Johan Teleman

Data processing methods and quality control strategies for label-free LC-MS protein quantification.

Protein quantification using different LC-MS techniques is becoming a standard practice. However, with a multitude of experimental setups to choose from, as well as a wide array of software solutions for subsequent data processing, it is non-trivial to select the most appropriate workflow for a given biological question. In this review, we highlight different issues that need to be addressed by software for quantitative LC-MS experiments and describe different approaches that are available. With focus on label-free quantification, examples are discussed both for LC-MS/MS and LC-SRM data processing. We further elaborate on current quality control methodology for performing accurate protein quantification experiments. This article is part of a Special Issue entitled: Computational Proteomics in the Post-Identification Era. Guest Editors: Martin Eisenacher and Christian Stephan.

DIANA—algorithmic improvements for analysis of data-independent acquisition MS data

MOTIVATION Data independent acquisition mass spectrometry has emerged as a reproducible and sensitive alternative in quantitative proteomics, where parsing the highly complex tandem mass spectra requires dedicated algorithms. Recently, targeted data extraction was proposed as a novel analysis strategy for this type of data, but it is important to further develop these concepts to provide quality-controlled, interference-adjusted and sensitive peptide quantification. RESULTS We here present the algorithm DIANA and the classifier PyProphet, which are based on new probabilistic sub-scores to classify the chromatographic peaks in targeted data-independent acquisition data analysis. The algorithm is capable of providing accurate quantitative values and increased recall at a controlled false discovery rate, in a complex gold standard dataset. Importantly, we further demonstrate increased confidence gained by the use of two complementary data-independent acquisition targeted analysis algorithms, as well as increased numbers of quantified peptide precursors in complex biological samples. AVAILABILITY AND IMPLEMENTATION DIANA is implemented in scala and python and available as open source (Apache 2.0 license) or pre-compiled binaries from http://quantitativeproteomics.org/diana. PyProphet can be installed from PyPi (https://pypi.python.org/pypi/pyprophet). SUPPLEMENTARY INFORMATION Supplementary data are available at Bioinformatics online.

Automated Selected Reaction Monitoring Software for Accurate Label-Free Protein Quantification.

Selected reaction monitoring (SRM) is a mass spectrometry method with documented ability to quantify proteins accurately and reproducibly using labeled reference peptides. However, the use of labeled reference peptides becomes impractical if large numbers of peptides are targeted and when high flexibility is desired when selecting peptides. We have developed a label-free quantitative SRM workflow that relies on a new automated algorithm, Anubis, for accurate peak detection. Anubis efficiently removes interfering signals from contaminating peptides to estimate the true signal of the targeted peptides. We evaluated the algorithm on a published multisite data set and achieved results in line with manual data analysis. In complex peptide mixtures from whole proteome digests of Streptococcus pyogenes we achieved a technical variability across the entire proteome abundance range of 6.5–19.2%, which was considerably below the total variation across biological samples. Our results show that the label-free SRM workflow with automated data analysis is feasible for large-scale biological studies, opening up new possibilities for quantitative proteomics and systems biology.

Numerical Compression Schemes for Proteomics Mass Spectrometry Data

The open XML format mzML, used for representation of MS data, is pivotal for the development of platform-independent MS analysis software. Although conversion from vendor formats to mzML must take place on a platform on which the vendor libraries are available (i.e. Windows), once mzML files have been generated, they can be used on any platform. However, the mzML format has turned out to be less efficient than vendor formats. In many cases, the naïve mzML representation is fourfold or even up to 18-fold larger compared with the original vendor file. In disk I/O limited setups, a larger data file also leads to longer processing times, which is a problem given the data production rates of modern mass spectrometers. In an attempt to reduce this problem, we here present a family of numerical compression algorithms called MS-Numpress, intended for efficient compression of MS data. To facilitate ease of adoption, the algorithms target the binary data in the mzML standard, and support in main proteomics tools is already available. Using a test set of 10 representative MS data files we demonstrate typical file size decreases of 90% when combined with traditional compression, as well as read time decreases of up to 50%. It is envisaged that these improvements will be beneficial for data handling within the MS community.

