Software:SpectraST

From SPCTools

Revision as of 18:52, 16 June 2008

SpectraST (short for "Spectra Search Tool" and rhymes with "contrast") is a spectral library building and searching tool designed primarily for shotgun proteomics applications. It is developed at the Institute for Systems Biology (ISB), in the research group of Professor Ruedi Aebersold. The main developer is Henry Lam.

The latest version of SpectraST is 3.0, to be released sometime in 2008, pending the publication of a manuscript. It is distributed by ISB under the LPGL license, as a component of the Trans Proteomic Pipeline (TPP) suite of software, distributed under the same license. The source code repository is at [1], and the official download site for the Windows installer is at [2]. Currently, the released version (2.0) with TPP can only perform library searching and does not have library creation functionalities.

Introduction to Shotgun Proteomics and Spectral Searching

The goal of proteomics is the systematic identification and quantification of all proteins in a biological system. In one of the most frequently practiced workflows, commonly known as shotgun proteomics, a protein sample of interest is first digested with a proteolytic enzyme (trypsin being the most common) to yield peptides that are amenable to LC-MS/MS analysis. The peptides in the resulting mixture are chromatographically resolved, ionized by techniques such as electrospray ionization (ESI) or matrix-assisted laser desorption ionization (MALDI) before being analyzed by a mass spectrometer. A fraction of the peptide ions are selectively isolated by the mass spectrometer and subjected to collision-induced dissociation (CID), in which the peptide ions are bombarded with noble gas atoms to induce fragmentation. (Other types of fragmentation techniques are also rapidly maturing.) The fragment ions are detected and reported by the mass spectrometer as tandem mass (MS/MS) spectra. Because peptide ions tend to fragment mostly along the peptide backbone in a somewhat predictable manner, the MS/MS spectra contain information that can be used to deduce the peptide sequence.

Traditionally, the inference of the peptide sequence from its characteristic tandem mass spectra is done by sequence (database) searching. In sequence searching, a target protein (or translated DNA) database is used as a reference to generate all possible putative peptide sequences by in silico digestion. The search engines then use various rules to predict the theoretical fragmentation pattern of each of these putative peptides, and compare the experimentally observed MS/MS spectra to these theoretical spectra one-by-one. Presumably, a positive identification is made if the experimental spectrum is sufficiently similar to one of the theoretical spectra. Several popular computational tools developed for this purpose have emerged over the years, each employing different algorithms and heuristics to achieve an acceptable balance of sensitivity and accuracy. Unfortunately, traditional sequence searching is a challenging, error-prone, and computationally expensive exercise. Despite the tremendous improvement in computer hardware and software over the past decade, this step often remains the bottleneck of any given proteomics experiment. The requirement of computational resources is also substantial, limiting the use of this powerful technique to only those research groups that can afford the costly computational infrastructure.

Spectral searching is an alternative approach that promises to address some of the shortcomings of sequence searching. In spectral searching, a spectral library is meticulously compiled from a large collection of previously observed and identified peptide MS/MS spectra. The unknown spectrum can then by identified by comparing it to all the candidates in the spectral library for the best match. This approach has been commonly employed for mass spectrometric analysis of small molecules with great success, but has only become possible for proteomics very recently. The chief difficulty, that of generating enough high-quality experimental spectra for compilation into spectral libraries, has been overcome by the recent explosion of proteomics data and the availability of public data repositories. Several attempts at creating and searching spectral libraries in the context of proteomics have been published within the past year, all demonstrating the tremendous improvement in search speed and the great potential of this method in complementing, if not replacing, sequence searching in many proteomics applications.

Advantages of Spectral Searching

1. Speed

Spectral searching benefits from a much reduced search space compared to sequence searching. In spectral searching, only peptide ions that are observed and identified in previous experiments will be included in spectral libraries and considered as candidates, whereas in sequence searching, all putative peptide sequences -- plus all permutations of post-translational modification sites, if specified -- in a protein database are considered. Most of these putative peptide ions considered in sequence searching are never observed in practice for a variety of reasons. With typical search parameters, the search space of spectral searching can be several orders of magnitude smaller. It is therefore not surprising that spectral searching can also be several orders of magnitude faster. SpectraST can achieve a top speed of 0.001 to 0.01 second per query spectrum (against a library of about 50,000 entries) on a modern personal computer. In contrast, SEQUEST, one of the most popular sequence search engine, needs about 5 to 20 seconds per query spectrum (against a human IPI database).

