SPRDA: a matrix completion approach based on the structural perturbation to infer disease-associated Piwi-Interacting RNAs

Abstract

<jats:title>Abstract</jats:title><jats:p>Emerging evidence suggests that PIWI-interacting RNAs (piRNAs) are one of the most influential small non-coding RNAs (ncRNAs) that regulate RNA silencing. piRNA and PIWI proteins have been confirmed for disease diagnosis and treatment as novel biomarkers due to its abnormal expression in various cancers. However, the current research is not strong enough to further clarify the functions of piRNA in cancer and its underlying mechanism. Therefore, how to provide large-scale and serious piRNA candidates for biological research has grown up to be a pressing issue. The main motivation of this work is tantamount to fill the gap in research on large-scale prediction of disease-related piRNAs. In this study, a novel computational model based on the structural perturbation method is proposed, called SPRDA. In detail, the duplex network is constructed based on the piRNA similarity network and disease similarity network extracted from piRNA sequence information, Gaussian interaction profile kernel similarity information and gene-disease association information. The structural perturbation method is then used to predict the potential associations on the duplex network, which is more predictive than other network structures in terms of structural consistency. In the five-fold cross-validation, SPRDA shows high performance on the benchmark dataset piRDisease, with an AUC of 0.9529. Furthermore, the predictive performance of SPRDA for 10 diseases shows the robustness of the proposed method. Overall, the proposed approach can provide unique insights into the pathogenesis of the disease and will advance the field of oncology diagnosis and treatment.</jats:p>

Databases and computational methods for the identification of piRNA-related molecules: A survey

“…For example, deep feedforward neural networks (DFNN) [30] and CNN [19] have been employed to learn feature representations of piRNA sequences, whereas GAT [23] has been built to learn the structural topology of piRNA–disease association networks for piRNA identification. Meanwhile, some studies also retained handcrafted features, which helped alleviate the problem of data sparsity [56] , [58] , [59] . Fig.…”

“… The disease GIP similarity is calculated by the known association matrix between piRNA and diseases. piRNA sequence similarity [59] Where: Where denotes the sequence of piRNA, and denotes the frequency score for each k-mer in i-th . Where is the mean of of each piRNA sequence.…”

“…Two types of similarity knowledge are employed to build a piRNA–disease association network: piRNA similarity and disease similarity. The first employs piRNA sequence similarity [59] , which works by calculating the similarity of two piRNA sequences via the k-mer algorithm and piRNA Gaussian Interaction Profile (GIP) kernel similarity [132] . GIP similarity measures the similarity of interaction patterns of biological entities [133] by extracting their associated features via Gaussian kernel functions and is widely utilized for similarity measurements of miRNA [133] , [134] , [135] , circular RNA (circRNA) [136] and long ncRNA (lncRNA) [137] , [138] .…”

“… Simply integrating different feature representation methods may cause inconsistency and conflict, which affects the stability of the model. None 0 SPRDA [59] Matrix Completion Disease semantic similarity, Disease functional similarity, piRNA sequence similarity and GIP similarity of piRNA and diseases 501 22 1212 0.9529 Using structural perturbation method and multi-source information for prediction, effectively avoiding the problem of lacking high-quality negative samples. Because the structural perturbation method is based on randomly selected edges to evaluate the predictability of the network, it may cause some important edges to be ignored or some irrelevant edges to be considered.…”

PDA-PRGCN: identification of Piwi-interacting RNA-disease associations through subgraph projection and residual scaling-based feature augmentation

“…GAPDA treated each known piRNA-disease association pair as a node in their reconstructed graph and employed graph attention network to make representation learning [12]. SPRDA applied piRNA/ disease similarity network to form a duplex network, then predicted PDAs as a matrix completion problem by structural perturbation algorithm [13].…”