DeepGrind

DeepGrind is a snakemake workflow that identifies peptides from LC/MS-MS data. It combines the well-established peptidomics identification tools Comet (http://comet-ms.sourceforge.net/) and Percolator (https://github.com/percolator/percolator) with the Deep Learning approaches offered by pDeep (https://github.com/pFindStudio/pDeep) and Prosit (https://github.com/kusterlab/prosit). These tools can predict fragmentation spectra (and, in the case of Prosit, chromatographic retention times) of arbitrary peptides which can then be used to rescore the PSMs found by Comet. The new scores subsequently allow Percolator to separate true from false discoveries with more confidence, which results in a higher number of identifications and smaller False Discovery Rates.

Installation

Requirements

In order to run DeepGrind, you need to have Conda installed. All other packages required (including Comet, Percolator, and Snakemake) will be installed by Conda on-the-fly. Furthermore, you need a Prosit instance that is able to return predicted spectra in Mascot Generic Format. The vanilla version of Prosit (as of July 2020) does not do that, but you can e.g. use my fork of Prosit (https://github.com/jmueller95/prosit). If you also want to use pDeep for MSMS prediction you'll need a server instance of pDeep, which you can obtain from https://github.com/jmueller95/pDeep.

Installation Steps

1. Clone this repository.
2. Create the Conda environment for the workflow by running conda env create -f envs/deepgrind_env.yml. This will also install Comet and Percolator.
3. Activate the environment: conda activate deepgrind_env.

Usage

The input spectra need to be provided in Mascot Generic Format (http://www.matrixscience.com/help/data_file_help.html). DeepGrind expects each spectrum to contain a local parameter IRT with the indexed (standardized) retention time. If you do not possess indexed retention times for your spectra, the workflow will simply skip the RT-based rescoring step (it is not possible to use the RTINSECONDS parameter for rescoring). The output is simply the tab-delimited PSM list that Percolator usually outputs (using the -r flag as described here).

The workflow parameters are given in the form of a configuration file. See config_template.yaml for an example and explanation on each parameter. Create your own config file, put your custom parameter values in it and save it to config.yaml.

Then, start the workflow by cding into the directory containing the Snakefile and calling snakemake --cores all. Instead of all, you can also put an integer specifying the number of processor cores you want to make available to the workflow.