Navigating through metaproteomics data: A logbook of database searching

Abstract

Metaproteomic research involves various computational challenges during the identification of fragmentation spectra acquired from the proteome of a complex microbiome. These issues are manifold and range from the construction of customized sequence databases, the optimal setting of search parameters to limitations in the identification search algorithms themselves. In order to assess the importance of these individual factors, we studied the effect of strategies to combine different search algorithms, explored the influence of chosen database search settings, and investigated the impact of the size of the protein sequence database used for identification. Furthermore, we applied de novo sequencing as a complementary approach to classic database searching. All evaluations were performed on a human intestinal metaproteome dataset. Pyrococcus furiosus proteome data were used to contrast database searching of metaproteomic data to a classic proteomic experiment. Searching against subsets of metaproteome databases and the use of multiple search engines increased the number of identifications. The integration of P. furiosus sequences in a metaproteomic sequence database showcased the limitation of the target-decoy-controlled false discovery rate approach in combination with large sequence databases. The selection of varying search engine parameters and the application of de novo sequencing represented useful methods to increase the reliability of the results. Based on our findings, we provide recommendations for the data analysis that help researchers to establish or improve analysis workflows in metaproteomics.

Comparative genomic, proteomic and exoproteomic analyses of three Pseudomonas strains reveals novel insights into the phosphorus scavenging capabilities of soil bacteria

“…Observing the global exoproteomic response of just three Pseudomonas species revealed new insights into the P scavenging capabilities of this genus and has provided a number of markers (Muth et al ., 2015) that can be utilized to investigate P‐mobilization directly in the rhizosphere. Given the enormous task of identifying proteins in situ from complex communities, the data presented in this paper will serve as a platform to investigate the key enzymes and microbial taxa involved in P‐mobilization at the level of functional entities (proteins) in situ .…”

“…The majority of exoproteins detected are associated with nutrient acquisition, motility, cell attachment, defence, communication as well as antagonism (Christie‐Oleza et al ., 2012). Therefore, exoproteomics is the ideal method of choice to study the bacterial mechanisms for scavenging extracellular P. Due to the complexity and technical challenges associated with metaproteomics (Muth et al ., 2015), this study aimed to identify Pi‐responsive proteins that can be used as markers for future studies investigating Pi mobilization within the rhizosphere. Therefore, exoproteomic analyses were performed on three plant‐associated Pseudomonas strains grown under Pi‐deplete conditions (50 μM).…”

The effect of microbial colonization on the host proteome varies by gastrointestinal location

“…DNA sequencing technologies have enabled the profiling of microbial communities and their association with human health and disease (Eckburg et al, 2005;Gill et al, 2006;Ley et al, 2006). Although these studies yielded key insights into host-microbe relationships, the direct effects that microbes have on host proteomes have only begun to be measured (Verberkmoes et al, 2009;Erickson et al, 2012;Lichtman et al, 2015;Muth et al, 2015). New tools to discover host-derived factors within the intestinal lumen hold tremendous potential for discovering molecular mediators of host-microbe interactions and deriving mechanistic understandings of this complex, biologically important system.…”

The challenge of metaproteomic analysis in human samples

“…Perhaps more importantly, use of such metaproteomic databases also creates additional problems because of low identification rates at the typical, stringent 1% FDR via classical targetdecoy-based approaches [19]. Moreover, iterative search strategies that expand into ever-increasing search spaces can lead to substantial underestimation of the actual FDR [19].…”