QuBAS interpolation analysis uses a slightly different mathematical approach from the standard process. Called the Bursa-Yellowlees method, a
blog describing it is available.
In the standard process, the concentration of the unknown data points is back transformed from the standard curve and adjusted for dilution of the sample. If there is more than one dilution or replicate the results are averaged to get the final concentration result.
Underlying this method is an assumption that all the unknown data points lie on a (dilution) curve that is parallel to the standard curve. If this were not the case, the averaging method would not be correct, and the sample would not be biologically equivalent to the standard.
Quantics improved method
Knowing that the dilution points must lie on a parallel curve allows all points to be used, even if they lie outside the range of the standard. It can be shown that the unknown value is represented by the shift in the C parameter between the standard and unknown curves on appropriate axis.
For details please contact Quantics.
The uncertainty in results from dilution point that are near the standard asymptotes, or even outside them is reflected in the confidence intervals for the goodness of fit, which in turn contribute to the CI of the final result.
Not only does this process allow all points to be used, and therefore, on average, fewer dilutions are required for the same precision of answer, but it also greatly simplifies the sample suitability criteria.
For a single dilution the results are identical to the standard method.
QuBAS assumes that all replicate samples are pseudo replicates, i.e. they are not truly independent. This is the most common situation in practise. Management of replicates as independent or not does not impact the results, but treating non-independent replicates as independent creates confidence intervals that are falsely narrow.
