MS Quality Monitoring System To

Apr 19, 2012 - Besides avoiding the loss or degradation of precious samples, a continuous monitoring system would have the advantage of maximizing ...
6 downloads 4 Views 3MB Size
Download PDF
Technical Note pubs.acs.org/jpr

SprayQc: A Real-Time LC−MS/MS Quality Monitoring System To Maximize Uptime Using Off the Shelf Components Richard A. Scheltema and Matthias Mann* Department of Proteomics and Signal Transduction, Max-Planck Institute of Biochemistry, Am Klopferspitz 18, D-82152 Martinsried, Germany ABSTRACT: With the advent of high-throughput mass spectrometry (MS)-based proteomics, the magnitude and complexity of the performed experiments has increased dramatically. Likewise, investments in chromatographic and MS instrumentation are a large proportion of the budget of proteomics laboratories. Guarding measurement quality and maximizing uptime of the LC−MS/MS systems therefore requires constant care despite automated workflows. We describe a real-time surveillance system, called SprayQc, that continuously monitors the status of the peripheral equipment to ensure that operational parameters are within an acceptable range. SprayQc is composed of multiple plug-in software components that use computer vision to analyze electrospray conditions, monitor the chromatographic device for stable backpressure, interact with a column oven to control pressure by temperature, and ensure that the mass spectrometer is still acquiring data. Action is taken when a failure condition has been detected, such as stopping the column oven and the LC flow, as well as automatically notifying the appropriate operator. Additionally, all defined metrics can be recorded synchronized on retention time with the MS acquisition file, allowing for later inspection and providing valuable information for optimization. SprayQc has been extensively tested in our laboratory, supports third-party plug-in development, and is freely available for download from http://sourceforge.org/projects/sprayqc. KEYWORDS: real-time monitoring, quality control, LC−MS, LTQ Orbitrap, column oven, temperature programming

■

INTRODUCTION Analysis of proteins by mass spectrometry has evolved from highly manual and specialized experiments for individually purified peptides or proteins to complex shotgun LC−MS/MS workflows in which thousands of proteins are measured in a single project.1−8 One aspect of this transition is reflected in the data volume, which has gone up from a few spectra to tens or hundreds of thousands of spectra over many runs. Nowadays laboratories often employ a number of mass spectrometers in parallel in an automated and continuous manner. This evolution has been aided by the higher reliability and robustness of chromatographic pumps and mass spectrometers over the past 10−20 years. However, for maximum performance current LC−MS/MS systems are often pushed to their limits. Current setups employ very high pressures and columns with small inner diameters as well as sophisticated mass spectrometers with complex electronic components, which can still make routine operation challenging. For instance, overpressure on the LC column will cause a run to abort, potentially leading to loss of precious or even irretrievable sample. Likewise, the electrospray can be compromised by droplet formation, which may affect data quality. Because of the automated nature of proteomics LC−MS/MS systems, these conditions are often detected only when data acquisition has © 2012 American Chemical Society

stopped for a considerable time or upon inspection of the already recorded data. That quality control is an important aspect of proteomics experiments has been recognized,9,10 and large lists of metrics have been proposed to ascertain whether the recorded data is of sufficient quality.11 A number of efforts have been directed at standardized quality control of proteomic data sets.12−19 These typically involve addition of standards or common references to the samples to be analyzed. In the most comprehensive effort to date, Paulovich et al. used the performance of the LC−MS/MS setup from a number of laboratories on yeast protein extracts as a common reference to ascertain quality data in other laboratories and to pinpoint failure areas.20 The above papers addressed quality control after data had already been acquired. This is an important step but does not address quality control of ongoing measurements. Besides avoiding the loss or degradation of precious samples, a continuous monitoring system would have the advantage of maximizing LC−MS/MS uptime. This equipment is still quite expensive and usually comprises the major non-personnel-related expense of proteomics laboratories (write down of a single system costing Received: December 11, 2011 Published: April 19, 2012 3458

