A Reference Sample Quality (RSQ) warning or error indicates that one or more probes were too variable in the sample population used for normalisation. An RSQ warning or error can be triggered if even a single probe has too much variation.

When dedicated reference samples are used, the RSQ is based on the standard deviation over each probe in the reference samples. When no dedicated reference samples are used, a corrected median absolute deviation (MAD) for each probe in the test and undefined samples is used instead.

Background

Coffalyser digitalMLPA calculates an inter ratio for relevant digitalMLPA probes and an estimated confidence interval based on the variation present during probe normalisation. For all relevant probes, the software also calculates the variation present in the reference sample population used for inter-normalisation. This variation is used to determine the RSQ in an experiment.

In digitalMLPA, there are two possible methods for inter-normalisation in an experiment:

1. Using dedicated reference samples for normalisation. This is recommended as it allows for reliable normalisation in all experiments.
2. Using all test samples in the experiment for normalisation. This option is only feasible for certain applications, e.g. large experiments where copy number changes are expected to be rare.

Analysis with dedicated reference samples (situation 1)

If reference samples are defined in an experiment, the standard deviation of the inter ratios per probe in this reference population is calculated. This calculated standard deviation is used as a measure of the variation observed in the reference population used for normalisation. If it exceeds a certain threshold for one or more relevant probes, this triggers an RSQ warning or error. 

Analysis without dedicated reference samples (situation 2)

If no reference samples are defined in an experiment, all test and undefined samples are used as the reference population during normalisation. The MAD of the inter ratios per probe in this reference population is calculated and used as a measure of the variation observed in the reference population used for normalisation. The MAD is multiplied by a correction factor to make the result comparable with the value of the standard deviation. If it exceeds a certain threshold for one or more relevant probes, this triggers an RSQ warning or error.

Factors that influence the RSQ

There are many factors that influence the RSQ in a digitalMLPA reaction. The most important factor is the quality of the experimental data, which is influenced by experimental execution and sample properties. The RSQ is also affected by probemix characteristics, including the inherent variability of probes, as well as whether reference samples have or have not been designated for normalisation. The number of samples that are used as the reference population during normalisation when no references are defined in the experiment also affects the RSQ, because the observed variation tends to be higher when a lower number of samples is used (see this article for details and an example).

Information about troubleshooting RSQ warnings and errors can be found in this article.