Adopting a practical statistical approach for evaluating assay agreement in drug discovery.

The authors assess the equivalence of 2 assays and put forward a general approach for assay agreement analysis that can be applied during drug discovery. Data sets generated by different assays are routinely compared to each other during the process of drug discovery. For a given target, the assays used for high-throughput screening and structure-activity relationship studies will most likely differ in their assay reagents, assay conditions, and/or detection technology, which makes the interpretation of data between assays difficult, particularly as most assays are used to measure quantitative changes in compound potency against the target. To better quantify the relationship of data sets from different assays for the same target, the authors evaluated the agreement between results generated by 2 different assays that measure the activity of compounds against the same protein, ALK5. The authors show that the agreement between data sets can be quantified using correlation and Bland-Altman plots, and the precision of the assays can be used to define the expectations of agreement between 2 assays. They propose a scheme for addressing issues of assay data equivalence, which can be applied to address questions of how data sets compare during the lead identification and lead optimization processes in which assays are frequently added and changed.