TY - JOUR AU - Thallinger, Gerhard G. AU - Obermayr, Eva AU - Charoentong, Pornpimol AU - Tong, Dan AU - Trajanoski, Zlatko AU - Zeillinger, Robert PY - 2012 TI - A Sequence Based Validation of Gene Expression Microarray Data JF - Current Research in Bioinformatics VL - 1 IS - 1 DO - 10.3844/ajbsp.2012.1.9 UR - https://thescipub.com/abstract/ajbsp.2012.1.9 AB - Problem statement: Quantitative Reverse Transcription PCR (RT-qPCR) is often used to validate microarray data. Previous studies show different levels of correlation, without further investigation of influencing factors. Approach: We compared expression levels of 381 genes obtained from microarray hybridizations and from TaqMan based RT-qPCR assays. Correlation of expression levels was determined by comparing: (i) single genes across samples, (ii) all genes within a sample and (iii) the expression ratios of all genes in a sample using another sample as the reference. The influence of several parameters on the correlation was analyzed: (i) variation in transcript set targeted by the microarray probe and the PCR assay, (ii) variation in amplicon probe position relative to 3' end of transcript, (iii) variation in efficiency of the PCR reaction and (iv) normalization of the PCR data. Results: The 381 genes covered by RT-qPCR had 494 matching probes on the microarray. 397 probes with a matching transcript set were identified via a rigid sequence-based validation. Correlation was significantly higher among matching transcript sets and probes closer to the 3' end. Adjustments for different amplification efficiencies had either no influence or decreased correlation. Normalization of qPCR data consistently reduced correlation for all analysis approaches. Conclusion: Current clinical research uses microarrays to select genes of interest and evaluates these genes using qPCR. Therefore, it is important that expression levels measured by both techniques be highly correlated. High correlation can be achieved if the targeted transcript sets match, whereas normalization and efficiency correction can have a negative influence.