dx.doi.org/10.1021/pr201219e | J. Proteome Res. 2012, 11, 3458−3466

Journal of Proteome Research

Technical Note

$600,000 and used for 5 years is more than $300 per day). Furthermore, the monitoring system can reduce wear and tear, for example, by avoiding overpressure and consequent potential damage to column and HPLC pump. Xu et al. inserted a bovine serum albumin (BSA) sample for quality control as every 10th run. If the BSA results do not pass a threshold, the queue can be aborted.21 While useful, this approach still introduces overhead (i.e., every 10th run is unproductive), it does not actually query the different components of the LC−MS/MS setup, and by its nature cannot prevent errors in ongoing measurements. The task of a real-time quality control system is to extract user-definable settings and operational parameters from the different equipment components, analyze these data for deviations from the norm, and provide early notification to the operator and ideally to take corrective action or indicate failure points in the equipment to the operator. The software also needs to be “lightweight”, i.e., it should not consume a large proportion of the acquisition computer’s resources. It is not intended to replace post acquisition quality control efforts, which can provide meaningful insights and statistics over a large batch of related samples. Here we introduce a real-time quality control system, which monitors HPLC pumps, the temperature of an LC column oven, the electrospray, and the continuity of data acquisition. “SprayQc” is realized as an open source project and has a plug-in-based, user-extensible software architecture. The current implementation provides support for the Easy-nLC (Thermo Scientific) and Eksigent pumps (AB Sciex) and Orbitrap family mass spectrometers (Thermo Scientific), but the software and concepts are easily extensible to a wide range of instrumentation. The SprayQc system is free of charge and open source, requires only inexpensive off the shelf components, and is provided with step-by-step instructions for setup and operation.

■

Figure 1. Design of the SprayQc application with diverse plug-ins. The system currently supports the Thermo Easy-nLC system, the video system of the Thermo Scientific nanosource system, the Sonation column oven, and Thermo Xcalibur RAW files. However, this opensource plug-in system can in principle be extended to any device. In cases where two or more plug-ins have collaborative functionality, secondary plug-ins (such the “nLC/Column oven” plug-in depicted at the top) connect the functionality of these plug-ins.

list. At this number of data points the information in the graph is not notably affected barring incidental outliers. All of these can be defined centrally in the application. Furthermore, we employ a sliding window of recent real-time data and operational parameters extracted from the peripheral equipment, which is analyzed to detect deviant behavior. These data are stored in a circular buffer ensuring that performance remains within required limits.

METHODS

Plug-in Development

The plug-in structure ensures that the application is easily extensible to include additional peripheral equipment into the application. To ease this process, well-documented interfaces and programming examples are provided with the distribution at http://sourceforge.net/projects/sprayqc allowing for third party development of plug-ins in any of the .NET programming languages. For support of additional devices it is necessary for the manufacturer to provide the appropriate software development kits (for short SDK; complete with documentation) to read out metrics from and, if available, to control the device. Most companies already do this, and the SDK can usually be requested through their website. With a physically present device a plugin can be developed and tested quickly.

SprayQc

The application SprayQc is implemented in the programming language C# using the Microsoft .NET framework. It is designed as a central application, where individual software modules are combined and provided with the required information (Figure 1). Such software modules are called plug-ins and are widely used in modern software design, web browsers being a prominent example. They can be developed independently adhering to common interface requirements, adding specific functionalities to the system. Each of the plugins is required to assume a role within this framework, which can be as basic as simply monitoring a peripheral device. A more extensive role is reserved for a single, privileged plug-in that is responsible for detecting and notifying the SprayQc application and all plug-ins about the overall state of the measurement, with the goal of synchronizing the application to the LC retention time. The software has a user-friendly interface (Figure 2) and makes use of a large number of real-time updated graphs displaying information from each component; for this we make use of the open-source library ZedGraph (http://sourceforge. net/projects/zedgraph/). To keep updates of these graphs efficient during long gradients, we implemented a time limit for when the equipment is idle. Functionality is provided to limit the number of displayed data points to a maximum, which is currently done by removing every fourth data point from the

SprayQc.filesystem

The task of this plug-in is to monitor continuing data acquisition by the mass spectrometer and prevent errors due to disk space shortage. During normal operation, the data acquisition software (Xcalibur in the case of Thermo mass spectrometers) continuously writes out the acquired data to a file, as the data is too large to be kept in memory. Such a file will be created before the measurement starts, and this event is used by the software as a cue to a pending measurement. The file system is therefore continuously monitored using .NET built-in functionality, which can be narrowed to a specific section (such as a directory where all files are written) and/or 3459

dx.doi.org/10.1021/pr201219e | J. Proteome Res. 2012, 11, 3458−3466

Journal of Proteome Research

Technical Note

Figure 2. Screenshot of the user interface of the SprayQc application. It contains a number of panels, providing information to the operator. Panel A contains a list of all active plug-ins and their associated controls. Panel B contains information about the measurement, such as the retention time and the measurement state. Panel C depicts the currently selected plug-in. Here the SprayQc.Easy.nLC overview control is shown, which provides information on this controller’s state (such as pump pressure, gradient, and progress on sample loading). Panel D contains a list of messages from the different plug-ins.

file extension (“RAW” for files generated with Xcalibur). From the filename a directory name is constructed where all telemetry data from the various components can be recorded for later analysis; the user can enable this centrally. In our laboratory a specific naming convention is used for the RAW files, which encodes information about the machine identification, sample type, and operator. The structure of this convention can easily be captured in an expression like the one below:

SprayQc.xcalibur

This module extracts limited parameters from the recorded MS data file. Access to the Thermo RAW files is obtained through the MSFileReader interface provided by Thermo. This is freely available from Thermo Scientific and provides the option to run SprayQc on a separate PC without the need to install Xcalibur. The RAW files are currently not readable during the measurement and can only be accessed when the measurement has been successfully finalized. Consequently, metrics from the RAW file are collected for the whole measurement and presented in a history of one or multiple RAW files.

