An application of a relational database system for high-throughput prediction of elemental compositions from accurate mass values

Bioinformatics. 2013 Jan 15;29(2):290-1. doi: 10.1093/bioinformatics/bts660. Epub 2012 Nov 18.

Abstract

Summary: High-accuracy mass values detected by high-resolution mass spectrometry analysis enable prediction of elemental compositions, and thus are used for metabolite annotations in metabolomic studies. Here, we report an application of a relational database to significantly improve the rate of elemental composition predictions. By searching a database of pre-calculated elemental compositions with fixed kinds and numbers of atoms, the approach eliminates redundant evaluations of the same formula that occur in repeated calculations with other tools. When our approach is compared with HR2, which is one of the fastest tools available, our database search times were at least 109 times shorter than those of HR2. When a solid-state drive (SSD) was applied, the search time was 488 times shorter at 5 ppm mass tolerance and 1833 times at 0.1 ppm. Even if the search by HR2 was performed with 8 threads in a high-spec Windows 7 PC, the database search times were at least 26 and 115 times shorter without and with the SSD. These improvements were enhanced in a low spec Windows XP PC. We constructed a web service 'MFSearcher' to query the database in a RESTful manner.

Availability and implementation: Available for free at http://webs2.kazusa.or.jp/mfsearcher. The web service is implemented in Java, MySQL, Apache and Tomcat, with all major browsers supported.

Contact: sakurai@kazusa.or.jp

Supplementary information: Supplementary data are available at Bioinformatics online.

Publication types

  • Research Support, Non-U.S. Gov't

MeSH terms

  • Algorithms
  • Databases, Chemical*
  • Mass Spectrometry / methods*
  • Metabolomics / methods*