BioSunMS: a plug-in-based software for the management of patients information and the analysis of peptide profiles from mass spectrometry

System architecture

BioSunMS provides a relational database and client-server architecture suitable for multi-user workgroups. It is a RCP application extending the Eclipse framework. It was developed in Java, and provided a domain-specific platform where the plug-ins can be integrated. End-users can select related features from the BioSunMS plug-ins freely.

BioSunMS system includes four functional modules, namely, data management, spectrum processing, MS profile analysis and security module (Figure 1). The data management module provides a robust, client-server, relational database system for the management of patients information and MS data. Data are primarily stored in a relational database MySQL, while some raw data is stored on the file system (mainly to provide backup for MS data). The spectrum processing module performs spectrum import, spectrum export, and related analysis such as calibration, normalization and peak detection [20]. The MS profile module is designed for sample classification and identification of potential biomarkers. Background computations are mostly done by R package and libSVM [21]. BioSunMS has a security system for protecting the data. Access to any data can be controlled. Access is granted via user groups rather than individual users. The main features of BioSunMS can be found in Table 1.

User Interface

BioSunMS is built using Standard Widget Toolkit (SWT). Created as part of the Eclipse project, SWT allows developers to build efficient and portable applications. In contrast to Swing/AWT (which provide their own graphical environment), SWT has the same look and feel of the operating system on which the application runs. AWT/Swing components can be wrapped in SWT, and this feature is utilized in BioSunMS to integrate Java components built on these toolkits [22]. In an Eclipse RCP application, the user interface is composed of five main graphical units, namely, View, Editor, Perspective, Menu and Wizard. A View is a graphical window to provide information for the users. An Editor is another type of window to let the users edit and save data. A perspective is a visual container for a set of views and content editors. An example in BioSunMS is the Spectral Perspective, which displays Views, Editors, and Menu options for spectral data. Wizards are used to guide users to finish particular object through a sequential set of tasks. The internal placement and size of components within a perspective are not fixed but can be changed at the user preference and is saved between sessions.

Workflow

To access the BioSunMS server at client side, users must first log on with an authorized username and a given password. Users could also log on for analyzing spectral data without connecting to the server, when their computers are offline. The workflow of BioSunMS was illustrated in Figure 2, which was divided into four main parts as follows.

Enter new records

The first step in the workflow is to input patients general information and laboratory test results. To add a new record, click on the top-right button "Append" in the toolbar. A form will be displayed to enter data according to the module. Fields include equipment, sample ID, gender, age, specimen group and so on. Marked fields must be specified. Users could group samples with the Query View.

Import spectra to the BioSunMS

The second step in the workflow is to import spectral files to the BioSunMS. After importing a spectral file, BioSunMS uses the Bioconductor PROcess package developed by Li [23] to process the spectral data. The PROcess package contains a collection of functions for removing baseline drifts, detecting peaks and aligning peaks for a set of biomarkers. In the Spectra View which uses the JFreeChart package [24] for visualization of spectral information, users can identify or erase the peaks in a spectrum manually. Labeled peaks are shown in the Peak View. After data processing, the peak list can be saved to the database corresponding to the sample ID (Figure 3). Users can upload the "raw" spectral files to the ftp server for backup.

Group spectra into folders

The third step in the workflow is to group spectral files into the Bioresource folder. To analyze data, spectra must be grouped together into a folder. There are many ways to organize spectra by querying the database for spectra meeting desired criteria, such as research group, user, sample state, sample type, patient and characteristic description. Users can also define the condition by selecting a series of fields to be used to separate spectra into groups for classification and prediction. After grouping, a web page with a table of characteristics of patients can be created (Figure 4).

Apply analysis algorithms

The fourth step in the workflow is to analyze the experimental results for identification of potential biomarkers and early diagnosis. BioSunMS provides the Wizard for general users, including extraction of MS peaks, construction of m/z matrix, sample classification and sample prediction. Advanced users can process and analyze the MS data in the Data Analysis Perspective. After building a Support Vector Machine (SVM) model on the training dataset and make prediction on the test dataset, BioSunMS uses a Receiver Operation Characteristic (ROC) curve to assess the models performance (Figure 5).

Security and access

BioSunMS is a network-based system like an Intranet. It can be accessed from all computers from the same local network. It can even be accessed through the internet. Therefore, data access and management can be password protected. The system always stores actions made by any users. The administrator can check the action history at any time.

