NIC AU Research: Variational Graph Neural Network Predicts LncRNA–Disease Associations

NIC AU congratulates Trần Bình Minh (NIC AU), together with co-authors Luong Hai Nguyen and Dang Hung Tran, on a new book chapter accepted into the ACIIDS 2026 proceedings (Springer, Communications in Computer and Information Science): “A Variational Graph Neural Network with Multi-view Integration for Predicting LncRNA–Disease Associations.”

The Problem

Long non-coding RNAs (lncRNAs) play a critical role in disease biology, but experimentally validating which lncRNAs are linked to which diseases is slow and expensive. Existing computational approaches often struggle with sparse, noisy biological networks.

The Approach

The authors propose an end-to-end deep learning framework that combines attention-weighted multi-view graph convolutional networks, a variational graph autoencoder, and a gated multilayer perceptron — unifying similarity networks of lncRNAs, diseases and miRNAs into a single heterogeneous graph.

The model achieved strong performance on a primary benchmark (AUC 0.9584, AUPR 0.9590), with a second independent dataset confirming its generalisability — offering a robust new tool for prioritising lncRNA-disease candidates and accelerating therapeutic discovery.

Read more on the Resources page or the published chapter on Springer Link.