Background: Patients undergoing chemotherapy for colon cancer are at significant risk for developing moderate-to-severe side effects as a result of their treatment regimens. These side effects can be debilitating to the patient and are often associated with a number of negative health and economic consequences. There is currently no accurate method to identify which patients are at risk for chemotherapy-induced side effects. If such a prediction tool were available, it would provide opportunities for directed prophylactic interventions. This study adopted a novel approach to filling the unmet clinical need for an accurate risk prediction tool. It capitalized on the growing body of evidence that genetic factors (in particular, networks of interacting genes) play a role in determining the likelihood of a patient’s risk for developing side effects. Specifically, the study was designed to assess the feasibility of identifying SNP-BNs 1 that could accurately predict the risk for 6 common chemotherapy-induced side effects: chemotherapyinduced nausea and vomiting (CINV), diarrhea, oral mucositis (OM) 2 , cognitive dysfunction (CD), peripheral neuropathy (PN), and fatigue. Methods: Patients (n=57) with colon cancer who received at least 3 cycles of FOLFOX6 +/bevacizumab (along with standard supportive care strategies) were enrolled. After informed consent, saliva samples were collected, DNA was isolated, and SNPs were analyzed on Illumina Omni microarrays (2.5 x 10 6 SNPs). Side effects under consideration were observed using Patient Care Monitor © , a validated patient-reported symptom assessment instrument. BNs were developed for each of the 6 side effects and cross-validated using robust statistical analyses. Results: The percentage of patients who experienced moderate-to-severe side effects was notable despite supportive care measures. Side effects included the following: CINV (32%), diarrhea (16%), OM (26%), CD (21%), PN (26%), and fatigue (56%). SNP-BNs were defined for each of the 6 side effects and were found to predict risk with a high degree of accuracy (>90%) and receiver operating characteristic (ROC) curves (>0.920).