Limits...
Simple regression for correcting ΔCt bias in RT-qPCR low-density array data normalization.

Cui X, Yu S, Tamhane A, Causey ZL, Steg A, Danila MI, Reynolds RJ, Wang J, Wanzeck KC, Tang Q, Ledbetter SS, Redden DT, Johnson MR, Bridges SL - BMC Genomics (2015)

Bottom Line: ΔCt values are then used to derive ΔΔCt when compared to a control group or to conduct further statistical analysis.We propose to regress the Ct values of the target genes onto those of the reference genes to obtain regression coefficients, which are then used to adjust the reference gene Ct values before calculating ΔCt.This method can be applied to both low density RT-qPCR arrays and individual RT-qPCR assays.

View Article: PubMed Central - PubMed

Affiliation: Department of Biostatistics, School of Public Health, University of Alabama at Birmingham, Birmingham, AL, 35294, USA. xcui@uab.edu.

ABSTRACT

Background: Reverse transcription quantitative PCR (RT-qPCR) is considered the gold standard for quantifying relative gene expression. Normalization of RT-qPCR data is commonly achieved by subtracting the Ct values of the internal reference genes from the Ct values of the target genes to obtain ΔCt. ΔCt values are then used to derive ΔΔCt when compared to a control group or to conduct further statistical analysis.

Results: We examined two rheumatoid arthritis RT-qPCR low density array datasets and found that this normalization method introduces substantial bias due to differences in PCR amplification efficiency among genes. This bias results in undesirable correlations between target genes and reference genes, which affect the estimation of fold changes and the tests for differentially expressed genes. Similar biases were also found in multiple public mRNA and miRNA RT-qPCR array datasets we analysed. We propose to regress the Ct values of the target genes onto those of the reference genes to obtain regression coefficients, which are then used to adjust the reference gene Ct values before calculating ΔCt.

Conclusions: The per-gene regression method effectively removes the ΔCt bias. This method can be applied to both low density RT-qPCR arrays and individual RT-qPCR assays.

Show MeSH

Related in: MedlinePlus

Negative correlation between the mean of the ΔCtvalues from the target genes and Ctvalues of the reference genes after normalization via conventional subtraction. The lower right panel is based the Ct mean of all five reference genes while the others are based on individual reference gene. Ref, reference; dCt, ΔCt; r, Pearson correlation coefficient; p, p value from testing the correlation coefficient against 0.
© Copyright Policy - open-access
Related In: Results  -  Collection

License 1 - License 2
getmorefigures.php?uid=PMC4335788&req=5

Fig2: Negative correlation between the mean of the ΔCtvalues from the target genes and Ctvalues of the reference genes after normalization via conventional subtraction. The lower right panel is based the Ct mean of all five reference genes while the others are based on individual reference gene. Ref, reference; dCt, ΔCt; r, Pearson correlation coefficient; p, p value from testing the correlation coefficient against 0.

Mentions: The commonly used normalization method for RT-qPCR data is subtracting the Ct values of the internal reference genes from those of the target genes to obtain the difference in the Ct (ΔCt). The premise is that differences in the loading amount of template would be represented by the different Ct values of the reference genes. Therefore, subtracting the Ct of the reference genes (or taking the ratio on the exponential scale) would adjust for these RNA loading differences. To assess the validity of this premise, we plotted the mean Ct values of the target genes from a low-density PCR-based array (SAB array), which represent the average signal strength of the target genes, against the reference gene Ct values. If the premise were correct, there would be a positive correlation. As expected, the mean Ct values of the target genes were indeed positively correlated (r between 0.68 and 0.86) with the Ct values of the reference genes (Figure 1). However, after subtracting the reference gene Ct values, a negative correlation (r between −0.84 to −0.44) was generated between the mean of the ΔCt values of the target genes and the Ct values of each reference gene (Figure 2). This finding indicates a systematic over-correction (bias). If there were no bias, there would be no significant correlation between the mean ΔCt values of the target genes and the reference gene Ct values. All five reference genes showed similar negative correlation although the degree varied, which indicates that this is a general phenomenon instead of the property of a particular reference gene. The negative correlation remained present (r = −0.83) when the geometric mean of multiple reference genes (instead of individual reference genes) was used (Figure 2).Figure 1


Simple regression for correcting ΔCt bias in RT-qPCR low-density array data normalization.

