Tumor purity predicted by statistical methods

Cancer is one of the major health problems for human and patients with advanced stage cancer have ultra-low 5-year survival rate. Therefore, improving early cancer detection accuracy rate is necessary. The measurement of tumor purity plays an important role in the early diagnosis of cancer and the tracking of patient’s condition. In addition, how to estimate tumor purity accurately is also a significant biological problem. Predicting tumor purity based on DNA methylation is a prevalent method. In this article, we discuss using different linear regression models to estimate tumor purity. Two main questions need to be solved to accurately estimate tumor purity. We first discussed the features selection methods in biological and mathematical aspect which reduce the collinearity between cell types and make deconvolution more stable. In addition, we also compared 4 different linear regression model usually used in this question and analyzed their advantages and disadvantages respectively.