AEMDA

AEMDA predicts associations between microRNAs (miRNAs) and diseases by integrating disease semantic similarity, miRNA functional similarity, and heterogeneous related interaction data to learn dense embeddings via a deep autoencoder for computational identification of disease–miRNA interactions.

Key Features:

• Learning-Based Methodology: Integrates disease semantic similarity, miRNA functional similarity, and heterogeneous related interaction data to derive dense, high-dimensional embeddings of diseases and miRNAs.
• Deep Autoencoder Architecture: Employs a deep autoencoder to learn latent representations and capture complex patterns in high-dimensional biological data without requiring negative samples for training.
• Reconstruction Error as Predictive Measure: Uses the autoencoder reconstruction error as a scoring metric to predict potential disease–miRNA associations, avoiding the need for extensive negative sampling.

Scientific Applications:

• Disease Diagnosis and Prevention: Supports identification of candidate disease-associated miRNAs to inform early diagnosis and studies of disease mechanisms.
• Research and Development: Enables exploration and prioritization of novel miRNA–disease interactions for genomics and bioinformatics research.

Methodology:

Integrates disease semantic similarity, miRNA functional similarity, and heterogeneous related interaction data to construct representations of diseases and miRNAs, trains a deep autoencoder without negative samples, and uses autoencoder reconstruction error to score and predict disease–miRNA associations.