From the User and Group Management View, BioSunMS administrator can add or edit the user and the group. If a user group has been permitted to access a kind of data, any users belonging to the group are permitted to read the data in the Sample View. Any user who doesn't belong to the user group will not be permitted to read the data in the Sample View. Any user can create a new record, and modify a record which belongs to the same user group. However, only administrators can create new users and user groups. By default a new record is accessible to all users. To restrict access to a record, assign that record to the desired user group.

Sample and clinical information

Recording detailed information on collection, processing and storage of samples is crucial both for efficient reporting on biomedical study and for subsequent data analysis[25]. Many patients and sample related variables, such as gender, age, genetic factors, drug treatment, and symptom, dietary, and family history are particularly important in the context of clinical proteomics study. BioSunMS provides an efficient link between sample data and experiment results. Users can define the condition by selecting a series of fields to be used to automatically segregate spectra into groups. For example, in a typical profiling study, a condition might consist of the array type, sample type, different subjects (e.g. healthy and diseased), clinical laboratory result and diagnosis. After grouping the data by condition, BioSunMS will create a table with characteristics of the group of patients information (Figure 4). Finally, the data produced by the above steps can be exported for further processing outside of BioSunMS system using the export function provided by Bioresource View.

Pre-processing and visualization

BioSunMS is useful for the users with the need to store, process and analyze MALDI-TOF or SELDI-TOF MS data. It supports general mass spectrometry file format, such as mzML, mzXML, mzData and CSV. For mzML data format standard released by the HUPO-PSI and Institute for Systems Biology in June 2008, BioSunMS reads and writes it using a package of ProteomeCommons.org IO Framework [26] and ProteoWizard[27]. Currently, the majority of instruments store mass spectra in vendor specific formats. Users can convert raw file to universal file, for example, mzXML format using the converter programs provided by the vendors. At present, we have developed a plug-in which integrates CompassXport[28] for converting raw file of Bruker Corporation[29] to mzXML. Other plug-ins for format converting is under development.

BioSunMS provides basic visualization tools and advanced processing algorithms. The method and related procedure, BioConductor PROcess was used for baseline subtraction, smoothing, normalization and peaks identification. Our data visualization module aims to support analysts in finding interesting peaks in spectra, selecting them for further analysis and visualizing the selected peaks by automatic or manual detection.

Sample classification and sample prediction

The analysis of several spectra coming from biological samples belonging to different subjects (e.g. healthy and diseased) focuses on identifying discriminant values in spectra related to diseases. The BioSunMS allows users to select features using t-test, construct models using SVM, and predict new samples by the wizard with the default parameters. Users can compare the generalization performance of a range of classifiers by plotting their performance on the test set in ROC curve. BioSunMS also provides heat maps and hierarchical clustering. In the future, we will gradually incorporate classification system Tclass [30] and sample class discovery system SamCluster [31] into BioSunMS, which were developed by our center for gene expression profile-based analysis.

R package

BioSunMS uses the R statistical programming language and some third part packages for processing and analysis of MS data. We implemented the communication between the BioSunMS and the R package with the class ncbaSpanker.System.R.RPackage. This class provides some methods for executing R scripts and commands. All communication with R language takes place via the file system. In general, BioSunMS writes a R script, makes the R system execute the script file, and output the results. Then BioSunMS reads an R output file to retrieve the results. The RPackage class hides all the details of communication with R. So users with little knowledge about R language can accomplish their task easily. Because many programs for gene expression profile-based analysis have been provided in R package, and some algorithms can be directly used in peptide profiling analysis, we use R package as the background computation. In addition, many packages for MS data analysis and visualization are also provided, for example, GALGO[32], caMassClass [33] and so on. We will incorporate them into BioSunMS system in the future.

Development environment

Here we want to emphasize an important characteristic of BioSunMS. It can not only be used as the standard-alone program, but also can be a framework for developers. By taking full advantage of the editing and visualization components provided by Eclipse, the developers can focus entirely on the problems at hand. Moreover, BioSunMS was implemented in Eclipse platform, which made it more flexible and easier to adopt an algorithm as a plug-in. For example, we have successfully developed a plug-in, cn.org.biosun.knn, for sample classification using k-nearest neighbour (kNN) method.