Data Processing Has Major Impact on the Outcome of Quantitative Label-Free LC-MS Analysis

High-throughput multiplexed protein quantification using mass spectrometry is steadily increasing in popularity, with the two major techniques being data-dependent acquisition (DDA) and targeted acquisition using selected reaction monitoring (SRM). However, both techniques involve extensive data processing, which can be performed by a multitude of different software solutions. Analysis of quantitative LC-MS/MS data is mainly performed in three major steps: processing of raw data, normalization, and statistical analysis. To evaluate the impact of data processing steps, we developed two new benchmark data sets, one each for DDA and SRM, with samples consisting of a long-range dilution series of synthetic peptides spiked in a total cell protein digest. The generated data were processed by eight different software workflows and three postprocessing steps. The results show that the choice of the raw data processing software and the postprocessing steps play an important role in the final outcome. Also, the linear dynamic range of the DDA data could be extended by an order of magnitude through feature alignment and a charge state merging algorithm proposed here. Furthermore, the benchmark data sets are made publicly available for further benchmarking and software developments.

Dinosaur: A Refined Open-Source Peptide MS Feature Detector

In bottom-up mass spectrometry (MS)-based proteomics, peptide isotopic and chromatographic traces (features) are frequently used for label-free quantification in data-dependent acquisition MS but can also be used for the improved identification of chimeric spectra or sample complexity characterization. Feature detection is difficult because of the high complexity of MS proteomics data from biological samples, which frequently causes features to intermingle. In addition, existing feature detection algorithms commonly suffer from compatibility issues, long computation times, or poor performance on high-resolution data. Because of these limitations, we developed a new tool, Dinosaur, with increased speed and versatility. Dinosaur has the functionality to sample algorithm computations through quality-control plots, which we call a plot trail. From the evaluation of this plot trail, we introduce several algorithmic improvements to further improve the robustness and performance of Dinosaur, with the detection of features for 98% of MS/MS identifications in a benchmark data set, and no other algorithm tested in this study passed 96% feature detection. We finally used Dinosaur to reimplement a published workflow for peptide identification in chimeric spectra, increasing chimeric identification from 26% to 32% over the standard workflow. Dinosaur is operating-system-independent and is freely available as open source on https://github.com/fickludd/dinosaur.

Improvements in Mass Spectrometry Assay Library Generation for Targeted Proteomics

In data-independent acquisition mass spectrometry (DIA-MS), targeted extraction of peptide signals in silico using mass spectrometry assay libraries is a successful method for the identification and quantification of proteins. However, it remains unclear if high quality assay libraries with more accurate peptide ion coordinates can improve peptide target identification rates in DIA analysis. In this study, we systematically improved and evaluated the common algorithmic steps for assay library generation and demonstrate that increased assay quality results in substantially higher identification rates of peptide targets from mouse organ protein lysates measured by DIA-MS. The introduced changes are (1) a new spectrum interpretation algorithm, (2) reapplication of segmented retention time normalization, (3) a ppm fragment mass error matching threshold, (4) usage of internal peptide fragments, and (5) a multilevel false discovery rate calculation. Taken together, these changes yielded 14-36% more identified peptide targets at 1% assay false discovery rate and are implemented in three new open source tools, Fraggle, Tramler, and Franklin, available at https://github.com/fickludd/eviltools . The improved algorithms provide ways to better utilize discovery MS data, translating to substantially increased DIA performance and ultimately better foundations for drawing biological conclusions in DIA-based experiments.

