Traditional Chinese medicine studies for AD based on Logistic Matrix Factorization and Similarity Network Fusion

Alzheimer's disease (AD) is a neurological disorder with complicated pathogenesis. The approved AD drugs cannot block or reverse the pathologic progression of AD. In this study, a method based on Logistic Matrix Factorization and Similarity Network Fusion (MLMFSNF) is proposed for screening out the Traditional Chinese medicines (TCMs) and active ingredients targeting AD targets. Firstly, TCMs for AD are obtained from the AD drug reviews, the active ingredients and related targets are collected from various databases. Secondly, the similarity networks are constructed by an improved Gaussian interaction profile kernel and other metrics for active ingredients and targets. The synthesized similarity networks are integrated based on similarity network fusion (SNF). The filling of missing activity ingredient-target associations is achieved by the logistic matrix factorization. Finally, the association scores between active ingredients and targets are calculated and ranked. We screen out TCMs for AD by the logistic function transformation. The results demonstrated that the MLMFSNF algorithm is effective for association prediction.