Machine Learning for Predicting lncRNA–Disease Associations Related to Periodontitis-Associated Alveolar Bone Loss
DOI:
https://doi.org/10.35699/2178-1990.2026.59071Palavras-chave:
periodontitis, machine learning, long noncoding RNAResumo
Introduction: Periodontitis-associated alveolar bone loss is characterized by the gradual destruction of alveolar bone and is regulated by intricate genetic and immunological processes. Although non-coding RNAs (lncRNAs) have become important regulatory components in these processes, little is known about how they relate to disease, especially in periodontitis. Although there are few comparative evaluations of various architectures, graph-based machine learning presents promising tools for modeling such biological associations.
Aim: To predict and reconstruct lncRNA–disease associations related to periodontitis-associated alveolar bone loss, this study compares and assesses the performance of Graph Auto-Encoder (GAE) and Adversarial Regularized Graph Auto-Encoder (ARGAE).
Materials and Methods: A heterogeneous bipartite graph with ncRNA and disease nodes was created using experimentally verified data from the LncRNADisease database. A straightforward one-hot scheme was used to encode node features. A two-layer Graph Convolutional Network (GCN) encoder was used in the GAE and ARGAE models. An extra discriminator for adversarial regularization was incorporated into ARGAE. The models were evaluated using embedding coherence (cosine similarity, t-SNE visualization), clustering quality (Silhouette Score, ARI), and link prediction metrics (ROC-AUC, Average Precision).
Results: While ARGAE added embedding diversity that allowed for the exploratory identification of less evident lncRNA–disease relationships, GAE showed a strong reconstruction capacity with superior clustering interpretability. The study found clear trade-offs between latent space generalization and prediction accuracy.
Conclusions: This comparative analysis fills a substantial knowledge gap by examining the structural and predictive abilities of GAE and ARGAE in modeling lncRNA–disease associations related to alveolar bone loss commonly observed in periodontitis. The results provide useful information for scientists using graph-based models to search for lncRNA biomarkers and aid in advancing periodontology precision diagnostics and treatments.
Referências
Sun J, Shi H, Wang Z, Zhang C, Liu L, Wang L, et al. Inferring novel lncRNA-disease associations based on a random walk model of a lncRNA functional similarity network. Mol Biosyst. 2014;10(8):2074-2081.
Ramesh A, Varghese SS, Doraiswamy JN, Malaiappan S. Herbs as an antioxidant arsenal for periodontal diseases. J Intercult Ethnopharmacol. 2016;5(1):92-96.
Panda S, Sankari M, Satpathy A, Jayakumar D, Mozzati M, Mortellaro C, et al. Adjunctive Effect of Autologus Platelet-Rich Fibrin to Barrier Membrane in the Treatment of Periodontal Intrabony Defects. J Craniofac Surg. 2016;27(3):691-696.
Kaarthikeyan G, Jayakumar ND, Padmalatha O, Sheeja V, Sankari M, Anandan B. Analysis of the association between interleukin -1β (+3954) gene polymorphism and chronic periodontitis in a sample of the south Indian population. Ind J Dent R. 2009;20(1):37-40.
Xuan P, Fan M, Cui H, Zhang T, Nakaguchi T. GVDTI: graph convolutional and variational autoencoders with attribute-level attention for drug-protein interaction prediction. Brief Bioinform. 2022;23(1):bbab453.
Lei S, Lei X, Liu L. Drug repositioning based on heterogeneous networks and variational graph autoencoders. Front Pharmacol. 2022;13:1056605.
Yuan Y, Zhang L. Long non-coding RNA regulates bone metabolism via Wnt/β-catenin signaling pathway. Sheng Wu Gong Cheng Xue Bao. 2021;37(7):2342-2350.
Gao MM, Cui Z, Gao YL, Wang J, Liu JX. Multi-Label Fusion Collaborative Matrix Factorization for Predicting LncRNA-Disease Associations. IEEE J Biomed Health Inform. 2021;25(3):881-890.
Zhang Y, Ye F, Xiong D, Gao X. LDNFSGB: prediction of long non-coding rna and disease association using network feature similarity and gradient boosting. BMC Bioinformatics. 2020;21(1):377.
Wang B, Liu R, Zheng X, Du X, Wang Z. lncRNA-disease association prediction based on matrix decomposition of elastic network and collaborative filtering. Sci Rep. 2022;12(1):12700.
Wang L, Shang M, Dai Q, He PA. Prediction of lncRNA-disease association based on a Laplace normalized random walk with restart algorithm on heterogeneous networks. BMC Bioinformatics. 2022;23(1):5.
Ray S, Lall S, Mukhopadhyay A, Bandyopadhyay S, Schönhuth A. Deep variational graph autoencoders for novel host-directed therapy options against COVID-19. Artif Intell Med. 2022;134:102418.
Wang C, Yuan C, Wang Y, Chen R, Shi Y, Zhang T, et al. MPI-VGAE: protein-metabolite enzymatic reaction link learning by variational graph autoencoders. Brief Bioinform. 2023;24(4):bbad189.
Salha G, Hennequin R, Remy JB, Moussallam M, Vazirgiannis M. FastGAE: Scalable graph autoencoders with stochastic subgraph decoding. Neural Netw. 2021;142:1-19.
Niknam G, Molaei S, Zare H, Pan S, Jalili M, Zhu T, et al. DyVGRNN: DYnamic mixture Variational Graph Recurrent Neural Networks. Neural Netw. 2023;165:596-610.
Chen G, Wang Z, Wang D, Qiu C, Liu M, Chen X, et al. LncRNADisease: a database for long-non-coding RNA-associated diseases. Nucleic Acids Res. 2013;41(D1):D983-6.
Xiao X, Zhu W, Liao B, Xu J, Gu C, Ji B, et al. BPLLDA: Predicting lncRNA-Disease Associations Based on Simple Paths With Limited Lengths in a Heterogeneous Network. Front Genet. 2018;9:411.
Li Y, Li J, Bian N. DNILMF-LDA: Prediction of lncRNA-Disease Associations by Dual-Network Integrated Logistic Matrix Factorization and Bayesian Optimization. Genes (Basel). 2019;10(8):608.
