scheduling strategies. Numerical results are presented to illustrate the performance of the proposed scheduling schemes.

A task' s end-to-end delay in its execution is a key requirement to real-time systems. This paper presents a configurable state class method based on time Petri nets for their quantitative analysis. The proposed method has a flexible state class structure. A firing domain is separated into a kernel domain that supports the basic evolution of state classes, and a configurable domain that is used to evaluate end-to-end delays. Since both domains adopt a uniform representation for time constraints, end-to-end delays can be computed synchronously with the evolution of state classes via the same firing rules. Firing rules are decomposed into basic timing operations. This treatment not only makes the calculation of end-to-end delays more flexible, but also provides a scalable way to add new timing operations into a time Petri net model. The proposed method computes arbitrary end-to-end delays along a trace with time O(ml2) and space O(l2), where m is the number of firing transitions along the trace and l is the maximum number of transitions in configurable and kernel domains. Compared with the existing state class methods, it has better performance and flexibility in on-the-fly computation of end-to-end delays.