One of the assays in the Neuro Exploratory panel, TDGF1 (Teratocarcinoma-derived growth factor 1, Uniprot number: P13385), shows a strong binominal pattern in human plasma and serum. Due to this binominal pattern, the intensity normalization applied to all projects with randomized samples is not recommended for this marker. Therefore, the results for TDGF1 will always be delivered normalized against the Inter-Plate Controls (IPCs), even if the remaining data in the project is intensity normalized.
Data normalization of normally distributed assays
Intensity normalization of the data from an assay with normally distributed data will help reduce potential technical variation. The figures above show an example from a project spanning over 3 sample plates. Each color indicates a different plate and the plate median is visualized by a dotted line. Using the IPC’s for normalization produces good data, but for a study with randomized samples where the plate median for each assay can be assumed to be the same over all plates, the added intensity normalization will help cancel out technical variations to a higher degree.
Data normalization of TDGF1
Unlike when the data of a normally distributed assay is intensity normalized, intensity normalization of TDGF1 data can increase the variation between plates rather than eliminate it. As the distribution of samples between the two groups (the high vs. the low group) is almost 50-50, one or a few samples more in one of the groups on a plate will skew the plate median upon which we base the intensity normalization. I.e. a few samples more in the group with high values will result in a higher plate median compared to a plate with a few more samples in the group with low values.
Using intensity normalization for this assay will therefore introduce more variation rather than decrease technical variation, and potentially biologically interesting findings cannot be seen as clearly.
Data normalization FAQ
Sample randomization FAQ
How is the data pre-processed? FAQ
For more information see our white paper, “Data normalization and standardization“.