Background: The development of non-invasive biomarkers has the potential to revolutionize clinical care for colorectal cancer (CRC) patients. The presence of bacteria in CRC tumor biopsies has been shown to contribute to CRC development. In a previous study, our group showed some intra-tumor microbes in CRC tumor biopsies correlated with overall survival in CRC patients. However, the correlations between microbes in tumor vs blood, and between non-invasive serum marker carcinoembryonic antigen (CEA) and microbes are unknown. We hypothesize that tumor microbes will also be found in blood, and that CEA will correlate with certain microbes. Methods: We obtained RNA-seq data from CRC tumor biopsies from patients treated at The Ohio State University Comprehensive Cancer Center as part of the Oncology Research Information Exchange Network (ORIEN). Reads were aligned to human and exogenous genomes using TopHat2 and Kraken2/Bracken, respectively. RNA-seq from CRC tumor biopsies as well as peripheral blood at the Cancer Genome Atlas (TCGA) consortium were processed by the same method. Results: The analyzed ORIEN cohort included 93 CRC patients with an age range from 30-83 years, 60.2% male, 87.1% adenocarcinoma, and 47.3% with metastatic CRC. The TGCA cohort included 495 CRC patients with an age range from 31-90 years, 53.3% male, 85.1% adenocarcinoma, and 15.5% with metastatic CRC. Over fifteen exogenous phyla (including bacteria, viruses, fungi) were observed in both ORIEN and TCGA cohorts. Several of the samples were dominated by viral sequences while others by bacteria, suggesting considerable tumor microbiome heterogeneity. Evaluation of the fraction of microbes in tumor and blood showed that nearly all the microbes found in blood (97.6%) were also observable in tumor in the TCGA cohort. Microbial abundances of various taxa, including Fusobacterium, significantly correlated between blood and tumor. Several bacteria including members of the genera Bacillus and Staphylococcus were positively associated with tumor stage (metastatic vs non-metastatic), but microbial relative abundances were not correlated with the location of tumor in colon (right, left, transverse colon). Certain microbial species from the ORIEN cohort were found to positively correlate with CEA, (including from the genera Fusobacterium, Lactobacillus, Pseudomonas, Vibrio, Clostridium) and these associations remained when adjusted for alcohol and smoking by multivariate analysis. Conclusions: Nearly all the microbes found in blood were found in tumor and abundances of various taxa were significantly correlated, suggesting that blood-based cancer microbiome analysis has great potential. Serum CEA has a low diagnostic ability when used alone, but combining this with blood microbiome could improve diagnostic/prognostic utility as a non-invasive biomarker.