ANMF

ANMF predicts novel drug–disease associations for computational drug repositioning by combining drug–drug and disease–disease similarities with autoencoder-derived latent features and generalized matrix factorization.

Key Features:

• Integration of Similarities: Integrates drug–drug and disease–disease similarities to enrich representations and mitigate data sparsity.
• Autoencoder-Based Feature Extraction: Employs a variant of autoencoders to extract latent features from drug and disease data.
• Collaborative Filtering with GMF: Integrates Generalized Matrix Factorization (GMF) to perform collaborative filtering using extracted latent features.
• Negative Sampling Techniques: Incorporates negative sampling during training to reduce overfitting and improve generalization.
• Implementation: Implemented in Keras with Theano as the backend.

Scientific Applications:

• Computational Drug Repositioning: Predicts novel therapeutic indications for existing drugs by identifying drug–disease associations.
• Benchmarking and Validation: Validated on the Gottlieb and Cdataset datasets and reported to outperform existing state-of-the-art methods in predicting drug–disease interactions.

Methodology:

Computational steps explicitly include integrating drug–drug and disease–disease similarities, applying an autoencoder variant for latent feature extraction, combining latent features with Generalized Matrix Factorization for collaborative filtering, and using negative sampling during training, with implementation in Keras/Theano and validation on Gottlieb and Cdataset.