Representation of selected-reaction monitoring data in the mzQuantML data standard

The mzQuantML data standard was designed to capture the output of quantitative software in proteomics, to support submissions to public repositories, development of visualization software and pipeline/modular approaches. The standard is designed around a common core that can be extended to support particular types of technique through the release of semantic rules that are checked by validation software. The first release of mzQuantML supported four quantitative proteomics techniques via four sets of semantic rules: (i) intensity‐based (MS1) label free, (ii) MS1 label‐based (such as SILAC or N15), (iii) MS2 tag‐based (iTRAQ or tandem mass tags), and (iv) spectral counting. We present an update to mzQuantML for supporting SRM techniques. The update includes representing the quantitative measurements, and associated meta‐data, for SRM transitions, the mechanism for inferring peptide‐level or protein‐level quantitative values, and support for both label‐based or label‐free SRM protocols, through the creation of semantic rules and controlled vocabulary terms. We have updated the specification document for mzQuantML (version 1.0.1) and the mzQuantML validator to ensure that consistent files are produced by different exporters. We also report the capabilities for production of mzQuantML files from popular SRM software packages, such as Skyline and Anubis.

Automated quality control system for LC-SRM setups.

UNLABELLED Selected reaction monitoring (SRM) is emerging as a standard tool for high-throughput protein quantification. For reliable and reproducible SRM protein quantification it is essential that system performance is stable. We present here a quality control workflow that is based on repeated analysis of a standard sample to allow insight into the stability of the key properties of a SRM setup. This is supported by automated software to monitor system performance and display information like signal intensities and retention time stability over time, and alert upon deviations from expected metrics. Utilising the software to evaluate 407 repeated injections of a standard sample during half a year, outliers in relative peptide signal intensities and relative peptide fragment ratios are identified, indicating the need for instrument maintenance. We therefore believe that the software could be a vital and powerful tool for any lab regularly performing SRM, increasing the reliability and quality of the SRM platform. BIOLOGICAL SIGNIFICANCE Selected reaction monitoring (SRM) mass spectrometry is becoming established as a standard technique for accurate protein quantification. However, to achieve the required quantification reproducibility of the liquid chromatography (LC)-SRM setup, system performance needs to be monitored over time. Here we introduce a workflow with associated software to enable automated monitoring of LC-SRM setups. We believe that usage of the presented concepts will further strengthen the role of SRM as a reliable tool for protein quantification. This article is part of a Special Issue entitled: Standardization and Quality Control in Proteomics.

Targeted proteomics and absolute protein quantification for the construction of a stoichiometric host-pathogen surface density model

Sepsis is a systemic immune response responsible for considerable morbidity and mortality. Molecular modeling of host-pathogen interactions in the disease state represents a promising strategy to define molecular events of importance for the transition from superficial to invasive infectious diseases. Here we used the Gram-positive bacterium Streptococcus pyogenes as a model system to establish a mass spectrometry based workflow for the construction of a stoichiometric surface density model between the S. pyogenes surface, the surface virulence factor M-protein, and adhered human blood plasma proteins. The workflow relies on stable isotope labeled reference peptides and selected reaction monitoring mass spectrometry analysis of a wild-type strain and an M-protein deficient mutant strain, to generate absolutely quantified protein stoichiometry ratios between S. pyogenes and interacting plasma proteins. The stoichiometry ratios in combination with a novel targeted mass spectrometry method to measure cell numbers enabled the construction of a stoichiometric surface density model using protein structures available from the protein data bank. The model outlines the topology and density of the host-pathogen protein interaction network on the S. pyogenes bacterial surface, revealing a dense and highly organized protein interaction network. Removal of the M-protein from S. pyogenes introduces a drastic change in the network topology, validated by electron microscopy. We propose that the stoichiometric surface density model of S. pyogenes in human blood plasma represents a scalable framework that can continuously be refined with the emergence of new results. Future integration of new results will improve the understanding of protein-protein interactions and their importance for bacterial virulence. Furthermore, we anticipate that the general properties of the developed workflow will facilitate the production of stoichiometric surface density models for other types of host-pathogen interactions.