628P - Detection of 5-hydroxymethylcytosine in circulating-free DNA for early diagnosis of colorectal cancer

Abstract

BackgroundA convenient and accurate method to improve the effectiveness of colorectal-cancer screening is still under exploration. This may be achieved by a diagnostic model based on 5-hydroxymethylcytosine (5-hmC). MethodsWe collected peripheral venous blood from 285 colorectal cancer patients and 285 healthy people at Zhongshan Hospital, Shanghai. The 5-hmC level of each gene locus in circulating-free DNA (cfDNA) was detected using a highly sensitive sequencing method. Of 285 subjects, 178 in each group were randomly selected as training set. The first 100 gene loci with the most significant difference of 5-hmC level were selected for establishing a diagnostic model (Logistic regression model) based on machine-learning method. The rest subjects in each group were used as validating set to validate this model. Further, we analysed the influence of several clinicopathological factors on 5-hmC level. ResultsIn training set, the comprison of the 5-hmC level suggested that the expression of 5-hmC in cfDNA was significantly different between the two groups. Of the 100 gene loci, 35 were included in the model. In internal validation, the sensitivity and specificity of the model were 88.8% and 89.9%, respectively. While the sensitivity and specificity 81.3% and 89.7%, respectively, in external validation using the validating set. The area under ROC curve of external validation was 0.946. For colorectal cancer stages I-IV, the diagnostic sensitivities were 78.7%, 81.9%, 88.2% and 95.6%, respectively. We also found that age, gender, Ras status or the site of primary tumor had no significant effect on the expression of 5-hmC. However, there was a statistically difference of 5-hmC level between stage I-III and stage IV. ConclusionsOur study identified the gene loci with significantly different 5-hmC level between colorectal cancer patients and healthy people, then we established a diagnostic model of colorectal cancer with high sensitivity and specificity. However, the utility of the model still need further verification in large-scale screening experiments. Clinical trial identificationNCT03599947 09/16/2018. Editorial acknowledgementShanghai Yibien Gene Technology Co., Ltd. Provides Testing Platform. Legal entity responsible for the studyThe authors. FundingHas not received any funding. DisclosureAll authors have declared no conflicts of interest.