Bioinformatics Methods for Mass Spectra Analysis

我和我的同事们正在开发将肽质谱与蛋白质序列数据库中的条目相匹配的方法。我们的目标是推导出易于解释的统计显著性测试，研究按来源物种限制搜索数据库的效用，并提供可供其他研究人员进一步测试和改进的开放源代码。迄今为止的工作重点是开发光谱噪声的客观测量方法及其与使用现有商业匹配算法成功识别蛋白质的关系。我们发现，令人惊讶的是，频谱噪声是一个相当小的因素，匹配算法的局限性往往是导致识别失败的原因。我们还开发了OMSSA，一种开放质谱搜索算法。测试表明，OMSSA中使用的搜索性能与商业算法一样快，而且OMSSA的显著性检验统计提供了某种程度上改进的信噪比特性。为OMSSA提供交互式用户界面的进一步工作正在进行中。

My colleagues and I are developing methods for matching peptide mass spectra to entries in a protein sequence database. Our goals are to derive easily-interpreted tests of statistical significance, investigate the utility of restricting the search database by species of origin, and provide open source code that can be further tested and improved by other investigators. Work to date had focused on development of an objective measure of spectral noisiness and its relationship to success in protein identification using exisiting commercial matching algorithms. We have found, surprisingly, that spectral noise is a rather minor factor, with limitations in the matching algorithms being more often responsible for identification failure. We have also developed OMSSA, an Open Mass Spectrometry Serch Algorithm. Tests have shown that the search hueristics used in OMSSA are as fast as those in commercial algorithms, and furthermore that OMSSA's significance test statistics offer somewhat improved signal-to-noise properties. Further work to provide a an interactive user interface for OMSSA is in progress.