Peptide selection pipeline

A comprehensive pipeline for selecting peptides for targeted mass spectrometry, integrating multiple prediction tools and filters.

Overview

This pipeline integrates several tools to facilitate the selection of optimal peptides for targeted mass spectrometry experiments:

• In silico digestion
• Basic property analysis (peptide length and methionine content)
• Uniqueness evaluation with Unipept
• Detectability prediction with DeepMSPeptide and DeepDetect
• Charge state prediction with CPred
• MS2 spectra prediction with MS²PIP

Environment setup

This pipeline uses multiple conda environments to manage tool dependencies. All environment files are provided in the environments/ directory.

Creating environments

# Create and activate environments from YAML files
conda env create -f environments/digest_env.yml
conda env create -f environments/deepms_env.yml
conda env create -f environments/deepdetect_env.yml
conda env create -f environments/cpred_optimized_env.yml
conda env create -f environments/ms2pip_env.yml

Environment descriptions

Environment	Purpose	Key Dependencies
digest_env	Basic operations and file conversions	Python, pandas, numpy
deepms_env	DeepMSPeptide predictions	Python 3.6, TensorFlow 1.13.1
deepdetect_env	DeepDetect predictions	Python 3.6, TensorFlow 1.10.0
cpred_optimized_env	CPred charge state predictions	Python 3.9, TensorFlow 2.15.1
ms2pip_env	MS²PIP spectra predictions	Python 3.9, ms2pip 4.1.0

Installing the Unipept CLI

This pipeline relies on the Unipept CLI for peptide uniqueness analysis. The CLI must be installed separately. Follow the installation instructions on the official repository: https://github.com/unipept/unipept-cli

Before running

When using the pipeline multiple times, it's recommended to remove all input and output files from previous runs to avoid issues with file overwriting.

Pipeline steps

1. In silico digest

Skip this step if starting from a list of peptides of interest

This step performs in silico digestion of proteins to generate peptides.

# Activate environment
conda activate digest_env

# Execute for a single protein
python in_silico_digest.py <UniProtID> [missed_cleavages]

# OR for multiple proteins in a FASTA file
python in_silico_digest_fasta.py file.fasta [missed_cleavages]

Output: peptides.csv

2. Basic properties analysis

This step calculates peptide length and methionine content.

If step 1 was omitted, first create a peptides.csv file containing:

• At minimum, a column named ‘peptide’ with peptide sequences
• Optionally, a column nmaed ‘Protein_ID’ with corresponding parent sequence identifiers

Example peptides.csv

Protein_ID,peptide
P48632,ETPQSITVVTR
Q9HWL3,LNFNNLFDK
...,...

# Activate environment
conda activate digest_env

# Execute
python basic_properties.py

Output: peptides.csv with added columns ‘length’ and ‘M_count’

3. Unipept peptide uniqueness analysis

Can be skipped if uniqueness is not a concern

3.1 Create txt file

# Activate environment
conda activate digest_env

# Execute
python peptides2txt.py peptides.csv

# Deactivate environment
conda deactivate

Output: peptides.txt containing all unique peptides from peptides.csv

3.2 Run Unipept

# Execute Unipept CLI (may take some time)
unipept pept2prot --output peptides_unipept.txt --select peptide,uniprot_id,protein_name,taxon_id --input peptides.txt

Output: peptides_unipept.txt with Unipept results

3.2 Process Unipept results

# Activate environment
conda activate digest_env

# Execute (adjust parameters as needed)
python retrieve_unipept.py --taxon_id 208964 --protein_id P48632 --keywords "Ferripyoverdine receptor,Ferripyoverdine,FpvA" --unipept_file peptides_unipept.txt --peptides_file peptides.csv

Details:

• Keywords are used to identify matches corresponding to the protein of interest
• The script counts UniProt matches that correspond to the given taxon ID but not to the specified protein_ID or keywords
• Adds columns 'unique' (boolean) and 'problematic_matches' in case of non-unique to peptides.csv
• Peptides not found in Unipept results will have missing values

