QSAR Analysis Involving Assay Results Which are only Known to be Greater Than, or Less Than Some Cut-off Limit

It is often impossible to measure compound properties above or below some detection limit, or cut-off level, in which cases result values can only be recorded as lying above or below that censoring limit. Such result values carry information, but are often excluded from QSAR studies or forcing the use of classification methods where regression models are more appropriate. This paper describes a regression method which makes use of censored values. The method works by estimating (imputing) the values underlying censored results. The algorithm was tested for stability and accuracy in the context of Free-Wilson R-group activity score calculations.