Prediction of anti-breast cancer compound activity based on Gradient Boosting Decision Tree ensemble learning
Haoyun Yuan, Yang Liu, Jingze Du, Ziqian Zhou · 2021
The number of breast cancer cases worldwide has risen sharply in the past 20 years. In 2000, the total number of cases was 1.05 million. By 2018, the number had increased to 2.09 million, an increase of 99.05 percent, with an average annual growth rate of more than 5 percent. Breast cancer has gradually become the most common cancer for women, so the development of anti-breast cancer drugs has become a hot topic in the current medical field. In this paper, a series of compound descriptors and their biological activity data were collected for ERα, a target associated with breast cancer, and compound activity was predicted by establishing a compound activity prediction model. According to the compound descriptors, more important features were selected. The Quantitative structure-activity Relationship (QSAR) model of compounds was constructed, and then the GBDT algorithm was used to predict the model. Finally, through comparative analysis of results, the method was fast and accurate. It plays a key role in drug research.