Probability profile method - new approach to data analysis in tandem mass spectrometry
peer-reviewed · Proceedings. 2004 IEEE Computational Systems Bioinformatics Conference, 2004. CSB 2004 · 2004
| Date | 2004-11-08 |
| Type | peer-reviewed |
| Venue | Proceedings. 2004 IEEE Computational Systems Bioinformatics Conference, 2004. CSB 2004 |
| Publisher | IEEE |
| Contribution | algorithm |
| DOI | 10.1109/csb.2004.1332475 |
| Citations (OpenAlex) | 0 |
Abstract
Tandem mass spectrometry (MS/MS) is one of the leading proteomics technologies, applicable to a wide range of experiments involving composition analysis of protein mixtures. Currently only /spl sim/10-20% of MS/MS spectral data lead to the successful peptide identifications, and the rate of false positives remains to be high. We propose probability profile method (PPM) as a new route for the development of MS/MS data analysis algorithms. The principal idea can be described as a probabilistic “labeling” of the individual peaks, or as a detailed analysis of the spectra leading to peak separation into specific categories (b-ion, y-ion, double charged b-ion, etc). PPM “assignments”, conducted on large and diverse data sets (/spl sim/60,000 spectra), indicate that a large majority of MS/MS peaks can be identified with a surprising level of confidence, providing the foundation for a range of novel algorithmic approaches: spectra can be edited by selecting desirable peak categories; overall characteristics of MS/MS spectra, such as parent ion charge or total number of the present ions, can be rapidly estimated with a high precision; labeled peaks of the same category (e.g. b-ions) can be efficiently connected into de novo tag peptides.
Methods and tools
- Probability profile method (ORNL): Learns neighbourhood patterns of b, y and related peaks from resolved spectra to assign each peak a category probability, giving anchor points for de novo sequence assembly.