Identifying proteins from mass spec data
This tutorial describes how to identify a list of proteins from tandem mass spectrometry data.
Analyses of this type are a fundamental part of most proteomics studies. The basic idea is to match tandem ms spectra obtained from a sample with equivalent theoretical spectra from a reference protein database. The process is referred to as "protein database search" or even "protein sequencing", although amino acid sequences are not obtained de novo with this method.
The data used in this tutorial were obtained from a single run on an Orbitrap Elite mass spectrometer. The sample itself corresponds to a purified organelle from Mouse cells.
The aim of the tutorial will be to create a list of all proteins that can be confidently said to be present in the sample, and then to use this list to guess the identity of the "mystery" organelle.
This tutorial uses free software including;
- The X!Tandem search engine
- The Trans Proteomic Pipeline (TPP) for post-search validation
- The Protk tool suite for various conversion tasks and to make working with X!Tandem and the TPP easier
- The Galaxy platform to bring all these tools together
Login to Galaxy
Open a browser and go to a Galaxy server.
- You can use a galaxy server of your own or
- Galaxy Tute at genome.edu.au
Use a supported browser. Firefox/Safari/Chrome all work well
If you use your own galaxy server you will need to make sure you have the protk proteomics tools installed.
Register as a new user if you don’t already have an account on that particular server
Import mass spec data
- Create a new history in Galaxy and name it "Organelle Tutorial"
Download datasets using the Galaxy uploader tool.
Open this tool by clicking the button as shown below
After opening the tool select Paste/Fetch data and paste the following URL into the box that appears. Then click Start to initiate the download.
After the download is finished you should have a single item in your history.
Rename the history item by clicking the pencil icon beside it to "Edit attributes".
This should bring up a dialog box where you can edit the name.
Change the name by removing everything up to the last forward slash "/"
Your item should then be named OrganelleSample.mzML
Dont forget to click "Save"
Basic properties of the data
Mass spectrometry data comes in many different formats and the first step in a proteomics analysis often involves data conversion or pre-processing. You can read more about mass spectrometry data formats here
Try clicking the title bar on the data in your galaxy history. This will toggle display of some additional information about the data.
MS versus MS/MS:
A key feature of Tandem mass spectrometry is the acquisition of mass spectra (spectra) that measure the masses of precursor ions (whole peptides) as well as spectra that measure the fragment masses of a single selected peptide. These two types of measurements are called MS and MS/MS spectra respectively. The following schematic shows how an MS/MS scan results from the fragmentation of a selected product ion. Often multiple MS/MS spectra are obtained for each MS scan, selecting different precursor masses each time so that as many peptides as possible can be analyzed.
Number of spectra:
Click the eye icon on the history item to view the mzML file as text. The file is almost impossible to read by hand but with some text searching we will be able to deduce the number of MS and MS/MS spectra in the file.
Now try searching for the text "MS1 spectrum" in the page using your web browser's search function. Looking closely you should see that this text appears once for every MS1 spectrum in the file (plus it occurs one extra time at the top of the file as part of the file description). The file is large though and the browser can only see the first megabyte of it.
Now search for the text "spectrumList count". It should bring you to a line in the file that says spectrumList count="24941". There are a total of 24941 spectra in the entire file including both MS and MS/MS spectra.
To answer this question you will need to use the Select tool from the Filter and Sort submenu to select lines matching the text "MS1 spectrum" in the whole file. Then use the Line/Word/Character count tool from the Text Manipulation submenu to count the number of lines returned by running the Select tool.
The text "MS1 spectrum" also appears at the top of the file as part of its description so you will need to subtract 1 from your answer
In the previous exercise we used two galaxy tools in succession to find out the number of items in a file that matched some text. Future exercises use the same technique so you might find it useful to create a tiny workflow to automate this proceedure. See the instructions here
Prior to the development of tandem mass spectrometry, peptides and proteins were detected purely by matching MS peaks against the masses of whole peptides via Peptide Mass Fingerprinting. This has largely been superceded by tandem mass spectrometry which gains much greater specificity by using the MS/MS spectra. In this tutorial only the MS/MS spectra will be used.
Use the fact that the file contains a total of 24941 spectra with your answer to the previous question about MS spectra.
Alternate data formats
Another format you are likely to encounter for tandem mass spectrometry data is Mascot Generic Format or mgf. Mascot Generic Format (mgf) is the data file format preferred by the Mascot search engine. It is a text based format is much easier to read by hand than the mzML file. Each spectrum appears between "BEGIN IONS" and "END IONS" statements and simply consists of (mz,intensity) pairs. Additional summary information about the precursor (whole peptide) ion such as its mass, retention time and charge are included.
Download the Organelle Sample data in mgf format
Use the Paste/Fetch data tool again and paste the following URL into the box that appears. Then click Start to initiate the download.
Inspect the data manually by viewing it in Galaxy. Try to get a feel for the way data is organised within the file.
Obtain a Search Database
Setting up a search database is a critical step. For this tutorial we have created a database for you, but if you need to create a database for your own data you'll need to consider the following key issues;
- Database size
- Whether to include decoys
- What types of variants to include if any
- How to format your database identifiers
More details are provided here.
Download a database of Mouse proteins in fasta format
Use the Paste/Fetch data tool again and paste the following URL into the box that appears. Then click Start to initiate the download.
Inspect the first few items in the database in Galaxy. The file is in Fasta format which means that each entry has a single description line that starts with a ">" followed by a unique identifier and then some general descriptive information. The actual sequence of amino acids is given after the description line. Take note of the format of the database identifiers. They are in Uniprot format and look like this;
The database also includes decoy sequences, appended at the end. They have identifiers like this;
Decoys are easiest to search for because they all start with "decoy_". The total number of database entries can be found simply expanding the fasta file in your history (by clicking on its title).
