One of the key to progress in biomarker discovery is the correct identification of peptides and proteins. Substantial progress has been made in developing enhanced ion separation and transmission capabilities in mass spectrometry based instrumentation that has become a primary tool used in biomarker discovery. Such instrumentation is capable of producing massive amounts of data in a single run. Improved methods are needed to analyze and fully utilize this data to achieve the end goal of accurately identifying peptide/proteins and characterizing their biological function.Researchers at PNNL have developed several patented and patent pending methods for improving the confidence and reliability of peptide identification. These include:

• A method (US 7,136,759) for improving the confidence in peptide identification based upon comparing the actual elution time of a peptide from a chromatography separation to its predicted elution time. An artificial neural network approach is used to derive the predicted elution time of a peptide based upon a massive data base of peptide elution times derived from PNNL experiments. The correlation of the predicted elution times derived from the ANN model to the actual elution times is normally greater than 95%. By comparing the actual measured elution time of a peptide from an LC/MS experiment to a predicted elution time, one can increase the confidence level of peptide identification.
• A method (US 7,756,646) of predicting whether a peptide will be detected by analysis with a mass spectrometer. The method utilizes the results of repeated analyses of the same sample with a mass spectrometer to develop probabilities of specific peptides being detected by the mass spectrometer. The probability estimates can then be used to account for the relative peptide identification efficacy of the mass spectrometry platform as part of the process of determining the presence or absence of a particular peptide in a sample. This approach can reduce both false positive and false negative peptide identifications.
• A patent pending method for improving peptide identification accuracy based upon a probability score that a spectral peak from a peptide fragment produced by collision induced disassociation tandem mass spectrometry is a match to a model-generated spectral peak for frequently observed fragmented peptide ions.   
• A patent pending method to identify small molecules based upon on their fragmentation spectra as obtained during electrospray ionization-collision-induced dissociation (ESI-CID) or electrospray ionization-collision-activated dissociation (ESI-CAD) experiments. The method is based upon an artificial neural network approach to model and predict specific features of small molecules using empirically derived information. The candidate predicted metabolite spectra are then compared with experimental metabolite spectra to increase confidence and accuracy in metabolite identification.