Background: Timely diagnosis and intervention of colorectal cancer (CRC) at curable stages is essential for improving patient survival. Stool samples carry exfoliation of intestinal epithelium, therefore providing excellent opportunity for non-invasive diagnosis of CRC as well as precancerous lesions. In this study, we aimed to conduct multi-dimensional analysis of fecal DNA and investigate the utility of different types of biomarkers for CRC detection. Method: In this case-control study, we performed comprehensive analyses of the genomic, epigenomic, and metagenomic features of fecal DNA from CRC patients, individuals with advanced precancerous lesions (APLs) and controls. DNA methylation markers were identified by whole genome bisulfite sequencing of paired colorectal cancer and normal tissues. A multi-gene fecal DNA methylation test was then developed based on three marker genes (SDC2, ADHFE1 and PPP2R5C) using quantitative methylation-specific PCR (qMSP), and validated on fecal DNA samples. Genomic mutation profiles as well as microbiome signatures of fecal DNA were analyzed using high-throughput sequencing. Results: The methylation-based fecal DNA test demonstrated an overall sensitivity of 88% for CRC and 46.2% for APL respectively, and a specificity of 91.8% for controls. On the other hand, the mutation-based diagnostic model yielded limited sensitivity, and combined detection of methylation markers and mutation in fecal DNA did not improve the assay performance. Meanwhile, a diagnostic model based on the relative abundance of bacterial species showed inferior performance than the methylation-based model. Finally, integrated diagnostic model combining both methylation and microbial markers showed an enhanced performance (AUC= 0.95) compared to methylation markers alone. Conclusions: The multi-gene fecal DNA methylation test provided remarkable diagnostic performance for CRCs and APLs. Furthermore, multi-target assay integrating both methylation and microbial markers may further improve the diagnostic performance. Our findings may aid in the development of novel diagnostic tools for CRC.