Establishment of a protein frequency library and its application in the reliable identification of specific protein interaction partners.

The reliable identification of protein interaction partners and how such interactions change in response to physiological or pathological perturbations is a key goal in most areas of cell biology. Stable isotope labeling with amino acids in cell culture (SILAC)-based mass spectrometry has been shown to provide a powerful strategy for characterizing protein complexes and identifying specific interactions. Here, we show how SILAC can be combined with computational methods drawn from the business intelligence field for multidimensional data analysis to improve the discrimination between specific and nonspecific protein associations and to analyze dynamic protein complexes. A strategy is shown for developing a protein frequency library (PFL) that improves on previous use of static "bead proteomes." The PFL annotates the frequency of detection in co-immunoprecipitation and pulldown experiments for all proteins in the human proteome. It can provide a flexible and objective filter for discriminating between contaminants and specifically bound proteins and can be used to normalize data values and facilitate comparisons between data obtained in separate experiments. The PFL is a dynamic tool that can be filtered for specific experimental parameters to generate a customized library. It will be continuously updated as data from each new experiment are added to the library, thereby progressively enhancing its utility. The application of the PFL to pulldown experiments is especially helpful in identifying either lower abundance or less tightly bound specific components of protein complexes that are otherwise lost among the large, nonspecific background.

Near Real-Time Processing of Proteomics Data Using Hadoop.

This article presents a near real-time processing solution using MapReduce and Hadoop. The solution is aimed at some of the data management and processing challenges facing the life sciences community. Research into genes and their product proteins generates huge volumes of data that must be extensively preprocessed before any biological insight can be gained. In order to carry out this processing in a timely manner, we have investigated the use of techniques from the big data field. These are applied specifically to process data resulting from mass spectrometers in the course of proteomic experiments. Here we present methods of handling the raw data in Hadoop, and then we investigate a process for preprocessing the data using Java code and the MapReduce framework to identify 2D and 3D peaks.