The inventors found that metabolic gene signatures of human pan-tumor associated myeloid cells correlate with patient survival and cancer cell mitotic index. Indeed by crossing i) metabolic genes expression from single cell RNA-seq data of myeloid cell subsets, and ii) their functional metabolism by SCENITH, a method to determine in parallel the phenotype and metabolic state of the immune, stromal and tumor cells, the inventors identified glycolytic and respiratory metabolic gene signatures which predict survival time in patients. They found that in different human tumors, the glycolytic signature was associated with significantly reduced patient survival, while a respiratory gene signature correlated with increased survival. Moreover they demonstrated that the presence of glycolytic myeloid cells in the tumor correlates with malignancy.rnThus the present invention relates to new gene signatures that are suitable for predicting survival time in patients suffering from cancer and for diagnosing malignancy tumor.rn