Quality Control

Quality Control

This section contains quality control plots and statistics for the methylation data.

Quality Control Box Plots

Each box plot below shows the signal distribution of quality control probes across all samples. The control box plots are separated by control types. Detailed description of the control probes is given in the RnBeads vignette.

Control probe type

Figure 1

Open PDF Figure 1

Quality control box plots.

Quality Control Bar Plots

The plots below visualize the exact signal levels at each quality control probe. Note that the scale is not standardized. Background signal is usualy at the level of 1000 to 2000.

Samples #:
Control probe ID

Figure 2

Open PDF Figure 2

Quality control bar plots.

Negative Control Box Plots

Negative control box plots visualize background intensity distributions of all analyzed samples. Samples with skewed distributions and high medians are likely to be of low quality and should be discarded from the analysis.

Samples #:

Figure 3

Open PDF Figure 3

Box plots of the negative control probes.

Visualization of Sample Mixups

SNP-based box and bar plots that facilitate identification of the sample mixups.

SNP Heatmap

SNP Heatmap allows for identification of sample mixups in particular for genetically matched designs.

Figure 4

Open PDF Figure 4

Heat map of the SNP probes. The samples with the same genetic background should cluster together.

SNP-based Distances

If we inspect the samples in the space defined by the SNP probes only, samples appear close to each other are genetically similar. We calculated the Manhattan distance between all vectors of SNP probes. The figure below shows the relative distances.

Display samples as

Figure 5

Open PDF Figure 5

Scatter plot showing the samples in the first two principal components of the space of their SNP probes.

The full table of all pairwise distances is stored in a dedicated comma-separated file accompanying this report.

Gender Prediction

RnBeads predicted the gender of the samples in the dataset using a logistic regression model. The results are summarized in the table below.

Gender Samples
1 female 2574
2 male 1460
3 unknown 0

Gender was predicted based on the increase (or decrease) of mean signal intensities in the sex chromosomes w.r.t. the corresponding value in autosomes. The figure below displays these characteristics of the samples.

Colors denote

Figure 6

Open PDF Figure 6

Gender prediction based on mean signal increase. The decision boundary between the two genders is visualized by a black line. Sample colors denote predicted male probability / gender.