Microarray Analysis and Co-expression Network Analysis- Colorectal and Breast Cancer

• Main Authors: CHANG, YU-TIEN, 張語恬
• Other Authors: CHU, CHI-MING
• Format: Others
• Language: en_US
• Published: 2017
• Online Access: http://ndltd.ncl.edu.tw/handle/fzf544

博士 === 國防醫學院 === 醫學科學研究所 === 106 === Backgrounds Gene expression profiling analysis and co-expression analysis facilitate further understanding of molecular function and identification of novel biomarkers of cancers. However, these kinds of studies were often limited by the small number of available samples. In order to overcome the limitation, we used pooled cDNA microarray data sets to 1) evaluate serum biomarkers for colorectal cancer (CRC) screening; 2) discover novel biomarkers and create prediction models on breast cancer (BRC) recurrence via co-expression network analysis. Methods and Materials 1)15 candidate genes pertaining to CRC prognosis were selected by literature view. We analyzed how they expressed in the blood samples of 227 CRC patients and 111 heath controls, which were matched by age and sex. 12 integrated independent microarray data sets were used to validate these biomarkers. 2)The gene profiles for BRCs recurrence was obtained by differential co-expression analysis in 4 pooled microarray data sets (recurrence: n=354, non-recurrence: n=553). Based on networks of different size, we created various prediction models and risk scores on recurrence by Cox regression. Results 1)Five genes, i.e., MDM2, DUSP6, CPEB4, MMD, and EIF2S3, were significantly associated with CRCs. The DUSP6 model displayed the best discrimination ability between CRCs and health cases, with an AUC of 0.804 (95%CI: 0.730-0.879) compared with the other one-gene models (AUC: 0.49-0.69). Distinct increases in the AUC of up to 0.905 (95%CI: 0.849-0.960) resulted from the combination of the five genes. It was validated by using 12 pooled microarray and the results showed R2 of 0.853, AUC of 0.978, accuracy of 0.949, specificity of 0.818 and sensitivity of 0.971. We created a new prediction model by selecting genes from all models using forward Cox regression. This model performed the best prediction with 7 genes of CPEB4, EIF2S3, ANXA3, TNFAIP6, IL2RB, MGC20553 and MS4A1. 2)The co-expression analysis results showed that hub genes related to immune play important roles in BRC recurrence. They are IGHA1, IGHD, IGHG1, IGHG3, IGLC2, and IGLJ3, which are all significantly associated with recurrence and node status (HR: 2.3-23, p<0.001). The models made of bigger size of networks have better prediction performance than that made of differential expressed genes (DEGs) on recurrence. The hazard ratio of risk score in biggest network-based model is 3.32 (p<10-8) under the control of node status. Conclusion We find five serum gene signature for CRCs screening with stable and great classification ability. This signature may serve as a promising screening tool after further analysis. We also found that the network-based models outperform DEGs based model. Through co-expression analysis, we succeeded in discovering 8 immune related genes and 4 novel biomarkers related to BRCs recurrence, IDUA (Iduronidase, Alpha-L-), CPZ (Carboxypeptidase Z), MGC27165 and C10orf56. Their functions in BRCs are still unknown. Further functional analysis is needed and they may serve as promising new target for BRCs treatments in the future.