COPS benchmark: interactive analysis of database search methods.

Abstract

The performance of sequence database search methods is usually judged by receiver operating characteristic (ROC) analysis. The proper interpretation of the results obtained and a fair comparison across different methods critically depends on the properties of the data set used for such an analysis; in particular, each query must have the same number of true positives and true negatives. Here, we present a novel web service based on a dataset specifically designed for ROC analysis and the investigation of alignment quality. The data set is derived from a quantitative classification of protein structures (COPS), while analysis and results are presented through an intuitive web interface. The analysis provides details such as false positives per query, and visualization of the structural similarity between query and targets. Most importantly, results obtained for a specific alignment method are immediately related to those obtained for several popular standard sequence alignment methods.