2. Preciseness

Spectral searching compares experimental spectra to experimental spectra; sequence searching compares experimental spectra to theoretical spectra. In general, the theoretical spectra considered in sequence searching are very simplistic (e.g., only including b- and y-type ions, at a fixed intensity), and do not resemble the experimental spectra that they are supposed to match. On the other hand, armed with previously observed experimental spectra compiled into spectral libraries, spectral searching can take full advantage of all spectral features, including actual peak intensities, neutral losses from fragments, and various uncommon or even unknown fragments, to determine the best match. The similarity scoring of spectral searching is therefore more precise, and will generally provide better discrimination between good and bad matches. This usually results in much superior statistics (e.g., sensitivity, false discovery rates) for the search results, compared to sequence searching.

Versions

What's new in SpectraST 3.0

• Creating libraries from sequence search results
• Library manipulation
  • Union/Intersect/Subtract operations
  • Consensus/Best-replicate library building
  • Filtering based on criteria
  • Quality filters
• Importing libraries from X!Hunter and BiblioSpec formats
• File list feature
• Logging
• Lib2HTML utility for visualizing library
• Monoisotopic mass support
• Various bug fixes and performance enhancements

What’s new in SpectraST 2.0

• Binary library format, enabling speed gain
• Library information and statistics in preambles of .sptxt and .pepidx files
• Searching of .dta files
• Detecting homologs in hit list
• Various bug fixes and performance enhancements

User's Guide

Installing SpectraST

SpectraST is an integral component of the Trans Proteomic Pipeline suite of software. Although it can be used alone without other TPP components, SpectraST users are strongly encouraged to download and install the entire TPP suite, which provides other useful functionalities such as raw data importation, automatic validation of search results, protein inference, and quantification and visualization.

Windows users: SpectraST is available as part of TPP for Windows, which is run in the cygwin (UNIX emulator) environment. Download the cygwin installer and follow installation instructions.

UNIX/LINUX users: Visit the Sashimi project page on SourceForge.net, and download the code as a tarball directly. Compiling, installation and configuration information is available in the README file.

Running SpectraST

SpectraST has two modes, the Create mode and the Search mode. In the former, SpectraST creates a searchable spectral library from various formats to prepare for searching. In the latter, SpectraST takes in unknown spectra and searches each of them against the spectral library.

The simplest way of running SpectraST is from the command line of your UNIX/LINUX or cygwin shell. The general usage is:

spectrast <options> <list of files of appropriate formats>

Options must be separated by space, and all begin with a hyphen ('-'). Search mode options always have an 's' following the hyphen; Create mode options a 'c'. SpectraST will perform the appropriate action based on the options specified, and complain when there are problems interpreting the command statement. The usage statement, and a list of options can be viewed by issuing the command spectrast by itself.

Once TPP is installed, SpectraST can also be run from the Petunia web interface, with limited options.

SpectraST Search Mode

SpectraST can perform spectral searching from the following data formats:

• .mzXML (all versions) format
• .mzData format
• .dta (SEQUEST) format, a simple peak list preceded by precursor information
• NIST (National Institute of Standards and Technology)’s .msp format

It requires a spectral library in SpectraST’s .splib format, which can be created in SpectraST Create Mode.

The results can be outputted to the following formats:

• .pepXML format
• .txt format, a fixed-width column text format
• .xls format, a tab-delimited column text format

The search mode is initiated with the option -s, or any of the search mode options. For instance, to search the MS/MS spectra in the file foo.mzXML against the spectral library bar.splib, using the parameters specified in the file spectrast.params, the command is simply:

spectrast -sFspectrast.params -sLbar.splib foo.mzXML

In the above, -sF and -sL are search mode options that the user can specify to customize the behavior of SpectraST. SpectraST will search all the MS/MS spectra in the file foo.mzXML against the spectral library bar.splib, using the parameters specified in the file spectrast.params. The result will be written to a file named foo.<ext> in the same directory where <ext> specifies the output format (.xml, .txt or .xls).

For a full list of options, see SpectraST Options.

SpectraST Create Mode

Importing Existing Libraries

SpectraST can create a searchable spectral library from the following formats:

• NIST (National Institute of Standards and Technology)'s .msp format (Download here)
• X!Hunter's .hlf format [3]
• BiblioSpec’s .ms2 format [4]

If files of these extensions are supplied, SpectraST simply converts those spectral libraries into a form suitable for SpectraST searches (.splib formats). (Note however that there is no study on how well SpectraST works with X!Hunter and BiblioSpec libraries.) For instance, to import the NIST yeast consensus library, and call the resulting library bar.splib and put it in the directory /dir/, the command is:

spectrast -cN/dir/bar yeast_consensus.msp

When it is done, it produces 5 files in the directory /dir/. The file bar.splib is the library itself; it’s in a binary (machine-readable) format. The file bar.sptxt is a text (human-readable) version of bar.splib. This .sptxt file is of no use to SpectraST; it can be deleted after manual inspection. The files bar.spidx and bar.pepidx are indices on the precursor m/z value and peptide, respectively. Keep the indices and the .splib file in the same directory for SpectraST to function properly. Lastly, a file spectrast.log is also created to document the command executed. Some useful information about the library is printed at the beginning of the bar.sptxt and bar.pepidx.

