It holds great promises to precisely stratify cancer subtypes to improve cancer diagnosis, therapy and prognosis. In the past, the diagnosis of pathological subtypes mainly relied on hematoxylin-eosin staining and immunohistochemistry. With the development of sequencing technologies, genotype and phenotype analysis of individuals has become possible and precision medicine is on the rise in healthcare. As different tumor subtypes have different cell-of-origin, risk factors and clinical phenotypes, they generate unique combinations of mutation types, termed “Mutational Signatures”. Herein, using the exome sequencing data from The Cancer Genome Atlas (TCGA), we evaluated the utility of mutational landscape for differentiating cell-of-origin within three common cancers (kidney, lung and esophageal cancers). We found that mutational signatures predicted histological subtypes of kidney cancers, clear cell renal cell carcinoma (KIRC) vs. chromophobe renal cell carcinoma (KICH), which had different cell-of-origin, with 100% accuracy (95% CI: 0.93-1.00). The mutational signatures also predicted histological subtypes of lung cancers (lung adenocarcinoma vs. lung squamous cell carcinoma) and esophageal cancers (esophageal adenocarcinoma vs. esophageal squamous cell carcinoma) with 78% (95% CI: 0.66-0.86) and 84% accuracy (95% CI: 0.60-0.97), respectively. Collectively, mutational signatures-based subtyping is good at pathological classification, personalized diagnosis, especially early detection for common cancers.