{DATE}_{MACHINE}_{OPERATOR}_{SAMPLETYPE}_

SprayQc.notification Example filename: 20110503_Velos5_RiSc_SA_HeLa3to1.RAW

The notification module provides the functionality for automatically sending information to the operator (the person performing the proteomics experiment and/or the person responsible for the mass spectrometric equipment). The current implementation sends email notifications through the standard .NET SMTP functionality, for which the settings can be changed in the user interface; however, this could be expanded to SMS messages. This component maintains a list of operators, containing the operator ids and the associated email addresses. When a notification needs to be sent to the operator,

The reserved keywords between the curly brackets are used by the software to determine what kind of information to expect at each location. Anything outside these brackets is considered a literal in the string, and any data not captured by the expression is ignored (in this case “HeLa3to1.RAW”). As not every laboratory is expected to maintain the same convention this expression can be changed in the user interface, for example, by leaving out the machine element. 3460

dx.doi.org/10.1021/pr201219e | J. Proteome Res. 2012, 11, 3458−3466

Journal of Proteome Research

Technical Note

the operator id is extracted from the current filename (the element {USER} in the filename expression) and matched to the corresponding email address in the list. When no SMTP service is available, hMailServer (http:// www.hmailserver.com/) is a free and easily configurable alternative. This application runs on Microsoft Windows platforms and can also be used by this plug-in to send out email notifications.

detected by monitoring the pixel value distribution of the ambient light condition, for which the change is required to be 3 standard deviations from the previous value. When this is detected, the system ignores changes during 30 s in order stabilize to the new condition (see below).

■

RESULTS AND DISCUSSION Here we describe the application SprayQc, which monitors the operational parameters of the different devices making up the LC−MS/MS system for behavior deviant from the norm and which then provides early warning to the operator. It has been designed in a modular fashion, where each device is monitored by a single plug-in software component. Metrics are continuously calculated from the device-specific data and are used to define different failure scenarios, which would compromise results (Table 1). With the ability to synchronize

SprayQc.easy.nLC

This module continuously monitors telemetry data from the Easy-nLC liquid chromatograph (Proxeon Easy-nLC; now designated Easy-nLC, Thermo Scientific) and stops the pump when necessary. The data is acquired using the “HPLC communication interface” (lcremote.dll). This interface allows retrieval of metrics pertaining to the machine state, pump flows and pressure, solvent mixture, and the gradient at a frequency of 1 Hz. This frequency has been set so as not to impede the data acquisition process. The SprayQc.easy.nLC plug-in additionally provides extra information about the state of the device as a progress bar, such as the duration of the sample load.

Table 1. Different Parameters As Monitored by SprayQc device

task

SprayQc.filesystem SprayQc.xcalibur

SprayQc.sonation.oven

data acquisition continuity storage space/backup device specific settings operator read-out ionization efficiency temperature/vacuum/voltages stability spray continuity nozzle positioning device specific settings error-state detection pressure stability and level device specific settings temperature stability overpressure prevention

A column oven provides a stable temperature environment for LC separations and the option to dynamically alter the temperature to a desired value. The column oven module continuously monitors telemetry data from and sends commands to the column oven (Sonation GmbH) through the “COControl remote interface” supplied by the manufacturer. This interface allows the collection of data at any time point, but data collection frequency is reduced to 1 Hz to stay consistent with the SprayQc.easy.nLC module. The collected telemetry data consists of the minimum and maximum allowed temperature and the current temperature. The interface also allows turning the oven on or off. Additionally, we have integrated support for analytical column detection by RFID tags that can be attached to the analytical column and fit easily inside the oven. The software is then able to retrieve information of the column such as date of manufacture, length, inner diameter, bead size, bead type, and total runtime (dynamic property that can be updated by the application).

to the retention time, capture different events (such as the start and stop of sample loading) and change parameters of the peripheral devices, the application can also influence the measurements in a dynamic fashion. In the following the different failure scenarios are described and how the plug-ins react to them.

