Skip to main content

Advertisement

Springer Nature is making SARS-CoV-2 and COVID-19 research free. View research | View latest news | Sign up for updates

Figure 2 | BMC Bioinformatics

Figure 2

From: caCORRECT2: Improving the accuracy and reliability of microarray data in the presence of artifacts

Figure 2

Effect of scratch artifact and removal on residual images. All 4 panels are heat maps with high positive values colored in red, values near zero in white, and large negative values colored in blue. The panels were not all generated identically, (i.e. panel C by RMAExpress, and others by caCORRECT), and so the scales and ranges may differ without loss of central meaning. All images show residuals between observed intensity and that expected by the underlying model. Thus, red color indicates higher than expected intensity, blue color indicates lower than expected intensity and white indicates good fit to the model. Panel A shows (caCORRECT) residuals from the original chip, panel B shows the (caCORRECT) residual after artifact flagging by caCORRECT, panel C shows the (RMA) residual produced by RMAExpress from the original chip, and panel D shows the (caCORRECT) residual produced after median-replacement form Harshlighting. Systematic blue color surrounding the red scratch (prominent in A and weaker in C) reveals poor model fit to the data, which is likely to cause an overestimate of gene expression for these probes. The figure suggests that caCORRECT has identified this scratch better than RMA has, and modeled the data better than Harshlighting has, resulting in a more accurate gene expression than either. The effective reduced image resolution for the RMA panel is due to the way that RMA handles pairs of perfect match (PM) and mismatch (MM) probes together when calculating residuals, whereas caCORRECT processes PM and MM probes independently of one another. Portions of this figure are reproduced with permission from [23].

Back to article page