4. DeepMSPeptide prediction

4.1 Create txt file

Skip if step 3 was completed

# Activate environment
conda activate digest_env

# Execute
python peptides2txt.py peptides.csv

# Deactivate environment
conda deactivate

Output: peptides.txt with unique sequences

4.2 Run DeepMSPeptide

# Copy text file to DeepMSPeptide directory
cp peptides.txt ../deepmspeptide/DeepMSPeptide/DeepMSPeptide

# Activate environment
conda activate deepms_env

# Navigate to DeepMSPeptide directory
cd ../deepmspeptide/DeepMSPeptide/DeepMSPeptide

# Execute
python DeepMSPeptide.py peptides.txt

# Rename results file
mv peptides_Predictions.txt peptides_DMP.txt

# Copy results back to pipeline directory
cp peptides_DMP.txt ../../../peptide_selection_pipeline

Output: peptides_DMP.txt with DeepMSPeptide predictions

4.3 Process DeepMSPeptide results

# Activate environment
conda activate digest_env

# Navigate to pipeline directory
cd peptide_selection_pipeline

# Execute
python retrieve_dmp_preds.py

Output: Updates peptides.csv with added column 'DMP_prob'

5. DeepDetect

5.1 Run DeepDetect

# Activate environment
conda activate deepdetect_env

# Navigate to DeepDetect directory
cd deepdetect/SourceCode/deepdetect_pred

# Download FASTA if needed
wget https://www.uniprot.org/uniprotkb/P48632.fasta

# Execute (adjust parameters as needed)
python main.py --input=P48632.fasta --output=peptides_DD.txt --regular='>(\S+)' --protease='Trypsin' --missed_cleavages=2 --min_len=8 --max_len=25

# Copy results to pipeline directory
cp peptides_DD.txt ../../../peptide_selection_pipeline

Output: peptides_DD.txt with DeepDetect results

5.2 Process DeepDetect results

# Activate environment
conda activate digest_env

# Navigate to pipeline directory
cd peptide_selection_pipeline

# Execute
python retrieve_dd_preds.py

Output: Updates peptides.csv with added column 'DD_prob'

6. CPred charge state prediction

6.1 Prepare CPred input

# Activate environment
conda activate digest_env

# Navigate to pipeline directory
cd peptide_selection_pipeline

# Execute
python peptides2cpredinput.py

Output: cpred_input.csv formatted for CPred

6.2 Run CPred

# Activate environment
conda activate cpred_optimized_env

# Navigate to CPred directory
cd cpred_optimized/CPred/CPred

# Execute Feature Engineering
python CPred_main.py FeatureEngineering -i ../../../peptide_selection_pipeline/cpred_input.csv -o output_FE

# Execute prediction
python CPred_main.py prediction -i ../output_FE.parquet -m Data/Models/CPred_model_v1.h5 -o predictions

# Rename predictions file
cd predictions
mv Model_predictions.csv peptides_CPred.csv

# Copy to pipeline directory
cp peptides_CPred.csv ../../../../peptide_selection_pipeline

Output: peptides_CPred.csv with CPred results

6.3 Process CPred results

# Activate environment
conda activate digest_env

# Navigate to pipeline directory
cd peptide_selection_pipeline

# Execute
python retrieve_cpred_preds.py

Output: peptides.csv with added columns 'CS' (charge state) and 'CS_prob' (corresponding probability)

7. MS²PIP prediction

7.1 Prepare MS²PIP input

# Activate environment
conda activate digest_env

# Navigate to pipeline directory
cd peptide_selection_pipeline

# Execute
python peptides2ms2pipinput.py

Output: ms2pip_input.tsv formatted for MS²PIP

7.2 Run MS²PIP

# Activate environment
conda activate ms2pip_env

# Navigate to pipeline directory
cd peptide_selection_pipeline

# Execute
ms2pip predict-batch ms2pip_input.tsv -f spectronaut -o peptides_ms2pip --model HCD

Output: peptides_ms2pip.spectronaut.tsv with MS²PIP predictions

7.3 Process MS²PIP results

# Activate environment
conda activate digest_env

# Navigate to pipeline directory
cd peptide_selection_pipeline

# Execute
python retrieve_ms2pip_preds.py

Output: peptides.csv with added columns 'valid_peaks' and 'average_peak_intensity'

8. Final peptide selection

# Activate environment
conda activate digest_env

# Navigate to pipeline directory
cd peptide_selection_pipeline

# Execute
python peptide_selection.py

Output: final_selection.csv containing peptides that meet all filter criteria