SprayQc.spray.video

File-System Analysis

This component monitors the tip of the electrospray needle in real time to detect a breakdown of the spray as manifested by the presence of a droplet. For this purpose we use a low cost video server, for example, Grand IP video server, which retails at about $150. The electrospray source (Themo Scientific) is already equipped with two built-in cameras recording the nanospray source (Thermo Scientific), which are normally used to position the spray needle, of which the side camera is used. When no such option exists, suitable cameras (designated “bullet camera”) retail at about $100. To retrieve and analyze the video signal (resolution of 320 × 240) we employ the .NET open source computer vision library AForge (http://www. aforgenet.com/). The acquired images are processed to detect the column tip and determine whether a droplet is present. To keep CPU requirements to a minimum, so as not to disturb the data acquisition process, 5 of the 25 frames per second are analyzed. To ensure reasonable lighting conditions, it is crucial that the red light source is centered toward the needle tip. This ensures that a usable image is acquired during night-time conditions when the ambient light is turned off. This is

The file-system is continuously monitored for new files that the data acquisition software is creating and for available storage. When a measurement is in progress the file is automatically tracked for the last write action until the measurement end trigger is received. The creation of a new RAW file is taken as a cue that a new measurement is pending. If there are no writes to the file for 30 min, a message is sent to the operator, indicating that most likely the data acquisition software has stopped functioning. Also, when for over 2 hours no new RAW file has been created after the last measurement end trigger, the operator is notified that the batch has potentially stopped unexpectedly. The file system where the RAW files are written is also monitored for available storage space. When this drops below a user-definable threshold (standard 4 Gb), the operator is notified. Additionally, there is the option to let SprayQc automatically delete the oldest RAW files in order to prevent run over of the storage space, data loss, and stopped runs. To avoid any data loss, there is an automatic backup option that writes RAW files to a different location, which is preferably

SprayQc.spray.video SprayQc.Easy.nLC

SprayQc.sonation.oven

3461

dx.doi.org/10.1021/pr201219e | J. Proteome Res. 2012, 11, 3458−3466

Journal of Proteome Research

Technical Note

Table 2. Performance of the Different Plugins of SprayQc on a Intel Dual Core 2.33 GHz with 2 Gb Internal Memory CPU [% of combined cores] SprayQc SprayQc.easy.nLc SprayQc.spray.video SprayQc.sonation.oven SprayQc.xcalibur SprayQc.notification

memory [Mb]

during measurement

post measurement

during measurement

post measurement

0 2 22 0 1 0

0 2 22 0 50 0

14 11 30 8 16 5

14 7 30 5 6 5

located on a server linked through a network file system protocol (i.e., SMB network drive) connection.

(Figure 3). The median of the greyscale pixels values (ranging between 0 and 255) in this area is defined as the ambient light value. This value is used to scale the image brightness. This stabilizes light conditions and ensures that standardized values can be used later on in the operations. An additional benefit is that large changes in ambient light conditions that would otherwise interfere with correct detection of the spray tip, such as a night guard turning off the light in the room, can be detected and taken into account. The column tip in the image is automatically detected with standard computer vision operations (for introduction see Szeliski27) as depicted in Figure 3. First a contrast-stretch improves the sharpness of the acquired image. Second, a Sobel edge detector is applied to highlight the outline of the column tip. Third, a threshold operation is applied with a user-definable cutoff value to binarize the image and remove background noise, after which a simple closing operator is applied to remove small gaps left over by the thresholding operation. Fourthly, a connected component algorithm is applied to detect the different elements in the image. Lastly, the element formed by the column tip is detected and its area is calculated by selecting the largest component on the left side of the image, around which a convex hull, an area with the property that any two contained points can be connected by a line inside the area, is constructed. (This necessitates that the camera is set up so the column tip starts in the upper left corner; see Figure 3). To prevent fluctuations in the calculated area caused by movement of the entire source, the area is limited to a width of 40 pixels of the originally detected column width. The final area is calculated from a running mean of 10 s to remove some of the larger variations due to noise in the recorded images. Droplet formation can be detected from the calculated area of the column tip. The presence of a droplet will expand the area, which is even exaggerated by the use of a convex hull making the algorithm sensitive to small droplets. The maximum allowed variation is 10%, for which a running mean over 30 s is employed to take slowly varying light conditions into account; these affect the binarization process and thus the calculated area. As the formation of a droplet occurs much faster, the calculated area will cross the threshold before it has the opportunity to catch up to the variation. When this event occurs a notification is triggered and the application waits until the running mean has stabilized. To avoid spurious notifications, this is disable for the first 5 min of the gradient (likely high water content, which promotes the formation of droplets), and the user can disable the notification completely or for the washout. Further features that aid in the setup are the detection of irregularities (e.g., hair) on the column tip and the distance of the column tip to the heated capillary.

