Migration

Data structures

One of the main differences between PLA 2.x and PLA 3.0 is the redesign of the information structure.

In PLA 2.x, data was organized in a hierarchy of projects, assays, and assay elements, namely standard, preparation and control (line) objects. This structure is highly flexible and has several advantages concerning practical handling. However, it also has severe disadvantages, which led to the development of the Quantitative response assay document type.

In the case of Quantitative response assay documents, all assay data including its results are kept in one single document. This document aims to calculate the potencies for one single assay run. The main difference comparing the document to PLA 2.x is that PLA 3.0 only supports one single interpretation of the standard curve within one single assay. If the definitions of a PLA 2.x assay allow different interpretations of the standard or the model (for example when automatic linear range detection is turned on with the 'Individual Range' option, two different ranges can be selected within PLA 2.x), this is no longer supported within PLA 3.0. In this case, the assay definitions need to be split per assay element.

Statistical differences between PLA 2.x and PLA 3.0

PLA 3.0 comes with many additional capabilities and allows a much more sophisticated assay setup.

From a statistical perspective, the main difference comes from the conceptual change of the assay perspective. In PLA 3.0, all assay elements have to be calculated with the same analytical setup, while in PLA 2.x the analytical setup has been controlled by the sample element. If the assay is set up with varying analytical settings, the assay has to be split in PLA 3.0.

The same conceptual change applies to the removal of the range optimizer option. In PLA 2.x, you could use this option to select different steps for the standard in parallel-linear models. If you need to select different range areas for the standard in PLA 3.0, you also have to split the assays per sample.

In the case of non-linear regression, the precision of PLA 3.0 has been improved. In a well-behaved assay, this leads to small differences in the estimated parameters and their standard errors. Usually, these deviations are very small. However, if the iteration number for non-linear regression is high (for example >50 iterations), this indicates an assay with a very shallow hyper-surface. In this case, the deviation can get larger. Especially the 5-parameter-fit might show this behavior. The reason for these differences is the application of different cut-off criteria during optimization.

Splitting assays

The migration task automatically splits assays if specific conditions are met. In this case, the migration process creates one assay document with all samples and an assay document for each sample. The Migration and Archiving Toolkit performs these split operations automatically.