One of the major challenges in the secondary screening of enzyme inhibitors (i.e., the determination of inhibition constants from dose-response curves) is the handling of occasional "outlier" data points. "Outliers" are data points that deviate from the theoretical pattern for unknown reasons, such as a random malfuncation of a laboratory robot or a human operator error.
Under usual circumstances, an outlier has to be spotted by the operator and manually deleted from the data set. Unfortunatetly, that method is laborious, error prone, and it introduces an undesirable degree of subjectivity. This paper describes a computational algorithm, based on Huber's Minimax method of robust regression, which allows a fully automatic treatment of outlier data points, including (optionally) a complete rejection of a clear outlier, using a heuristic rule.
The method has been successfully implemented in the software packages BatchKi and PlateKi.
- Full text (HTML)
This is the initially submitted version of the paper, before several minor editorial changes (wording, punctuation) were introduced by the page editor.
- Full text (PDF)
This is the final version of the paper in the Portable Document Format.