Comparison of Chemometric Explorative Multi-Omics Data Analysis Methods Applied to a Mechanistic Pan-Cancer Cell Model

Abstract

The analysis of single cell multi-omics data is a complex task, and many explorative data analysis methods are being used to draw information from such data. This paper compares several of these methods to visualize the output of a mechanistic model under various simulated conditions. The analysis methods include PCA, PARAFAC, ASCA, MASCARA, COVSCA, P-ESCA, and PE-ASCA. These techniques, applied to high-dimensional data such as gene expression and protein levels, assess correlations across time series and experimental conditions. The study uses a complex mechanistic model of MCF10A cancer cells, simulating interactions between signaling pathways related to cell growth and division. Results show that while methods like PCA PARAFAC and ASCA reveal time-dependent variations in protein data, mRNA data exhibit minimal systematic variation. MASCARA offers unique insights by identifying genes linked to specific pathways. This work highlights the potential and limitations of various data analysis methods in understanding multi-omics data, particularly in single-cell contexts where experimental variation and stochastic processes complicate interpretation.