Views
Section15:Introduction
From Assay Guidance Wiki
Contents |
Definitions of Result Levels
Raw data (from counters, detectors, etc.).
Normalized well level or individual data values (Inhibition, Stimulation, etc; in most cases Inhibition and Stimulation are expressed as a % of the dynamic range of the assay).
Aggregate median (preferred) or mean normalized well level data when replicates exist in a single run (Inhibition, Stimulation, etc…). This level provides for a consistent determination of n as it applies to in vitro results.
Derived data (ICx, Relative ICx, ECx, Ki, Kb, etc.).
Summarized data (geometric mean ICx, Relative ICx, Ki, Kb or average Inhibition, Stimulation, etc.).
Treatment of Replicate Data
When a compound defined by given serial number and lot number is tested more than once in an experimental run with the same reagents and conditions, only one result should be published, n=1. When this occurs, the result level data can be aggregated prior to publication. Proper data aggregation for results from a run requires the appropriate selection and application of either a geometric or arithmetic mean calculation. Geometric mean calculation must be used when aggregating results that are generated logarithmically (e.g. IC50, EC50). Arithmetic mean calculation must be used when aggregating results that are generated linearly (e.g. Inhibition, Stimulation).
For a compound with the same serial number and lot number to have results n=2 or higher, the compound material must be tested with reagents prepared separately and in a different ‘run’. Multiple preparations (i.e. lots) of a given chemical structure are considered to be “different” and, therefore, can have independent results published from the same experimental run.
Abs IC50, Rel IC50 or Rel EC50
For assays described in this chapter, Absolute IC50, Relative IC50 and Relative EC50 are predominantly used to derive a value that can be used to compare results within and across runs as well as between assays. Abs IC50 and Rel IC50 are used when different assumptions are applied; the selection of either is at the discretion of the scientist but should be applied consistently and not changed for a defined assay.
For consistency, Rel IC50 is used for inhibition assays while Rel EC50 is used for stimulation assays, even though there is no fundamental difference between them. Because of their relative simplistic composition, biochemical in vitro assays can be easily labeled as either Stim or Inh, while every biochemical whole cell assay can be either as Stim or Inh depending on multiple factors. Therefore, the guideline for defining whole cell biochemical assays is to use the label that better reflects the perceived pharmacology, regardless of the direction (increasing or decreasing with test substance concentration) of the raw signal. How an assay is defined within Reference Database can also drive which result type label to use. For instance, if an assay categorized by Cell Cycle Modulation is attempting to inhibit the cell cycle, the Rel IC50 should be used.
Guidelines for Curve Fitting
- Three or four parameter logistic curve fits are acceptable.
- Under appropriate conditions, the Top may be fixed to 100 and the Bottom may be fixed to 0, but a two parameter fit is not acceptable.
- The Hill Coefficient may not be preset to any fixed number, unless supported by a statistician.
- Cubic spline curve fits are not acceptable, unless supported by a statistician.
- The Fitting Error of the IC50/EC50 should be less than 40% of the IC50/EC50 and not exceed 100%, unless supported by a statistician. (It should be noted that this “standard error” is a measure of “goodness of fit” of the data to the curve fitting equation and not the “standard error” of aggregate data values).
Normalizing Data using a Positive Control Curve
In some cases, it is preferable to use a reference curve to define the dynamic range of the assay. In those cases, the fitted Top of the reference curve is substituted for the Max while the fitted Bottom of the reference curve is substituted for the Min in normalization calculations. This may be particularly useful in Agonist assays where the reference agonist curve is strongly recommended. It is still preferable to define the dynamic range on each plate so that individual plate drift is assessed and single plates can pass/fail. Additionally, the upper and lower asymptote of the reference curve must be established by the data in order to use them for dynamic range determination.
Application of a Standard Curve
Use of a standard curve is required wherever possible when the raw data is not a linear function of the biological response. For example, optical densities, fluorescence units and luminescence units often cannot be directly used for calculations of activity as they are often non-linear functions of the concentration of the relevant biological product. This standard curve is used to convert the raw data to concentration of biological substance. The calculated concentrations are then used to calculate the Normalized Result, as discussed in each assay section. The standard curve data should be generated with an appropriate number and spacing of points, fit by an appropriate dose response model so that bias and precision are within acceptable limits, and all raw data within the scope of the assay can be converted to the biological response.
Documenting Assay Changes in the Protocol
The written assay protocol should provide information not only to enable another person to repeat the assay but also descriptions of changes to the assay as they have occurred over time. Since some assays undergo changes (e.g. different instrument) but continue to report data to the same method, the notation of these changes comprehensively describe the body of data that is queried. A statistician should be consulted when assay changes are made to determine if additional validation steps are necessary. See Section II.F. for the validation steps necessary for various types of assay changes.
