Validating an Assay
Validation of all analytical methods is a regulatory requirement. Risk-analysis techniques such as FMEA (failure mode and effect analysis) have proven to be useful in this regard. Validation involves a risk assessment step that is followed by a close analysis of the robustness of each assay. As each protein requires a tailored quantification method this will to some extent determine the required validation process.
To illustrate the complexity of protein quantification, and the validation process, Protagen validated an assay for protein quantification according to ICH guideline Q2(R1), using a tumor lysate as a model matrix and bovine serum albumin (BSA) as a model protein. According to this ICH guideline, several parameters have to be validated for a quantitative assay, outlined in Table 1, with definitions in Table 2.
The aim was to develop a robust assay that is capable of quantifying the total protein content of a sample in a wide variety of buffers. The assay chosen was a modified protein-amidoblack-complex precipitation using triplicate determinations. The modifications included the introduction of a 96-well plate fitted with a membrane bottom, which allowed the reproducible isolation of the protein-dye precipitate.
The quantification was carried out relative to a BSA standard curve. To evaluate critical steps a FMEA was performed.
One of the critical steps identified was the risk of using commercially available BSA without checking the actual concentration of each lot. To overcome this, an NIST standard was employed to check the normal protein concentrations for the calibration curve. Another critical point was identified in the use of the multiwell plates for the assay—each lot of plates needed to be checked for a homogeneous read-out in all well positions.
Having addressed these and other outcomes of the FMEA, the robustness of the assay was considerably increased. Therefore, the validation characteristics of linearity, accuracy, precision (both repeatability and intermediate precision), detection limit, quantification limit, and range could be determined.