Abstract

Info

Title:: Predicting the proportions of tumor-infiltrating immune cells based on partial DNA methylation

Author(s):: SONG Chunhui1; 2; QIN Yufang1; 2; CHEN Ming1; 2; 1. College of Information Technology, Shanghai Ocean University, Shanghai 201306, China 2. Key Laboratory of Fisheries Information, Ministry of Agriculture and Rural Affairs of the Peoples Republic of China, Shanghai 201306, China

Keywords:: Keywords: cancer DNA methylation non-negative matrix factorization immune cell

Abstract:: Abstract: Understanding the proportions of immune cells in tumor tissues has important implications for the diagnosis and prognostic treatment of tumor patients, and DNA methylation can be used as a cellular molecular signature to quantify multiple cell types in complex tumor mixtures. Herein a prediction model for the proportions of tumor-infiltrating immune cells is established based on a non-negative matrix factorization algorithm combined with partial DNA methylation. When predicting the proportions of cell types in tumor mixtures, the known DNA methylation level matrix that is significantly different for each immune cell type is selected as the feature matrix, and then the matrix and the DNA methylation level of tumor mixtures are input into the non-negative matrix factorization algorithm for predicting the proportions of cell types. Simulation experiments at different levels of noises and tumor contents verify that the proportions of immune cells predicted by the model are highly consistent with the real proportions, with small root-mean-square error. Applying the model to 4 kinds of real tumor samples yielded the highest levels of infiltrating immune cells in each sample, and the results are consistent with previous biological findings by other scholars. The proposed model can accurately predict the proportions of cell types that make up tumor mixtures, providing a new approach for tumor immunology research.