Cui X, Yu S, Tamhane A, Causey ZL, Steg A, Danila MI, Reynolds RJ, Wang J, Wanzeck KC, Tang Q, Ledbetter SS, Redden DT, Johnson MR, Bridges SL - BMC Genomics (2015)

Negative correlation between the mean of the ΔCtvalues from the target genes and Ctvalues of the reference genes after normalization via conventional subtraction. The lower right panel is based the Ct mean of all five reference genes while the others are based on individual reference gene. Ref, reference; dCt, ΔCt; r, Pearson correlation coefficient; p, p value from testing the correlation coefficient against 0.
© Copyright Policy - open-access
Related In: Results  -  Collection

License 1 - License 2
Show All Figures
getmorefigures.php?uid=PMC4335788&req=5

Fig2: Negative correlation between the mean of the ΔCtvalues from the target genes and Ctvalues of the reference genes after normalization via conventional subtraction. The lower right panel is based the Ct mean of all five reference genes while the others are based on individual reference gene. Ref, reference; dCt, ΔCt; r, Pearson correlation coefficient; p, p value from testing the correlation coefficient against 0.
Mentions: The commonly used normalization method for RT-qPCR data is subtracting the Ct values of the internal reference genes from those of the target genes to obtain the difference in the Ct (ΔCt). The premise is that differences in the loading amount of template would be represented by the different Ct values of the reference genes. Therefore, subtracting the Ct of the reference genes (or taking the ratio on the exponential scale) would adjust for these RNA loading differences. To assess the validity of this premise, we plotted the mean Ct values of the target genes from a low-density PCR-based array (SAB array), which represent the average signal strength of the target genes, against the reference gene Ct values. If the premise were correct, there would be a positive correlation. As expected, the mean Ct values of the target genes were indeed positively correlated (r between 0.68 and 0.86) with the Ct values of the reference genes (Figure 1). However, after subtracting the reference gene Ct values, a negative correlation (r between −0.84 to −0.44) was generated between the mean of the ΔCt values of the target genes and the Ct values of each reference gene (Figure 2). This finding indicates a systematic over-correction (bias). If there were no bias, there would be no significant correlation between the mean ΔCt values of the target genes and the reference gene Ct values. All five reference genes showed similar negative correlation although the degree varied, which indicates that this is a general phenomenon instead of the property of a particular reference gene. The negative correlation remained present (r = −0.83) when the geometric mean of multiple reference genes (instead of individual reference genes) was used (Figure 2).Figure 1

Bottom Line: ΔCt values are then used to derive ΔΔCt when compared to a control group or to conduct further statistical analysis.We propose to regress the Ct values of the target genes onto those of the reference genes to obtain regression coefficients, which are then used to adjust the reference gene Ct values before calculating ΔCt.This method can be applied to both low density RT-qPCR arrays and individual RT-qPCR assays.

View Article: PubMed Central - PubMed

Affiliation: Department of Biostatistics, School of Public Health, University of Alabama at Birmingham, Birmingham, AL, 35294, USA. xcui@uab.edu.

ABSTRACT

Background: Reverse transcription quantitative PCR (RT-qPCR) is considered the gold standard for quantifying relative gene expression. Normalization of RT-qPCR data is commonly achieved by subtracting the Ct values of the internal reference genes from the Ct values of the target genes to obtain ΔCt. ΔCt values are then used to derive ΔΔCt when compared to a control group or to conduct further statistical analysis.

Results: We examined two rheumatoid arthritis RT-qPCR low density array datasets and found that this normalization method introduces substantial bias due to differences in PCR amplification efficiency among genes. This bias results in undesirable correlations between target genes and reference genes, which affect the estimation of fold changes and the tests for differentially expressed genes. Similar biases were also found in multiple public mRNA and miRNA RT-qPCR array datasets we analysed. We propose to regress the Ct values of the target genes onto those of the reference genes to obtain regression coefficients, which are then used to adjust the reference gene Ct values before calculating ΔCt.

Conclusions: The per-gene regression method effectively removes the ΔCt bias. This method can be applied to both low density RT-qPCR arrays and individual RT-qPCR assays.

Show MeSH
Related in: MedlinePlus