For a full list of SpectraST options, see SpectraST Options.

Creating Libraries from Sequence Search Results

SpectraST can create a spectral library from a .pepXML file, which contains peptide identifications from a previous shotgun proteomics experiment. For this purpose, it is preferable that the .pepXML has been processed with PeptideProphet, such that all the search hits have probabilities assigned. When importing from a .pepXML file, SpectraST scans through the .pepXML file for confident identifications, and attempts to extract the corresponding experimental spectra from .mzXML files. For instance, the command

spectrast -cNraw -cP0.9 dataset1.xml

will import all peptide identifications with probability at or above 0.9 from the file dataset1.xml, and put them in a library called raw.splib (with the accompanying raw.sptxt, raw.spidx and raw.pepidx files).

For a full list of SpectraST options, see SpectraST Options.

Manipulating SpectraST Libraries

SpectraST can convert one or more .splib libraries to another, performing various operations. For instance, to create a consensus library from all the entries in bar.splib and foo.splib, the command is:

spectrast -cNconsensus -cJU -cAC bar.splib foo.splib

SpectraST will take the union (specified by the option -cJU) of all the entries in bar.splib and foo.splib, and wherever a certain peptide ion is present as multiple entries (replicates), it will coalesce the replicates into a single consensus spectrum (specified by -cAC).

Some additional examples:

spectrast -cNphospho -cf”Mods =~ Phospho” bar.splib

This will screen the library bar.splib for all entries with a phosphorylation modification, and put the phosphopeptides in the library phospho.splib.

spectrast -cNcommon -cJI dataset1.splib dataset2.splib

This will take the intersection of the two libraries dataset1.splib and dataset2.splib, and put all entries of peptide ions that are seen in both files in the library common.splib.

spectrast -cNquality -cAQ -cL2 bar.splib

This will apply SpectraST’s quality filters to the library bar.splib; only those entries that pass the first 2 quality filters will be included in the library quality.splib.

For a full list of SpectraST options, see SpectraST Options. For a typical recipe for creating consensus libraries from sequence search results, see Creating Consensus Libraries.

Creating Consensus Libraries

A recipe for creating consensus libraries from TPP-processed sequence search results is detailed here. Consider the following example:

The following commands should be issued in succession:

? -cNrawA -cnAlpha
A-SEQ.xml
A-MAS.xml
? -cNrawB -cnBeta
B1.xml
B2.xml
? -cNrawG -cnGamma
G.xml
? -cJU -cAC -cNconsABC
rawA.splib
rawB.splib
rawC.splib
? -cAQ -cNconsABC_Q
consABC.splib

1. Importing the raw spectra into SpectraST
spectrast -cNrawA -cnAlpha A-SEQ.xml A.MAS.xml
spectrast -cNrawB -cnBeta B1.xml B2.xml
spectrast -cNrawG -cnGamma G.xml

These commands will create the raw libraries rawA.splib, rawB.splib and rawC.splib. Identifications from multiple .pepXML files of the same dataset are imported with the same dataset identifier. The same query with identifications from multiple search engines will be combined intelligently. The probability threshold above which identifications are imported can be specified with the option -cP<prob>, which defaults to 0.9. This will not coalesce replicates of the same peptide ion identification into a consensus spectrum yet. Remember that the .mzXML files must be in the same directories as their corresponding .pepXML files.

2. Creating a consensus spectral library
spectrast -cJU -cAC -cNconsABC raw*.splib

This will combine the three raw libraries, then replace multiple replicates of the same peptide ion identification with a consensus spectrum. Many options are available to fine-tune the algorithm; however, the default parameters are usually adequate.

3. Performing quality control of the consensus spectral library
spectrast -cAQ -cNconsABC_Q consABC.splib

This will run the consensus spectra through SpectraST's quality filters. With the default settings, spectra failing either or both of the first 2 filters will be removed, and spectra failing any of the other filters will be marked. Different quality levels can be set with the options -cL and -cl. It is recommended that a consensus spectral library is subject to some quality control before using it in spectral searching; the optimal quality level reflects the user's desired compromise between library comprehensiveness and library quality. This is to minimize mis-identified and low-quality spectra in the library. These questionable spectra can propagate errors from sequence searching, reduce the discriminating power of the spectral search engine, and induce false positive and false negative hits.