RAW File Analysis

A limited set of metrics can be retrieved from an Xcalibur RAW file, prior to extensive processing by proteomics data analysis software such as MaxQuant.22 Of these metrics we have selected cycle time, ion inject time, and pressure and temperature of the Orbitrap analyzer as key performance indicators. Retrieval of these values is reasonably CPU intensive (Table 2); however, this operation is scheduled directly after a run has finished, when the data acquisition software is idling. An automatic warning is dispatched only for the temperature and pressure, as the other values are sample-dependent for which hard thresholds cannot be implemented. The graphs in this component provide an overview of the selected metrics for an adjustable number of samples. The cycle time is defined as the time it takes for a normal full mass range scan and the dependent scans (e.g., MS2 scans of ions selected from the full scans; also referred to as TopN where N indicates the number of dependent scans) to complete. When this time increases dramatically for the same type of sample and method, this could indicate problems in the operating software, which could require a reboot to be resolved, or less efficient ion transfer from the source to the mass spectrometer, which could require full maintenance of the machine. The ion inject times (i.e., the time the instrument spends to collect enough ions to reach a user-defined threshold) are collected for all cycles that reach the full number of MS2 scans, for example, 10 in a Top10 method. This ensures that the collected values pertain to the chromatographic area where peptides are eluting, which are meaningful for evaluating the performance of the machine. During normal operation with complex mixtures and automatic gain control, it is not expected that a large proportion MS2 scans reach the maximum allowed ion inject time (i.e. there are insufficient ions in comparison to the requested value) as this “maxing out” would indicate compromised spectrum quality. On the LTQ Orbitrap Velos platform25 the average ion inject time rises over days or weeks, which can be caused by sample residues coating the S-lens. The collected statistics can be used to determine when more elaborate cleaning is required. Video Data Analysis

The SprayQc.spray.video component constantly observes the electrospray and performs image analysis using computer vision. Before the column tip is extracted from the camera image, fluctuations in brightness caused by varying light conditions are stabilized to ensure that the image processing does not produce randomly changing artifacts. This is achieved by monitoring a user-definable area in the image containing only background, which serves to detect the ambient light level 3462

dx.doi.org/10.1021/pr201219e | J. Proteome Res. 2012, 11, 3458−3466

Journal of Proteome Research

Technical Note

whose responsibility it then is to stop the sequence, as it is currently not possible to control the data acquisition process and prevent further measurements. The early warning of such events and consequent abortion of the runs by the operator already helps to preserve samples. The most important of the defined metrics is the continuity of the gradient flow. When the pressure on the pumps goes over the maximum, the device stops all flow. This can be detected by the absence of flow telemetry data for the pumps. Failure to report can then either be caused by loss of network connection or by one of the pumps going into overpressure, either of which is destructive to the data acquisition process. Additionally, the pressure on the pumps is monitored to verify that they do not exceed a pressure limit (typically 90% of the allowed maximum pressure). When this occurs, a general warning is distributed to the other plug-ins that a catastrophic error is pending. For example the column oven can in such an event be used to increase the temperature of the oven, which reduces viscosity and decreases the pressure, which can save the measurement and the sample. Continuously monitoring the flow for each pump enables detection of a “backflow condition”. During a normal measurement backflow is not expected and only occurs when one of the rotary seals is leaking. An error condition is reported when non-negligible (>5 nL/min) and sustained (>60 s) backflow is detected. Each pump additionally has a limited syringe volume (in this case typically 140 μL), which needs to be replenished when the syringe runs empty before the gradient can be continued. Should this occur, most commonly for very long gradients, pressure is lost during refilling, and hence this condition is reported to the operator. The back pressure on the pumps should not fluctuate too radically in order to achieve stable and reproducible chromatography. Therefore, the reported pressure from the pumps is continuously monitored for such fluctuations. This is done by comparing the difference between the first and the third quantile of the last 10 pressure reports scaled to a percentage of the maximum pressure for each pump separately. When this difference exceeds 10%, the pressure is deemed to fluctuate strongly, leading to an operator notification. Likewise, the pressure must not drop too low. At a certain point stability is affected and this could point to a leak in the t-piece connection to the column. Therefore, when the pressure drops below a definable pressure (standard 70 bar), a warning is issued. Additionally, the gradient is inspected for deviations from the intended gradient. Should one of the pump seals be leaking, the reported flow for that pump will be lower than what is expected. This produces shifts in the actual gradient and irreproducible chromatography. When this shift exceeds 5% of the percentage buffer B, a warning is issued.

Figure 3. Droplet detection by computer vision. The column tip is recognized by the program from the recorded video images. (A) Contrast stretching is applied, providing the highest dynamic range possible in the image, and the user-definable ambient light area (marked in red) is analyzed. (B) Brightness is scaled using the background area. (C) Sobel edge detection highlights sharp contrasts, exposing the column and capillary. (D) A simple threshold is applied to generate a binary image. (E) “Closing” fills any gaps left after thresholding. (F) Connected components (marked red and green in the image) are detected and their pixels uniquely marked to locate column tip and capillary. (G) A convex hull (marked in blue) is constructed around the element making up the column tip, from which the tip area is calculated. To prevent variation in area due to moving of the column tip, the width of the convex hull is restricted to 40 pixels (marked in red).

