The identification of a Hit in a DI.Screening Analysis single dose Screening Analysis follows an elaborate mathematical procedure to ensure high confidence in the results. This approach considers the quality of the reference and sample measurements with respect to deviations between the individual replicates and relates the quality of these measurements to the signal area determined for the tested ligands. For this purpose, DI.Screening Analysis performs several operations:

Assessment of reference measurement quality + calculation Hit Threshold:

To distinguish binding ligands from non-binding ligands, it is crucial to account for the intrinsic variance of the individual assay by assessment of reference measurements and to define a meaningful minimum signal required for the identification of binding ligands.

For Spectral Shift Analysis, the 670nm/650nm ratio between reference points is determined and compared to the assay points. Any ligand different than 3 Standard deviations is determined as a hit.

For TRIC analysis, the areas between measured reference traces are calculated. That results in a number of ((n²-n)/2) areas, where n denotes the number of references.

Now, depending on the number of references within one group and the number of groups, the software will decide between three approaches to calculate the hit threshold.

1. Statistics within normalization groups

As long as there is more than one normalization group and sufficient reference traces within each group (4 or more), the software will first calculate the median reference area of all possible reference areas within each group as well as the robust standard deviation (MAD within all groups), followed by establishing the mean of the median values and the MAD values across all groups. The hit threshold is then defined as the following.

Hit Threshold = meanM +z∙meanMAD

where,

• meanM = Mean of all median reference areas across all normalization groups.

• z = User-defined Z-score factor (Typically between 3 and 7), which can be defined on the “Data” page of the software.

• meanMAD = mean of all Median absolute deviation (also robust standard deviation) across all groups.

This approach is the most robust one, with the best possible statistics computed first within each individual group and then across all groups.

2. Statistics across normalization groups

In some cases, groups are so small (3 or fewer references) that no meaningful statistics can be computed within each group. Consider, for example, a row-wise assay with two reference values per row. Here, if each row is defined as an individual group, there would be only one value for that group. Hence, statistics will be computed using the following approach: first, the mean of all possible reference areas within each group is computed, followed by the computation of the median and MAD across all the group means.

Hit Threshold = medianM +z∙MAD

where, 

• medianM = Median of all available means of the individual groups

• z = User-defined Z-score factor (Typically between 3 and 7), which can be defined on the “Data” page of the software.

• MAD = Median absolute deviation (also robust standard deviation) of all mean values

3. Corner cases

Finally, some corner cases are to be considered include cases where neither approach 1 nor 2 would result in meaningful statistics. 

For example, in case one binding check is evaluated, only one group is available, and usually only three data points per group. In this (binding check) case, even though the number of references is smaller than four, approach 1 is used, but no mean value is calculated; instead, median and robust standard deviation from within this one group are used.

There are other cases where no or too few values can be calculated because, for example, there are only one or two valid reference values. In such case (two valid references), the software will use an empirical hit threshold of 0.25.

Tip: By changing the Z-score factor, the sensitivity of hit identification can be fine-tuned.

Assessment of ligand reproducibility:

If multiple wells for one ligand were measured, the values of all ligand replicates are calculated. Based on the statistical variation of the ligand values, the reproducibility of the ligand is assessed.

Calculation of hit identification in Spectral Shift assays:

The ratios between the ligand and each reference value in the same referencing group are averaged to yield the mean signal value. If multiple wells for one ligand were measured, the mean signal of these wells is also calculated and compared to the Hit Threshold. If the Hit Threshold is exceeded, the ligand is classified as a hit.

Calculation of signal area and hit identification in TRIC assays:

For each ligand well, the areas between the ligand trace and each reference trace in the same referencing group are calculated and averaged to yield the mean signal value. If multiple wells for one ligand were measured, the mean signal areas of these wells are also averaged and compared to the Hit Threshold. If the Hit Threshold is exceeded, the ligand is classified as a hit.

Assessment of overall signal quality:

Finally, to judge the overall quality of a single-dose measurement (for each ligand to the reference), also referred to as signal-to-noise, the DI.Screening Analysis software calculates a “Signal Quality” value that is generally a very robust measure for the reliability of the data obtained for a given ligand.