QSICPM: A Novel Quantitative Semantic Interaction Calculation Prediction Method for Drug Repositioning Based on Biomedical Literature Semantic Data (Preprint)

Abstract

The process of developing new drugs is very tortuous. Bringing new drugs to the market requires billions of dollars in investment, which takes an average of about 13-15 years. In order to overcome these difficulties, more and more companies and pharmaceutical companies have begun to adopt the strategy of “repositioning drugs” instead of new drug development.

Traditional drug repositioning methods often focus on relationships between entities, ignoring the semantic component of relationships. Therefore, we propose a new drug repositioning method to calculate the impact of pathogenic entities on disease mechanisms by quantifying semantic interactions.

The QSICPM proposed in this paper divides the relevant interactions into three quantitative calculation layers based on the cause of disease. Representing interactions between the same type of pathogenic entities, interactions between different types of pathogenic entities, and pathogenic entity – drug interactions, respectively. QSICPM calculates the influence of drugs on the disease mechanism by utilizing the positive semantic relationships in each layer of the quantitative calculation process. And the gene prioritization sorting method and protein prioritization sorting method are proposed to sort the calculation results. The higher the ranking of the drugs in the results, the more likely the drug becomes an effective drug for the disease.

We used QSICPM to perform drug repositioning experiments on Parkinson's disease, breast cancer, and Alzheimer's disease. The experimental results predicted 881 potential drugs or pharmacological substances for PD disease, 830 potential drugs or pharmacological substances for BC disease, and 1180 potential drugs or pharmacological substances for AD disease. What’s more, the result set was sorted according to gene prioritization sorting and protein prioritization sorting. In the top 25 parts of the ranking, the average precision of the three results reached 68%, 75%, and 64%, respectively. The accuracy and applicability of QSICPM were verified.

This paper proposes a new research method for drug repositioning based on semantic relationship quantitative calculation. The performance of the QSICPM method was verified by drug repositioning experiments for Parkinson's disease, Breast cancer, and Alzheimer's disease. The results prove that QSICPM is a drug repositioning method with strong prediction precision and wide applicability.