To evaluate the value of urinary normal and modified nucleosides in diagnosis and surgical monitoring of colorectal cancer (CRC). Between October 2002 and July 2003, 52 consecutive patients with pathological confirmed CRC were included in this study. Spontaneous urine samples were collected 1 d before and 8 d after surgery and 14 kinds of urinary nucleosides in the samples were determined by reversed-phase high-performance liquid chromatography (RP-HPLC) method. Another 62 healthy volunteers were also enrolled as controls. The routine clinical tumor markers, including serum CEA, CA199, CA125 and AFP levels of CRC patients were evaluated by electrochemical-luminescence immunoassay simultaneously. The mean levels of pseudouridine (Pseu), adenosine (A), cytidine (C), 1-methyladenosine (m1A), 1-methylinosine (m1I), 3-methyluridine + 5-methyluridine (mU), 2,2-methylguanosine (m22G), inosine (I), 1-methylguanosine (m1G), N4-acetylcytidine (ac4C), N6-methyladenosine (m6A) among 14 kinds of determined urinary nucleosides in CRC group were much higher than those of controls (P < 0.05). Based on principal component analysis, 76.9% of CRC patients were correctly identified, which was much higher than that of CEA (38.5%), CA199 (40.4%), CA125 (15.4%), and AFP (17.3%) (P < 0.01). ROC curve analysis of m1G, and Pseu showed good sensitivity-specificity profiles to CRC. Two classification equations, Y(normal) = -3.009 + 0.0272 x Pseu + 4.918 x m1G and Y(CRC) = -8.057 + 0.0667 x Pseu + 8.258 x m1G, were established by Bayes stepwise discriminate analysis for predicting carcinogenesis of CRC. The elevated levels of Pseu, C, U (uridine), m1A, m1I, m1G, ac4C, A, m22G dramatically decreased after curative resection of 40 cases of CRC. And our data also showed that the preoperative levels of Pseu, m1G, m1A and m22G were positively related with tumor size and the preoperative levels of m1A, m22G and ac4C were positively related with Duke's staging of CRC (P < 0.05). Normal and modified urinary nucleosides may become additional tumor markers which are feasible in the clinical setting and will prove helpful in the diagnosis, management and follow-up of CRC, and Pseu and m1G may be more promising for clinical application.