An inverted index for mass spectra similarity query and comparison with a metric-space method

Abstract

Query performance is a determining factor in the adoption of an indexing method for similarity query. Metric space indexing methods take great pride in their general applicability. However, it is usually hard for a general method to perform well for every domain. Therefore, it is of interest to investigate the performance of metric-space methods, comparing with domain specific methods, on a particular domain. This paper describes such an investigation for proteomic mass spectra. An inverted index method that exploits the sparsity of mass spectra binary format data and acts as a coarse filter before fine ranking is proposed and empirically compared with an existing metric-space indexing method. Results show that the inverted index method yields greater search efficiency and outperforms the metric-space method in query speed and index size.