Skip to content

Simple data analysis

Sarah Haynes edited this page Jul 7, 2021 · 7 revisions

This tutorial will show you how to use Philosopher for a complete proteomics data analysis, starting from a raw LC-MS file and ending with protein reports. Philosopher can also be used on Windows, though the commands in this tutorial are formatted for GNU/Linux

What are the basic steps?

The commands in this tutorial should be executed in a particular order. Consider this the "default" order in which to perform an analysis:

  1. Create a workspace
  2. Download a database
  3. Search with MSFragger
  4. PeptideProphet
  5. ProteinProphet
  6. Filter
  7. Quantify (optional)
  8. Report

Before we start

For this tutorial, we will use a publicly-available LC-MS data file from a human cell line sample described in this publication. Download the file 20190202_QExHFX2_RSLC8_PST_HeLa_10ng_1ulLoop_muPAC_1hr_15k_7.raw from the dataset FTP location (the full listing is here).

You can choose to use the .raw spectral format, or you can convert it to the mzML format (which is needed for quantification). A tutorial on raw file conversion can be found here.

For additional help on any of the Philosopher commands, you can use the --help flag (e.g. philosopher workspace --help), which will provide a description of all available flags for each command.

1. Create a workspace

(Note: use the full path to the Philosopher binary file in place of philosopher in the following steps.) Place the 20190202_QExHFX2_RSLC8_PST_HeLa_10ng_1ulLoop_muPAC_1hr_15k_7.raw in a new folder, which we will call the 'workspace'. We will create the workspace with the Philosopher workspace command, which will enable the program to store processed data in a special binary format for quick access. (If you have already initialized a Philosopher workspace in this same location, run philosopher workspace --clean to prepare it for a new analysis.)

Inside your workspace folder, open a new terminal window and run this command: philosopher workspace --init

From now on, all steps should be executed inside this same directory.

2. Download a protein database

For the first step we will download and format a database file using the database command, but first we need to find the Proteome ID (PID) for our organism. Searching the UniProt proteome list, we can see that the Homo sapiens proteome ID is UP000005640, so let's prepare the database file with the following command:

philosopher database --id UP000005640 --contam --reviewed

Philosopher will retrieve all reviewed protein sequences from this proteome, add common contaminants, and generate decoy sequences labeled with the tag rev_.

You should see that a new file was created in the workspace. (Note: Databases must be processed within the current Philosopher workspace for the analysis to finish properly. If you are not downloading the database with Philosopher in the current workspace, see building a custom database from existing sequences or annotating an existing database if you already have one formatted from a previous Philosopher analysis.)

3. Perform a database search with MSFragger

(Note: use the full path to the MSFragger.jar file in place of MSFragger.jar in the following steps.)

Run java -jar MSFragger.jar --config to print three MSFragger parameter files (closed, nonspecific, and open).

In the closed_fragger.params file (you can remove the other two .params files if desired), update the database_name parameter to the name of the database file we downloaded in the previous step (e.g. 2019-11-04-td-rev-UP000005640.fas). You can also change the calibrate_mass parameter near the top of the file from 2 to 0 to speed up the search even more.

Launch the search by running: java -Xmx32g -jar MSFragger.jar closed_fragger.params 20190202_QExHFX2_RSLC8_PST_HeLa_10ng_1ulLoop_muPAC_1hr_15k_7.raw. (Adjust the -Xmx flag to the appropriate amount of RAM for your computer.)

The search should be done in a few minutes or less. The search hits are now stored in a file called 20190202_QExHFX2_RSLC8_PST_HeLa_10ng_1ulLoop_muPAC_1hr_15k_7.pepXML.

4. PeptideProphet

The next step is to validate the peptide hits with PeptideProphet:

philosopher peptideprophet --database 2019-11-04-td-rev-UP000005640.fas --decoy rev_ --ppm --accmass --expectscore --decoyprobs --nonparam 20190202_QExHFX2_RSLC8_PST_HeLa_10ng_1ulLoop_muPAC_1hr_15k_7.pepXML

Note that tag identifying decoy sequences ("rev_" in this case) has been specified with the --decoy flag.

This will generate a new file called interact-20190202_QExHFX2_RSLC8_PST_HeLa_10ng_1ulLoop_muPAC_1hr_15k_7.pep.xml.

5. ProteinProphet

Next, perform protein inference and generate a protXML file:

philosopher proteinprophet interact-20190202_QExHFX2_RSLC8_PST_HeLa_10ng_1ulLoop_muPAC_1hr_15k_7.pep.xml

6. Filter and estimate FDR

Now we have all necessary files to filter our data using the FDR approach:

philosopher filter --sequential --razor --picked --tag rev_ --pepxml interact-20190202_QExHFX2_RSLC8_PST_HeLa_10ng_1ulLoop_muPAC_1hr_15k_7.pep.xml --protxml interact.prot.xml

Note that the "rev_" tag is specified again at this step. The filter algorithm can be applied in many different ways, use the --help flag and choose the best method to analyze your data. Scoring results will be shown in the console, and all processed data will be stored in your workspace for further analysis.

7. Optional: perform quantification (requires mzML format)

Filtered search results can be quantified using MS1 peak intensities at this point.

philosopher freequant --dir .

(Isobaric label-based quantification can also be performed at this point, see the labelquant command for more information.)

8. Report the results

Now we can inspect the results by printing the PSM, peptide, and protein reports: philosopher report


As an optional last step, backup your data in case you wish to print the reports again later.

philosopher workspace --backup

Concluding remarks

We've demonstrated how to run a complete proteomics analysis with TMT quantification using Philosopher. By providing easy access to advanced analysis software and custom processing algorithms, protein reports can be obtained from LC-MS files in just a few minutes.