MMP3C: an in-silico framework to depict cancer metabolic plasticity using gene expression profiles.

Metabolic plasticity enables cancer cells to meet divergent demands for tumorigenesis, metastasis and drug resistance. Landscape analysis of tumor metabolic plasticity spanning different cancer types, in particular, metabolic crosstalk within cell subpopulations, remains scarce. Therefore, we proposed a new in-silico framework, termed as MMP3C (Modeling Metabolic Plasticity by Pathway Pairwise Comparison), to depict tumor metabolic plasticity based on transcriptome data. Next, we performed an extensive metabo-plastic analysis of over 6000 tumors comprising 13 cancer types. The metabolic plasticity within distinct cell subpopulations, particularly interplay with tumor microenvironment, were explored at single-cell resolution. Ultimately, the metabo-plastic events were screened out for multiple clinical applications via machine learning methods. The pilot research indicated that 6 out of 13 cancer types exhibited signs of the Warburg effect, implying its high reliability and robustness. Across 13 cancer types, high metabolic organized heterogeneity was found, and four metabo-plastic subtypes were determined, which link to distinct immune and metabolism patterns impacting prognosis. Moreover, MMP3C analysis of approximately 60 000 single cells of eight breast cancer patients unveiled several metabo-plastic events correlated to tumorigenesis, metastasis and immunosuppression. Notably, the metabolic features screened out by MMP3C are potential biomarkers for diagnosis, tumor classification and prognosis. MMP3C is a practical cross-platform tool to capture tumor metabolic plasticity, and our study unveiled a core set of metabo-plastic pairs among diverse cancer types, which provides bases toward improving response and overcoming resistance in cancer therapy.