DeepRescore

Overview

DeepRescore is an immunopeptidomics data analysis tool that leverages deep learning-derived peptide features to rescore peptide-spectrum matches (PSMs). DeepRescore takes as input MS/MS data in MGF format and identification results from a search engine. The current version supports four search engines, MS-GF+, Comet, X!Tandem, and MaxQuant.

Installation

1. Download DeepRescore:

git clone https://github.com/bzhanglab/DeepRescore

2. Install Docker (>=19.03).

3. Install Nextflow. More information can be found in the Nextflow get started page.

4. Install nvidia-docker (>=2.2.2) for AutoRT and pDeep2 by following the instruction at https://github.com/NVIDIA/nvidia-docker. Please note GPU is required to run DeepRescore.

All other tools used by DeepRescore have been dockerized and will be automatically installed when DeepRescore is run in the first time on a computer. DeepRescore has been tested on Linux.

Usage

○ → nextflow run DeepRescore.nf --help
N E X T F L O W  ~  version 19.10.0
Launching `deeprescore.nf` [special_hamilton] - revision: 2817bc64da
=========================================
DeepRescore => Rescore PSMs
=========================================
Usage:
nextflow run DeepRescore.nf
Arguments:
  --id_file              Identification result.
  --ms_file              MS/MS data in MGF format. If the search engine is MaxQuant, this parameter is not useful.
  --se                   The name of search engine, msgf:MS-GF+, xtandem:X!Tandem, comet:Comet or maxquant:MaxQuant.
                         Default is "msgf" (MS-GF+).
  --ms_instrument        The MS instrument used to generate the MS/MS data. 
                         This is used by pDeep2 for MS/MS spectrum prediction. Default is "Lumos".
  --ms_energy            The energy used in MS/MS data generation. 
                         This is used by pDeep2 for MS/MS spectrum prediction. Default is 0.34.
  --out_dir              Output folder, default is "./output"
  --prefix               The prefix of output file(s).
  --decoy_prefix         The prefix of decoy proteins. Default is "XXX_".
  --cpu                  The number of CPUs
  --mem                  The memory for processing the data, default is 8. The unit is G.
  --help                 Print help message

Input

In general, the main inputs to run DeepRescore are identification result from one of the four search engines (MS-GF+, X!Tandem, Comet and MaxQuant) and the MS/MS data used for searching. If the identification software is MaxQuant, then the MS/MS data is not needed because MS/MS data is included in MaxQuant search result ( folder combined, mqpar.xml is also required to be present in the combined folder). Below is the table showing the detailed search result format and MS/MS data format supported for each search engine. Using MS-GF+, X!Tandem or Comet, raw MS/MS data must be converted to MGF format using ProteoWizard. Multiple MGF files (different fractions) from the sample or same TMT/iTRAQ experiment should be combined into one MGF file. Only oxidation of M is supported as variable modification. Please note if DeepRescore is used to rescore MaxQuant result, the FDR cutoff should be set as 100% when performing the MaxQuant search, otherwise target PSMs may be filtered by MaxQuant's FDR calculation before rescoring using DeepRescore.

For MGF file conversion, we recommend to use the following command line:

msconvert --filter "peakPicking true 1-2" --mgf *.raw

Search engine	Identification format	MS/MS data format
Comet	.pepxml	MGF
MS-GF+	.mzid	MGF
X!Tandem	.xml	MGF
MaxQuant	/combined/	-

Below is an example:

nextflow run DeepRescore.nf --id_file ./example_data/A1101.pep.xml \
	--ms_file ./example_data/A1101.mgf \
	--se comet \
	--ms_instrument Lumos \
	--ms_energy 0.34 \
	--out_dir out \
	--prefix d2 \
	--decoy_prefix XXX_ \
	--cpu 4 \
	--mem 8

It took about one and half hour to run the example on a Linux server (12 threads, 64 RAM, GPU: TITAN Xp). The example data can be downloaded through this link: test_data.

Output

The final output data can be found in this folder out_dir/DeepRescore_results. Here out_dir is the output directory specified through parameter --out_dir. There are two files in this folder: *_psm_final.tsv and *_pep_final.tsv. The first one is the result controled FDR at PSM level and the second one is the result controled FDR at peptide level. Below is an example of *_psm_final.tsv. The format of *_pep_final.tsv is the same with *_psm_final.tsv. Users can filter the result based on the column q-value (for example, q-value <= 0.01). The result files (*_psm_final.tsv or *_pep_final.tsv + the MS/MS data in MGF format) can be imported into PDV for visualization.

spectrum_title	Percolator_score	q_value	modification	Mod_Sequence	Label	RT	Mass	Abs_Mass_Error	Ln_Total_Intensity	Match_Ions_Intensity	Rel_Match_Ions_Intensity	Max_Match_Ion_Intensity	Score	Pep	Delta_Score	charge	peptide	Proteins	Delta_RT	SA	mz
YE_20180517_SK_HLA_A1101_3Ips_a50mio_R1_02.25098.25098.2	1.43274	5.43478e-05	Carbamidomethyl of C@23[0.0];	QVADEGDALVAGGVSQTPSYLSCK	1	59.187	2451.16256793088	0	14.0435918544821	11.9698824268126	0.125718571751381	24026.38671875	339.58	4.3576e-114	283.38	2	QVADEGDALVAGGVSQTPSYLSCK	uc003kfu.4	0.130932000000001	0.775038384865853	1226.58908396544
YE_20180517_SK_HLA_A1101_3IPs_a50mio_R1_01.21936.21936.3	1.36464	5.43478e-05	-	PLFVNVNDQTNEGIMHESK	1	52.926	2171.03328724	0	16.183782828437	15.7096998670651	0.622455610654202	520227.46875	268.55	1.5772e-35	235.24	3	PLFVNVNDQTNEGIMHESK	uc010fur.3;uc002vee.4	0.23695	0.669018716842519	724.685562413335
YE_20180517_SK_HLA_A1101_3Ips_a50mio_R2_01.21952.21952.3	1.34864	5.43478e-05	-	PLFVNVNDQTNEGIMHESK	1	52.495	2171.03151690128	0	14.6959014960537	14.1948714045837	0.605906199334659	119562.108398438	284.15	1.4821e-46	248.85	3	PLFVNVNDQTNEGIMHESK	uc010fur.3;uc002vee.4	0.855877999999997	0.677688435645182	724.684972300427
AC20171011_Broad_HLA_A1101_R1_Rep01.3055.3055.3	1.32707	5.43478e-05	-	RTLDAKMPRK	1	11.999	1214.69196592591	0	15.3894885906454	14.7015468520804	0.502609506923564	815314.75	201.84	0.0048553	201.84	3	RTLDAKMPRK	uc003lvo.4;uc021ygh.2	0.168488	0.884456468503026	405.905121975303
YE_20180517_SK_HLA_A1101_3IPs_a50mio_R1_01.19078.19078.2	1.29334	5.43478e-05	-	GILAADESVGTMGNR	1	46.712	1489.71904591213	0	15.0714146779268	14.7645814265636	0.735773279870115	423094.538085938	305.7	1.1819e-36	222.7	2	GILAADESVGTMGNR	uc004bbk.2	0.180505999999994	0.862802817911802	745.867322956064

How to cite:

Kai Li, Antrix Jain, Anna Malovannaya, Bo Wen, Bing Zhang (2020), DeepRescore: Leveraging Deep Learning to Improve Peptide Identification in Immunopeptidomics. Proteomics. doi:10.1002/pmic.201900334