Peripheral blood mRNA signature for colorectal cancer screening.
B. Mougin,S. Cai,Y. Xu,X. Ye,X. Meng,Q. Xu,F. Liu,S. Ni,F. Wu,X. Du
DOI: https://doi.org/10.1200/JCO.2011.29.15_SUPPL.E14030
IF: 45.3
2011-05-20
Journal of Clinical Oncology
Abstract:e14030 Background: Colorectal cancer is the second and third leading cause of cancer-related mortality for women and for men in the world. It is believed that more than 50% of these deaths may have been prevented through the use of screening tests. The compliance with current CRC screening recommendations remains poor. A simple blood-based test would represent an interesting alternative. We used DNA-microarrays and PAXgene to study the gene expression repertoires in peripheral blood, suitable for new biomarker discovery.
METHODS
119 CRC patients and 101 non-CRC controls were enrolled in our study. RNA extracted from peripheral blood samples collected in PAXgene tubes, were analysed after NuGEN WT-Ovation RNA amplification, using Affymetrix GeneChip HG-U133 Plus 2.0 Array. Molecular signature discovery was performed based on 5-fold cross validation method, which means 4/5 of samples were used as internal train fold, prediction model was created with the top 100 genes scored by RFE-SVM, and then model performance was assessed with rest of samples. The process was run for 1000 iterations. Eventually, accuracy, sensitivity and specificity were used to measure the overall performance. Furthermore, we estimated the signature performance in an independent validation set, with 45 CRC patients and 47 non-CRC controls.
RESULTS
The overall performance based on cross validation process reached 95% (94.9, 95.3) accuracy, 95% (94.8, 95.2) sensitivity and 95% (95.0, 95.4) specificity. The signature performance in the independent test set reached 87.7% (77.9, 92.7) accuracy, 73.3% (57.8, 84.9) sensitivity and 100% (90.6, 100.0) specificity. The majorities of signature genes represent MHC class II receptoe activity, scavenger receptor activity, T cell proliferation and signal transduction information, implying the molecular changes occurring during disease progression.
CONCLUSIONS
Our results demonstrated the feasibility of using gene expression profiling in peripheral blood to identify novel biomarkers. The CRC biomarker panel learned on the present dataset was considered as a promising transcriptomic biomarker for colorectal cancer screening.
Biology,Medicine