Column Oven Monitoring and Control

To achieve high quality chromatography and stable retention times, the temperature of the column can be stabilized with a column oven. We use a Peltier element cooled and thermo element heated “oven”, which enables a constant and controllable temperature environment for short or long columns28 (see Methods). At the start of the measurement, the actual temperature of the column oven is recorded, and during the measurement the recorded temperature value cannot deviate more than 0.5 °C from the starting value. This deviation is much higher than the accuracy with which the

Thermo nLC Metrics Analysis

The correct operation of all the components in the nano liquid chromatography (nLC) device is required to achieve high quality chromatography. However, these components can fail, and other problems can develop over the course of a measurement resulting in diminished or corrupted data quality. To detect and provide notification of such problems at an early stage, diverse metrics from the nLC are continuously monitored. In all cases only a warning is issued to the operator, 3463

dx.doi.org/10.1021/pr201219e | J. Proteome Res. 2012, 11, 3458−3466

Journal of Proteome Research

Technical Note

Figure 4. Temperature program editor for the column oven. A simple editor is provided for temperature programming, which operates similarly to the gradient editors provided by some LC manufacturers. Apart from the temperature gradient synchronized to the retention time, the plug-in offers the option to use a different temperature during sample load.

save the run in progress and ensure analysis of further queued samples. We also provide the option of only enabling temperature increase during the wash-out phase of the gradient. The module furthermore provides options for controlling the temperature at certain events within the measurement: initialization, sample load, gradient, wash-out, as well as for when the flow goes above a predefined value. Of these the most important is the ability to increase the temperature during loading. With the current trend toward longer columns, this can help to reduce the loading times by reducing the limiting effect of the maximum pressure, while operating the gradient under the standard temperature conditions. In most laboratories diverse column types are used, depending on the application, and it would be desirable to have this information present for later analysis and comparison but also in real time. We therefore decided to equip each column with a reusable RFID chip to make it individually traceable. The analytical column information available through the fitted RFID chip with the column oven is then used by this plugin to allow lower pressure limits to be defined for different column types (based on the length, inner diameter and bead size; these factors together with the flow rate on the LC affect the backpressure). When the pressure on the LC drops below the defined lower limit there is most likely a leaking connection in the lines linking the LC to the analytical column. A warning message is automatically sent to the operator on this occasion. For those cases where the analytical column information is not available, a fallback lower pressure limit can be defined. The RFID chip implementation for tracing columns is an example for the general capability of SprayQC to incorporate information on individualized reagents.

temperature can be controlled by the column oven (SD 0.04) and below the point where the fluctuations start to affect overall chromatography. Additionally, this plug-in offers the possibility to control the temperature of the column oven during a measurement. For this purpose an editor and batch control component are provided for temperature programs (Figure 4). A temperature program consists of loading temperature, a stable temperature, and a temperature gradient, which is matched to the retention time of the method that is running. As soon as a temperature program is triggered from the batch control included in the software, the temperature is regulated on the basis of the program, overriding all other temperature settings. SprayQc.lc.temp

This module combines the functionality of the above two plugins and handles the regulation of the column oven at specific events from the LC device, by gathering information from both the LC and the column oven modules. When the LC goes into an error state caused, for example, by overpressure or missing start signals from the MS data acquisition software, the flow stops and the column may dry out in the heated environment of the oven, potentially destroying an already loaded sample and the column. The module prevents this by automatically disabling the oven (and pump). An additional option is to detect when the pressure on the pumps exceeds a userdefinable threshold (e.g., 90% of the maximum allowed pressure). To prevent overpressure, the software can automatically increase the temperature of the column oven to reduce the pressure, for example, to the current limit of 55 °C. Even though this will cause changes in the retention time of the peptides on the column, this option may be preferable as it will 3464

