Machine-Learning Based Drug and Supplement Design Platform for Chronic Disease Control

Abstract

The emerging modern medical technology has provided treatment for a variety of major human diseases (including cancer, cardiovascular and cerebrovascular diseases, diabetes, etc.). However, the long-term control of these chronic diseases remains a challenge. These chronic diseases are the main cause of death and a major burden on public health. One key question in chronic disease management is how to develop efficacious and reliable drugs based on our emerging understanding of disease-related gene expression (from genes to drugs). In contrast, specific medical treatments such as surgery or chemotherapy of cancer often possess a small period over the entire course of chronic disease management, meanwhile the therapy outcomes of chronic diseases also largely depend on the long-term lifestyles of the patients, such as mood and diet. Given the large quantity and complexity of traditional medical data, it is difficult to manually evaluate the impact of diets or nutrition on chronic disease control (from drugs to supplements/nutrition). This project aimed to systematically analyze disease-related gene expression data for drug development and to specify a dietary plan for specific chronic disease recovery from health big data and individualized data of patients. Using multiple gene-to-drug algorithms combined with Random Forest machine learning tools, I have discovered decamethonium bromide as a broad-spectrum anti-cancer drug targeting currently “undruggable” oncogenes and tumor suppressors and investigated honeysuckle as a potential supplement. CCS Concepts· Applied computing · Life and medical sciences · Computational biology