structural-signatures-2.0

Leaner, faster, better,

Structural Signatures Generator! (now with more emojis 🎉)

To start, untar the tar.gz file in the database folder and run structural-signatures-2.0.sh.

Next export the homed variable to be the installation directory of structural signatures, add it to your .bashrc.

If you are in the installation directory and want it automatically done, you can use the following command on linux systems (Ubuntu, WSL), not OSX.

$ pwd | sed -E "s/(.+)/export homed='\1'/gi" >> ~/.bashrc

Otherwise just export the installation directory, using the following command, changing the /path/to/structural-signatures-2.0/ to your installation path:

export homed='/path/to/structural-signatures-2.0/' >> ~/.bashrc

Dependancies

R >=3.0

Rscript

Perl >=5.22.8

Perl DBI

Perl Parrallel Fork Manager

sqlite >3.0

Running structural-signatures-2.0.sh

Running structural-signatures-2.0.sh without any arguments will give the following:

To run, specify the -i, -t, n and -o arguments. A quick example can be run using the following code (provided that all the dependancies are met):

./structural-signatures-2.0.sh -i apoptosis.gene.list.head -t both -n gn -o test

This will generate the following output if successful:

And will generate the following output: (note: outputs will change given how you specify the -t option )

Where:

1. *-enrichment.csv files are the structural signatures output
2. *.info.csv contain information about what structures were found for which proteins (ipr.info.csv has domain information and scop.structure.info has SCOPe structure information)
3. *.uid.converted contains the uniprot identifiers for the input gene list

Adding custom backgrounds to compare signatures against.

Structural signatures generates pvalues and fold changes by comparing the counted structures against the structure counts from the human proteome defined by SwissProt (~20,000 proteins)

You may want to compute structural enrichment against a background that is a subset of human proteome from SwissProt.

To do this first you can run structural-signatures-2.0.sh -t both on the genelist you want to generate the background for.

Next for each of the structral signature files (*-enrichments.csv) run the following command to get the structure counts:

cut -f1,2 -d"," filename-structure-enrichment.csv > name.background.[ipr/scop].[domain/family/superfam/fold].csv

Where:

      filename

      is the output file name from structural signatures.

      structure

      is the output structure type: domain, fold, family, superfam or fold

      name

      is the name of the background you give

      [ipr/scop]

      if structure type is domain use ipr

      if fold, family, superfamily or fold use scop

      [domain/family/superfam/fold]

      assign for the specific input structure type you are working with

Next make a directory here: bin/files/backgrounds/**name** where name is the name of the backgrounds you assigned earlier (everything must have the same name). Move all the files to that directory.

To use the custom background just specify the -r option to structural-signatures-2.0.sh and input the name of the database. Structural-signatures will verify if the counts are in the correct format.

For examples of how you should format your custom background counts look at the bin/files/backgrounds/human_proteome/ directory files.

To Do:

• Make an installation script
• Backport structural-signatures-1.0 features(disordered regions, 2D protein feature enrichment, etc)
• Multiprocessing of protein structure extraction (currently runs one at a time)
• Dockerize
• Add various error tests