Meta-analysis of fecal metagenomes reveals global viral signatures and its diagnostic potential for colorectal cancer and adenoma

Introduction Gut microbiome plays an important role in maintaining human health. Although mounting evidence has revealed the critical function of the gut bacteriome in the progression of CRC, the contribution of gut viral community to CRC is rarely studied. Objectives The present study aimed to reveal the gut virome signatures of colorectal adenoma patients and CRC patients and decipher the potential viral markers to build clinical predictive models for diagnosis. Methods 1,282 available fecal metagenomes data from 9 published CRC studies were collected. A new virus database was constructed based on a reference-independent virome approach for further analysis. Viral markers were filtered by statistical methods and used to build machine learning models such as Random Forest and Least Absolute Shrinkage and Selection Operator (LASSO) to distinguish patients from controls. New fecal samples were collected to validate the generalization of predictive model. Results The gut viral composition of CRC patients was drastically altered compared with healthy, as evidenced by changes in several Siphoviridae viruses and a reduction of Microviridae, whereas the virome variation in adenoma patients was relatively low. The viral markers contained the phages of Porphyromonas , Fusobacterium , Hungatella , and Ruminococcaceae . In 9 cohorts and independent validation cohorts, a random forest (RF) classifier and LASSO model got the optimal AUC 0.830 and 0.906, respectively. While the gut virome analysis of adenoma patients identified 88 differential viruses and achieved an optimal AUC of 0.772 for discriminating patients from controls. Conclusion Our findings demonstrate the distinctly different composition of gut virome between healthy controls and CRC patients, and highlight the potential of viral markers for clinical diagnosis. ### Competing Interest Statement The authors have declared no competing interest.