Identification of Methylation Signatures and Rules for Sarcoma Subtypes by Machine Learning Methods

Figure 1

Flow chart of the whole analysis process. Methylation site features from sarcoma samples were analysed by Boruta, and remaining features were ranked in accordance with their relevance with three feature ranking algorithms, namely, LASSO, LightGBM, and MCFS. Subsequently, three ordered feature lists were fed into the incremental feature selection computational framework to access essential methylation sites, models with high performance, and quantitative classification rules.