Table 1 Overview on published methods to infer tumor purity based on WES/SNP array, gene expression arrays and methylation arrays
Publication | Method name | Statistical framework/technique | Datasets used for establishing the method/ validation of the method | Datatypes which can be used as input |
|---|---|---|---|---|
Carter et al. [4] | ABSOLUTE | Tumor purity inference based on somatic copy number aberrations in SNP arrays | TCGA | WES data/SNP array |
Yoshihara et al., 2013 [8] | ESTIMATE | Comparison of various published gene sets to delineate a) immune signature b) stromal signature - based on these signatures, calculation of purity score | TCGA | Affymetrix gene expression array data |
Aran, D. et al. 2015 [7] | LUMP (leukocytes unmethylation for purity) | Averaging of the methylation values 44 CpG sites, known to be hypomethylated in immune cells | TCGA | 450 K methylation array data |
InfiniumPurify | Tumor purity estimation: (PMID:28122605) comparison of tumor and normal samples to identify DMC (differentially methylated CpG sites) between tumors and an universal set of normal samples in the TCGA dataset followed by kernel density estimation to obtain tumor purity | TCGA | 450 K Methylation array data | |
Benelli et al. 2018 [6] | PAMES (Purity Assessment from clonal MEthylation Sites) | - Calculation of average methylation values per CpG island from TCGA entities. - Calculation of the Area under the curve for the ROC curves of each CpG island: If AUC < 0.2 or AUC > 0.8 a certain CpG site was considered discriminatory and taken into the model. - Tumor purity estimate based on the median of hypomethylated and hypermethylated sites | TCGA (generation of the model), Comparison to other TCGA samples and one additional dataset (333 prostata adenocarcinomas) | 450 K methylation array data |