SpectraST Parameter File

SpectraST allows the use of parameter files to simplify the process of spectral library building and searching. Namely, desired options can be specified in a text file, and supplied to SpectraST every time the same action is performed, saving the user from having to specify lengthy list of command-line options. To invoke the parameter files, specify the options -sF<parameter file> and -cF<parameter file> for Search Mode and Create Mode, respectively. Exemplary parameters file are provided below (these are essentially the defaults):

Search Mode: spectrast.params

Create Mode: spectrast_create.params

SpectraST File List Feature

SpectraST allows the user to list the files to be processed in a text file with extension .list. This can be useful when the number of files to be processed is very large, possibly overwhelming the UNIX command line. It is also an easy way to queue up multiple SpectraST tasks and to keep track of them. For example, if the file job.list contains the lines:

# This is a comment line ignored by SpectraST.
? -sLfoo.splib   # '?' signals the start of a new job; options for this job follow the '?'
1.mzXML
2.mzXML

? -sLbar.splib
3.mzXML
4.mzXML

Then running the command:

spectrast -sFspectrast.params job.list

is equivalent to running

spectrast -sFspectrast.params -sLfoo.splib 1.mzXML 2.mzXML

followed by

spectrast -sFspectrast.params -sLbar.splib 3.mzXML 4.mzXML

SpectraST Options

Commonly used options are shown in bold. The rest are advanced options that should rarely need to be used.

Other SpectraST Utilities

Plotspectrast

Plotspectrast is a spectrum viewer designed for SpectraST. It comes as two programs: a CGI that can be launched from a web page (e.g., from PepXMLViewer), and a stand-alone application. They are included in the TPP and no additional installation is necessary.

The most common use of Plotspectrast is for visualization of spectral matches from PepXMLViewer. When displaying SpectraST results, PepXMLViewer provides a link to invoke plotspectrast.cgi for each spectrum query. The query spectrum will be plotted as a "mirror image" of the best-matched library spectrum, enabling the user to quickly assess the quality of the match. Below the plot there is an ion table, and tables listing information about the library spectrum. The legend of the plots and ion table is as follows:

Various controls are available to the left of the plot to customize how the spectra are displayed:

The stand-alone plotspectrast application produces a static .jpg image in the same directory as the query spectrum file. It has the following usage:

plotspectrast <.splib file> <library file offset> <.mzXML file> <query scan number>

Plots the library spectrum at <library file offset> and the query spectrum of <query scan number> in the .mzXML file. The desired value of <library file offset> can be extracted from the .spidx, .pepidx or .sptxt file (BinaryFileOffset in the Comment field).

plotspectrast <.splib file> <library file offset> <.dta file>

Similar to above, except the query spectrum is in a .dta file.

plotspectrast <.splib file> <library file offset> <.none file>

Plots the library spectrum by itself. It will not actually look for the .none file, but the resulting .jpg file will be named with the same prefix as the .none file and place in the same directory.

plotspectrast <.msp file of library spectrum> <.msp, .dta, or .none file>

Similar to above, except the library spectrum is given in a .msp file.

Lib2HTML

Lib2HTML is an application that converts a SpectraST library into an HTML file for viewing. It is included in the TPP and no additional installation is necessary. In the resulting HTML file, replicates of the same peptide ion will be listed on one row, and links are provided to each replicate to view the spectrum using Plotspectrast. The usage is:

Lib2HTML <options> <full path from webserver root to .splib file>

Options include:

• -V : Verbose. Displaying more information for each entry.
• -N<num> : Specify the maximum number of replicates displayed for each unique peptide ion. Default is 10.
• -P<path> : Specify the full path from the webserver root to the plotspectrast.cgi binary.

Developer's Guide

The SpectraST source code contains detailed documentation. More tips to developers who want to modify SpectraST will be available shortly.

Where to Get Help

The SPC Tools Discussion Group: spctools-discuss.googlegroups.com

The SPC Tools Announcement Group: spctools-announce.googlegroups.com

Public spectral libraries are available for download at PeptideAtlas

External Links

• Institute for Systems Biology
• Sashimi Project on SourceForge.net
• Seattle Proteome Center Software
• spctools-discuss.googlegroups.com
• spctools-announce.googlegroups.com
• PeptideAtlas's Spectral Library Page
• GPM's X!Hunter Project
• BiblioSpec Project at University of Washington

Reference

• Keller, Andrew, et al. (2005) "A uniform proteomics MS/MS analysis platform utilizing open XML file formats". Molecular Systems Biology 1, 17. Full text

• Lam, Henry, et al. (2007). "Development and validation of a spectral library searching method for peptide identification from MS/MS". Proteomics 7 (5), 655-667. Abstract

• Craig, Robertson, et al. (2006). "Using annotated peptide mass spectrum libraries for protein identification". Journal of Proteome Research 5 (8), 1843-1849. Abstract

• Frewen, Barbara, et al. (2006). "Analysis of peptide MS/MS spectra from large-scale proteomics experiments using spectrum libraries". Analytical Chemistry 78 (16), 5678-5684. Abstract