Independent component analysis of cancer transcriptomes: optimization of parameters and improvement of interpretability. In Computer Technologies and Data Analysis (CTDA'2020) : Materials of II International Scientific-Practical Conference, Minsk, 23-24 April, 2020. (Book Chapter)

Here we estimated the optimal parameters of the independent component analysis method regarding the tasks of identification of glioblastoma and pancreatic cancer subtypes, prediction of patient survival and characterization of active biological processes. Analysis of deconvolution results of bulk and single-cell data allows sharing annotation between highly correlated components and improving interpretability of the results