The proteomics analysis pipeline consists of a suite of tools that support the design and analysis of mass-spec based proteomics and phosphoproteomic measurements. In addition to custom scripts, the suite also incorporates pre-existing tools such as GSEA, VIPER and Hotnet in order to make network level inference of differential pathway activity.
To install the Python dependencies of this package, run
pip install -r requirements.txtIn the first example below, a TMT-labeled mass spec dataset is merged with user-provided metadata. In addition, all outdated identifiers (Uniprot Id, Gene name) are identified and updated. The metadata file should be a csv file containing a column 'TMT_label' that correspond with the columns (tmt tags) in the dataset, and a second column caled 'Sample' which is a user provided name for the sample. The metadata file may optionally also contain additional columns that describe other categorical features of the samples.
import pandas as pd
from msda import preprocessing
meta_df = pd.read_csv('metadata_file.csv')
df = preprocessing.pd_import(data_file, meta_df)If the experiment samples are split across two or more 10-plex experiments and a bridge sample is present in each 10-plex, then the datasets can be normalized based on the bridge sample and merged into a single dataset before subsequent analysis
df = preprocssing.pd_import([data_file1, data_file2], meta_df)The next example shows how principal component analysis or hierarchichal clustering of the dataset can be done.
from msda import clustering
clustering.pca(df, meta_df, num_components=2, label=None, plot_prefix='pca_')
clustering.hierarchical_clustering(df, output_file='figure.png')If the metadata file contains additional columns with categorical information about the samples, then the PCA plots can also depict the categories using different colors. In order to do so, simply pass the name of the column in the metadata file to the argument 'label'.
clustering.pca(df, meta_df, num_components=2, label='tumor_subtype', plot_prefix='pca_')Mass-spec based proteomics measurements provide relative intensity measurements for each protein i.e intensities for a given protein can be compared between samples, but intensities between proteins cannot be comapred. Intensity based absolute quantifcation (iBAQ) is a normaliation methods that enables comparison across proteins. The example below show how to map the dataset onto iBAQ space.
from msda import ups2_calibration as uc
df_ibaq = uc.compute_ibaq_dataset(df, organism='human', meta_df)If UPS2-based calibration has been performed for one of the samples in the dataset (for eg:'Bridge'), then the intensity measurements can be mapped to concentrations (or copies/cell).
df_ups2 = pd.read_csv(ups2_dataset_file)
df_abundance = uc.compute_abundance(df, df_ups2, sample='Bridge')Below is an example of how gene set enrichment analysis can be performed using the GSEA tool from Broad Institiute. One can either use libraries from GSEA, Enrichr or custom libraries that are saved as '.gmt' files. GSEA requires an input '.rnk' file that contains genes and the associated fold-change values for any sample_pairs of interest. First, the rnk file is generated by providing the full dataset and the sample to be comapred. The df_rnk is to be saved at a specific location. The path to the rnk file, the desired library to be used and an output folder to save the results are to be specified as arguments to the get_gsea_enrichment function. The results can be plotted using the plot_nes function. The argument filter can be set to True if only the significanttly enriched scores are to be plotted (p < 0.05 and FDR < 0.25)
from msda import gsea_tool
df_rnk = gsea_tool.make_rnkfile(df, sample1, sample2)
df_rnk.to_csv(rnk_filename, index=False, sep='\t')
library = 'path to .gmt file'
df_nes = gsea_tool.get_gsea_enrichment(rnk_filename, library, output_folder)
gsea_tool.plot_nes(df_nes, filter=True, outfile='enrichment.png')VIPER is a tool that is developed by the Calefano Lab. It is used to infer Transcription factor activity based on differential activity of downstream targets. We wrote a wrapper that can take in RNA-seq/proteomics or phosphoproteomics datasets to infer TF or kinase activites respectively. VIPER can be run either in single sample ('ss') or multi-sample mode ('ms'). For multi-sample mode, the sample categories to be compared (test and ref) have to be provided in the function argument.
from msda import run_viper as rv
# single sample mode
df_nes = rv.run_viper(df, meta_df, type='ss', outfile='path to output file')
# multi sample mode
df_nes = rv.run_viper(df, meta_df, type='ms', label='tumor_subtype',
test='subtype A', ref='subtype B', outfile='path to output file')