Abstract
Small nucleolar RNAs (snoRNAs) are essential in biological processes and are linked to diseases such as cancer and neurodegenerative disorders. Identifying snoRNA-disease associations is challenging due to limited and imbalanced data. To address this, we propose a Siamese Neural Network (SNN) framework that utilizes features derived from 4 -mer sequence patterns and DNABERT for snoRNAs, along with binary profiles of diseases based on semantic relationships and phenotype associations. The SNN processes these features through parallel sub-networks, generating embeddings in a shared space to predict associations. Trained with binary crossentropy and Kullback-Leibler divergence losses, our model achieves superior performance compared to existing methods, with an AUC of 0.91 and accuracy of 0.91. These results demonstrate the effectiveness of our approach in identifying meaningful snoRNA-disease relationships, offering a scalable and efficient tool for biomarker discovery and therapeutic research.
Milad Besharatifard
Ph.D. student in Computer Science specializing in Soft Computing and Artificial Intelligence
Fatemeh Zare-Mirakabad
Associate Professor
My research interests include bioinformatics, computational biology and artificial intelligence.