Description of the Method
The method DNAm_wICA (m_WIC) has been intially described in the deconbench project1 and published in BioRxiv. It is an unsupervised method working on methylome data. It follows the pipeline implemented in the R package medepir (HADACA consortium et al., 2020).
The code is using the R package medepir, a pipeline for DNA methylation deconvolution2.
Feature selection
The feature selection is performed by medepir::feature_selection for keeping highly variable probes (variance higher than 0.02). T
Deconvolution
The method EDec (Onuchic et al., 2016) is used for the deconvolution part, with the function medepir::Edec and all the selected probes as “infloci” parameter.
Note that the Edec method offers the possibility to select biologically revelant probes as selected features, although it is not implemented here to keep it generic.
Results
The program returns an estimated proportion matrix and the corresponding estimated cell-type specific profiles matrix.
Descriptors
Omics_type = methylome
Cancer_type = all
Method_type = unsupervised
Cell_type = all