MSCNE:Predict miRNA-Disease Associations Using Neural Network Based on Multi-Source Biological Information.

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• MicroRNAs (miRNAs) play a significant role in human diseases.
• Experimental methods for identifying miRNA-disease associations are costly and inefficient.
• There is a need for high-efficiency computational approaches.

Purpose of the Study:

• To develop a high-efficiency algorithm for predicting miRNA-disease associations.
• To integrate diverse biological source information for improved prediction accuracy.
• To reduce the cost and blindness associated with experimental methods.

Main Methods:

• A novel algorithm combining a convolutional neural network (CNN) feature extractor and an extreme learning machine (ELM) classifier was proposed.
• Multi-source biological information, including semantic similarity of diseases, Gaussian interaction profile kernel similarity (miRNA, disease, lncRNA, EFs), and miRNA similarities (target, sequence, family, function), were fused.
• An autoencoder (AE) was used for dimensionality reduction, followed by CNN for deep feature extraction and ELM for prediction.

Main Results:

• The proposed multi-biological source information (MSCNE) model achieved an average AUC value of 0.9630.
• The MSCNE model demonstrated superior performance compared to other classic classifiers, feature extractors, and existing algorithms.
• The algorithm effectively predicted the correlation between miRNA and disease.

Conclusions:

• The MSCNE algorithm is effective and efficient for predicting miRNA-disease associations.
• Integrating multi-source biological information significantly enhances prediction accuracy.
• This computational approach offers a cost-effective alternative to traditional experimental methods.