This subproject is one of many research subprojects utilizing the resources provided by a Center grant funded by NIH/NCRR. The subproject and investigator (PI) may have received primary funding from another NIH source, and thus could be represented in other CRISP entries. The institution listed is for the Center, which is not necessarily the institution for the investigator. Accurate predictions of peptide retention times (RT) in liquid chromatography have many applications in mass spectrometry-based proteomics. Most notably such predictions are used to weed out incorrect peptide-spectrum matches, and to design targeted proteomic experiments. In this study, we describe a RT predictor, ELUDE, which can be employed in both applications. ELUDE's predictions are based on 60 features derived from the peptides'amino acid composition and optimally combined using kernel regression. When sufficient data is available, ELUDE derives a retention time index for the condition at hand making it fully portable to new chromatographic conditions. In cases when little training data is available, as often is the case in targeted proteomics experiments, ELUDE selects and calibrates a model from a library of pre-trained predictors. Both model selection and calibration are carried out via robust statistical methods and ELUDE can handle situations where the calibration data contains erroneous data points. We benchmarked our method against two state-of-the-art predictors and showed that ELUDE outperforms these methods and tracked up to 43% more peptides in a theoretical SRM method creation experiment. ELUDE is freely available under Apache License from http://per-colator.com
Showing the most recent 10 out of 583 publications
