Liu Tiyao, Wang Shudong, Qiao Sibo, Zhao Yawu, Zhang Kuijie, Tan Xiaodong, Wu Wenhao, Wang Shaoqiang
College of Computer Science and Technology, Qingdao Institute of Software, China University of Petroleum, Qingdao 266580, China; State Key Laboratory of Chemical Safety, Qingdao, 266580, China; Shandong Key Laboratory of Intelligent Oil & Gas Industrial Software, Qingdao, 266580, China.
The School of Software, Tiangong University, Tianjin 300387, China.
Int J Biol Macromol. 2025 Sep;321(Pt 3):145404. doi: 10.1016/j.ijbiomac.2025.145404. Epub 2025 Jul 12.
The interaction between miRNAs and lncRNAs has a crucial effect in the gene expression regulatory network, significantly influencing cellular functions and the development of diseases. Therefore, inferring potential miRNA-lncRNA interactions is essential. With the increase in data size, the computational complexity of models has grown, necessitating efficient methods for fast prediction of molecular interactions. Additionally, most existing models rely on a single similarity metric, introducing bias into computational modeling. The potential of similarity-based topologies has also not been fully explored or utilized. In this study, we propose a rapid prediction model for miRNA-lncRNA interactions utilizing Multi-view Projection Fusion and Random Parallel Matrix Factorization (MVPFDPC). First, we construct a comprehensive miRNA/lncRNA similarity profile. Second, we have developed a multi-view projection fusion technique that leverages multiple similarity network topologies from various spaces along with the original interaction matrix structure to update the interaction matrix. Third, we propose Divide-Parallel-Combine (DPC), a parallel acceleration framework utilizing truncated matrix factorization. DPC decomposes large-scale matrix tasks into smaller subproblems, solves each subproblem in parallel using truncated matrix factorization, and combines the solutions utilizing randomized matrix approximation techniques. Extensive experiments on three datasets showed that MVPFDPC is a highly accurate and fast predictive model for miRNA-lncRNA interactions.
微小RNA（miRNA）与长链非编码RNA（lncRNA）之间的相互作用在基因表达调控网络中具有关键作用，对细胞功能和疾病发展有显著影响。因此，推断潜在的miRNA-lncRNA相互作用至关重要。随着数据量的增加，模型的计算复杂度也在增长，这就需要高效的方法来快速预测分子相互作用。此外，大多数现有模型依赖单一的相似性度量，在计算建模中引入了偏差。基于相似性的拓扑结构的潜力也尚未得到充分探索或利用。在本研究中，我们提出了一种利用多视图投影融合和随机并行矩阵分解（MVPFDPC）的miRNA-lncRNA相互作用快速预测模型。首先，我们构建了一个全面的miRNA/lncRNA相似性概况。其次，我们开发了一种多视图投影融合技术，该技术利用来自不同空间的多个相似性网络拓扑结构以及原始相互作用矩阵结构来更新相互作用矩阵。第三，我们提出了划分-并行-合并（DPC），这是一种利用截断矩阵分解的并行加速框架。DPC将大规模矩阵任务分解为较小的子问题，使用截断矩阵分解并行解决每个子问题，并利用随机矩阵近似技术合并解决方案。在三个数据集上进行的大量实验表明，MVPFDPC是一种用于miRNA-lncRNA相互作用的高精度快速预测模型。