dx.doi.org/10.1021/pr201219e | J. Proteome Res. 2012, 11, 3458−3466

Journal of Proteome Research

■

CONCLUSIONS

■

AUTHOR INFORMATION

■

REFERENCES

(1) Biemann, K. The coming of age of mass spectrometry in peptide and protein chemistry. Protein Sci. 1995, 4 (9), 1920−7. (2) Burlingame, A. L.; Boyd, R. K.; Gaskell, S. J. Mass spectrometry. Anal. Chem. 1998, 70 (16), 647R−716R. (3) Aebersold, R.; Mann, M. Mass spectrometry-based proteomics. Nature 2003, 422 (6928), 198−207. (4) Bradshaw, R. A.; Burlingame, A. L. From proteins to proteomics. IUBMB life 2005, 57 (4−5), 267−72. (5) Cox, J.; Mann, M. Is proteomics the new genomics? Cell 2007, 130 (3), 395−8. (6) Fenyo, D.; Eriksson, J.; Beavis, R. Mass spectrometric protein identification using the global proteome machine. Methods Mol. Biol. 2010, 673, 189−202. (7) Nilsson, T.; Mann, M.; Aebersold, R.; Yates, J. R., 3rd; Bairoch, A.; Bergeron, J. J. Mass spectrometry in high-throughput proteomics: ready for the big time. Nat. Methods 2010, 7 (9), 681−5. (8) Savas, J. N.; Stein, B. D.; Wu, C. C.; Yates, J. R., 3rd Mass spectrometry accelerates membrane protein analysis. Trends Biochem. Sci. 2011, 36 (7), 388−96. (9) Martens, L. A report on the ESF workshop on quality control in proteomics. Mol. BioSyst. 2010, 6 (6), 935−8. (10) Martens, L.; Vizcaino, J. A.; Banks, R. Quality control in proteomics. Proteomics 2011, 11 (6), 1015−6. (11) Rudnick, P. A.; Clauser, K. R.; Kilpatrick, L. E.; Tchekhovskoi, D. V.; Neta, P.; Blonder, N.; Billheimer, D. D.; Blackman, R. K.; Bunk, D. M.; Cardasis, H. L.; Ham, A. J.; Jaffe, J. D.; Kinsinger, C. R.; Mesri, M.; Neubert, T. A.; Schilling, B.; Tabb, D. L.; Tegeler, T. J.; VegaMontoto, L.; Variyath, A. M.; Wang, M.; Wang, P.; Whiteaker, J. R.; Zimmerman, L. J.; Carr, S. A.; Fisher, S. J.; Gibson, B. W.; Paulovich, A. G.; Regnier, F. E.; Rodriguez, H.; Spiegelman, C.; Tempst, P.; Liebler, D. C.; Stein, S. E. Performance metrics for liquid chromatography-tandem mass spectrometry systems in proteomics analyses. Mol. Cell. Proteomics 2010, 9 (2), 225−41. (12) Tang, L.; Fitch, W. L.; Alexander, M. S.; Dolan, J. W. Expediting the method development and quality control of reversed-phase liquid chromatography electrospray ionization mass spectrometry for pharmaceutical analysis by using an LC/MS performance test mix. Anal. Chem. 2000, 72 (21), 5211−8. (13) Choi, B. K.; Ayer, M. B.; Siciliano, S.; Martin, J.; Schwartz, R.; Springer, J. P. Interpretation of high-throughput liquid chromatography mass spectrometry data for quality control analysis and analytical method development. Comb. Chem. High Throughput Screening 2005, 8 (6), 467−76. (14) Hong, H.; Dragan, Y.; Epstein, J.; Teitel, C.; Chen, B.; Xie, Q.; Fang, H.; Shi, L.; Perkins, R.; Tong, W. Quality control and quality assessment of data from surface-enhanced laser desorption/ionization (SELDI) time-of flight (TOF) mass spectrometry (MS). BMC Bioinf. 2005, 6 (Suppl 2), S5. (15) Bons, J. A.; de Boer, D.; van Dieijen-Visser, M. P.; Wodzig, W. K. Standardization of calibration and quality control using surface enhanced laser desorption ionization-time of flight-mass spectrometry. Clin. Chim. Acta 2006, 366 (1−2), 249−56. (16) Houde, D.; Kauppinen, P.; Mhatre, R.; Lyubarskaya, Y. Determination of protein oxidation by mass spectrometry and method transfer to quality control. J. Chromatogr. A 2006, 1123 (2), 189−98. (17) Vetter, W.; Gaul, S.; Melcher, J. Improved quality control in gas chromatography interfaced to stable isotope ratio mass spectrometry by application of derivative chromatography. Anal. Chim. Acta 2007, 590 (1), 49−54. (18) Cairns, D. A.; Perkins, D. N.; Stanley, A. J.; Thompson, D.; Barrett, J. H.; Selby, P. J.; Banks, R. E. Integrated multi-level quality control for proteomic profiling studies using mass spectrometry. BMC Bioinf. 2008, 9, 519. (19) Foster, J. M.; Degroeve, S.; Gatto, L.; Visser, M.; Wang, R.; Griss, J.; Apweiler, R.; Martens, L. A posteriori quality control for the curation and reuse of public proteomics data. Proteomics 2011, 11 (11), 2182−94.

With the recent proliferation of high-throughput proteomics experiments using LC−MS/MS equipment, quality control has become a key aspect to ensure that meaningful answers to the biological questions can be extracted from the recorded data. Any delays due to failing equipment and other problems need to be kept to a minimum to ensure that loaded samples are in the best condition for analysis. Additionally, the analysis time needs to be as short as possible to keep the experiments within budget and to ensure the optimal utilization of expensive proteomics equipment. The SprayQc system addresses this need by monitoring the peripheral equipment together with the mass spectrometer over the course of a large batch of measurements, without the direct involvement of the operator. The automatic notification of the responsible person for the measurement in the event of problems ensures that immediate action can be taken and keeps the time spent by the operator on checking the equipment to a bare minimum. Additionally, the recorded metrics from each of the peripheral machines assists in our efforts for off-line quality control and optimize data quality. Many more plug-ins for support of different devices and manufacturers can be envisioned. Currently, in addition to the Thermo Scientific Easy-nLC pumps and AB Sciex Eksigent pumps, work is planned for support of the Dionex chromatography devices as well as further integration with the column ovens product line (Sonation GmbH). This demonstrates that different hardware can readily be supported in the SprayQC framework. However, since SprayQc is opensource, in principle any device can be supported, and we extend an offer to other laboratories or companies to start developing appropriate plug-ins for their devices of interest. An area of active development is the mass spectrometry component, where we envisage a direct link to the mass spectrometer rather than a passive link through the metadata stored in the RAW file. This will allow us to do extend the functionality of the current Xcalibur plugin to a point where warnings pertaining to the acquired MS data can be generated during the run rather than afterward. Additionally, this will allow us to integrate automated analysis for known sample types, such as routine quality control standards.

