Raman spectra and gene expression correspondences reveal global stoichiometry conservation architecture in cells

Abstract

Cellular Raman spectra are optical signatures reflecting comprehensive molecular composition of cells. It has been shown recently that condition-dependent cellular transcriptome profiles are inferable from the changes in cellular Raman spectra. However, it remains elusive whether the other layers of omics are also inferable from Raman spectra and what constraints in gene expression profiles enable such inference. Here we show that proteome profiles ofEscherichia coliacross diverse conditions are predictable from cellular Raman spectra using simple ordinary linear regression. Employing methods in graph theory, we characterize core and peripheral stoichiometrically-conserved groups in the proteomes and reveal that major changes in cellular Raman spectra reflect low-dimensional proteome structure constructed from the global stoichiometry conservation relations. Mathematical analysis suggests a proportionality between stoichiometry conservation strength and condition-specificity of gene expression, which turns out universal across omics classes and organisms. Furthermore, stoichiometry conservation strength correlates with the essentiality and evolutionary conservation of each gene. These results provide an omics-level interpretation of cellular Raman spectra and uncover an architectural principle of global omics profiles.