PhosFake: Simulating proteoforms in bottom-up proteomics

Phosfake is a platform for the generation of an in-silico bottom-up proteomics data set with a ground truth on the level of proteoforms.

All the parameters that are used to generate the data are described in man/Parameters.qmd. The script that runs the entire pipeline is RunSims.R. Alternatively, you can use the Vignette. The simulations with multiple parameters can be set up and run. PhosFake also provides comparison of the results with the ground truth using benchmarking metrics and visual comparison between simulated data sets.

The pipeline is described in the figure PhosFakeLayout.svg and can be described as follows:

0) General functions to run the simulations: 00_BachRunFuncs.R 1) Generation of ground truth data at the proteoform leve:l 01_GenerateGroundTruth.R. 2) Digestion of the proteoforms from the ground truth: 02_Digestion.R. 3) In silico MS run 03_MSRun.R. 4) Functions for data analysis from the peptide to proteins: 04_DataAnalysis.R. 5) Statistical testing: 05_Statistics.R. 6) Benchmarking: 06_Benchmarks.R.

Installation

Install the package from GitHub with devtools: devtools::install_github("veitveit/PhosFake")

Running full batches and benchmarking

Running the Vignette allows running full batches without having to re-run the data sets which have been built with the same set of parameters. In addition, the pipeline is run hierarchically to avoid repetitive execution of identical down-stream analysis. This is done via creating hashes of the parameter configurations and writing intermediate and final results into respective tables.

Re-running the full batch with different assessment of the benchmarking metrics will avoid re-running the data set generation and analysis, and thus should be superfast.

Important remarks:

• You need to always define all parameters in the beginning of this script
• Be aware that changing downstream parameters (ground truth, digestion) can immensely increase the number of possible parameter settings
• Keep always the result files in the respective folder (resultFilePath) if you didn't change anything in the pipeline such as any of the methods in the sourced files. This will allow you to run the full batch without re-running the data set generation and analysis.
• For data set with only few quantified proteins (<100), benchmarking is discarding.