Corresponding Author

*E-mail: [email protected]. Notes

The authors declare no competing financial interest.

■

Technical Note

ACKNOWLEDGMENTS

We thank scientists at Thermo Fisher Scientific Odense, especially Jesper Matthiesen; scientists at Sonation GmbH, especially Wolfgang Schrader; scientists at NYU Langone Medical Center, especially Steven Blais and Thomas Neubert; and our colleagues at the Max Planck Institute, especially Korbinian Mayr, Felix Meissner, Dirk Walther, and Nagarjuna Nagaraj for enduring early testing, help, and fruitful discussions. This project was supported by the European Commission’s seventh Framework Program PROteomics SPECification in Time and Space (PROSPECTS, HEALTH-F4-2008-201648). 3465

dx.doi.org/10.1021/pr201219e | J. Proteome Res. 2012, 11, 3458−3466

Journal of Proteome Research

Technical Note

(20) Paulovich, A. G.; Billheimer, D.; Ham, A. J.; Vega-Montoto, L.; Rudnick, P. A.; Tabb, D. L.; Wang, P.; Blackman, R. K.; Bunk, D. M.; Cardasis, H. L.; Clauser, K. R.; Kinsinger, C. R.; Schilling, B.; Tegeler, T. J.; Variyath, A. M.; Wang, M.; Whiteaker, J. R.; Zimmerman, L. J.; Fenyo, D.; Carr, S. A.; Fisher, S. J.; Gibson, B. W.; Mesri, M.; Neubert, T. A.; Regnier, F. E.; Rodriguez, H.; Spiegelman, C.; Stein, S. E.; Tempst, P.; Liebler, D. C. Interlaboratory study characterizing a yeast performance standard for benchmarking LC−MS platform performance. Mol. Cell. Proteomics 2010, 9 (2), 242−54. (21) Xu, H.; Freitas, M. A. Automated diagnosis of LC−MS/MS performance. Bioinformatics 2009, 25 (10), 1341−3. (22) Cox, J.; Mann, M. MaxQuant enables high peptide identification rates, individualized p.p.b.-range mass accuracies and proteome-wide protein quantification. Nat. Biotechnol. 2008, 26 (12), 1367−72. (23) Olsen, J. V.; de Godoy, L. M.; Li, G.; Macek, B.; Mortensen, P.; Pesch, R.; Makarov, A.; Lange, O.; Horning, S.; Mann, M. Parts per million mass accuracy on an Orbitrap mass spectrometer via lock mass injection into a C-trap. Mol. Cell. Proteomics 2005, 4 (12), 2010−21. (24) Schlosser, A.; Volkmer-Engert, R. Volatile polydimethylcyclosiloxanes in the ambient laboratory air identified as source of extreme background signals in nanoelectrospray mass spectrometry. J. Mass Spectrom. 2003, 38 (5), 523−5. (25) Olsen, J. V.; Schwartz, J. C.; Griep-Raming, J.; Nielsen, M. L.; Damoc, E.; Denisov, E.; Lange, O.; Remes, P.; Taylor, D.; Splendore, M.; Wouters, E. R.; Senko, M.; Makarov, A.; Mann, M.; Horning, S. A dual pressure linear ion trap Orbitrap instrument with very high sequencing speed. Mol. Cell. Proteomics 2009, 8 (12), 2759−69. (26) Cox, J.; Michalski, A.; Mann, M. Software lock mass by twodimensional minimization of peptide mass errors. J. Am. Soc. Mass Spectrom. 2011, 22 (8), 1373−80. (27) Szeliski, R., Computer Vision: Algorithms and Applications; Springer: New York, 2010. (28) Thakur, S. S.; Geiger, T.; Chatterjee, B.; Bandilla, P.; Frohlich, F.; Cox, J.; Mann, M. Deep and highly sensitive proteome coverage by LC−MS/MS without prefractionation. Mol. Cell. Proteomics 2011, 10 (8), M110 003699.

3466

dx.doi.org/10.1021/pr201219e | J. Proteome Res. 2012, 11, 3458−3466