Run a search using X!Tandem
A large number of search engines now exist for proteomics data. This exercise uses X!Tandem which is one of the fastest and most widely used. Other search engines include OMSSA, MS-GF+ and Mascot.
- Select the X!Tandem Search tool
- Enter parameters as shown in the table below (leave all others at their defaults)
- Click Execute
|Uploaded Fasta File||
|Missed Cleavages Allowed||
|Fragment Ion Tolerance||
|Precursor Ion Tolerance||
The search should run for about 5-10 minutes and will produce an output file in X!Tandem xml format. A much more useful format is pepXML so the next step in the analysis will be to run a tool to convert from tandem to pepXML.
- Select the Tandem to pepXML tool
- Select the output from the previous step as input and click Execute
While the search is running, read some background theory on how search engine's work.
Convert Results to tabular format
Although the pepXML format is useful as input to other tools it is not designed to be read or analyzed directly. Galaxy includes a tool to convert pepXML into tabular (tab separated) format, which is much easier to read. Tabular format also has the advantage that it can be downloaded and opened using many other programs including Excel and R.
- Select the pepXML to Table tool
- Select the pepXML file produced in the previous step as input and click Execute
To get the most out of tabular files it is often necessary to know the column number corresponding to columns of interest. Explore the column assignments in your tabular file by clicking on its title in your galaxy history. This will show extra details about the item, including a handy preview where column numbers are displayed.
Sort tabular outputs
Examine the tabular output file from the previous step. It contains many columns, of which the most interesting are probably the raw search engine score and the name of the protein from which each peptide has been derived. Note the presence of plenty of decoy proteins among the results. These decoys should tend to have quite poor scores compared with real hits. The raw score for X!Tandem searches is an E-value. To push these decoys to the bottom of the list we can sort the data by raw score.
- Select the Sort data tool from the Filter and Sort menu in the left pane of Galaxy
- Choose to sort on raw_score. This is column c10
- Select Ascending order for the sort direction (small E-values are good) and click Execute
Browse the resulting dataset. The top of the file should now have very few decoys.
Convert raw scores to probabilities
Raw X!Tandem scores only give a rough estimate of the reliability of each peptide to spectrum match (PSM). A better estimate can be obtained by running a tool that uses global statistical properties of the search to assign a probability to each PSM of being correct. A number of tools exist for this, and in this tutorial we use Peptide Prophet, which can work with a wide variety of different search engine scoring systems. It is extremely useful as it effectively converts disparate scores to the common scale of probability. The probabilities produced by Peptide Prophet can be used to set a threshold for acceptance. For example we could decide to accept only PSM's with a probability greater than 0.95. Note that this is not the same as the False Discovery Rate which is computed by taking (1-p) for all the accepted PSM's and dividing by the total number of accepted PSM's. A widely used alternative to Peptide Prophet is Percolator.
- Select the Peptide Prophet tool
- Select the X!Tandem output in pepXML format generated earlier as input
- Check the box that says Use decoys to pin down the negative distribution.
- Convert the resulting pepXML file to tabular using the PepXML to Table tool
Take a look at the resulting tabular file. Note that this time the peptideprophet_prob column is populated and contains numbers between 0 and 1.
Use the Filter tool from the Filter and Sort submenu. Also remember that Peptide Prophet probability is given in a column called peptideprophet_prob. The syntax for "greater than or equal to" in the Filter tool is >=.
Use this text in match with condition field of the Filter and Sort tool.
To answer the second question use the Select tool on the filtered table to select lines matching "decoy_"
Perform Protein Inference
Up to this point we have looked at peptide to spectrum matches PSMs. Each of the peptides observed will have come from a protein sequence in the fasta file that we used as a database, and this protein is recorded along with the PSM itself in all of the result tables we've viewed so far. Unfortunately, the process of inferring the existence of proteins based on these PSMs is much more complicated than that because some peptides are found in more than one protein, and of course some proteins are supported by more than one PSM.
The Protein Prophet tool can be used to run a proper protein inference analysis, and assigns probabilities to individual proteins, as well as groups of related proteins.
- Select the Protein Prophet tool
- Choose the pepXML formatted output from Peptide Prophet as input and click Execute
- Convert the resulting protXML to tabular using the protXML to Table tool.
Filter on column 6 protein_probability
If you have time, read over these notes on the Protein Prophet output. Explore your output (use the unfiltered tabular output) to find examples of different kinds of Protein groupings.
Functional enrichment analysis
This step will allow you to discover the identity of the Organelle that was used to create the sample.
We use the GOrilla gene ontology enrichment analysis tool (a web based tool) to discover GO terms that are over-represented in proteins at the top of our list compared with those that are assigned very low probabilities (at the bottom).
- Start with unfiltered tabular protein prophet results
- Use the Cut columns tool from the Text Manipulation menu to extract the third column from the filtered protein table (contains protein_name).
- Convert the "pipes" that separate parts of the protein_name into separate columns using the Convert delimiters to TAB tool in the Text manipulation submenu of Galaxy. This should result in a file with 3 columns
- Use the Cut columns tool again to cut the second column from this dataset
- Download this file to your desktop and rename it to organelle.txt
- Open the GOrilla web page in your web browser
- Select Organism as Mouse
- Upload the organelle.txt file as a ranged gene list
- Choose Component for the ontology
Ignore terms relating to exosomes
In the resulting output look to the most enriched and most specific GO terms.