Repository associated with the manuscript "An atlas of plasma extracellular vesicle proteins distinguishing cancer and doxorubicin sensitivity"

This reposotory includes the code and key summary tables related to the manuscript by Van Tine et al. focused on plasma extracellular vesicles in patient with sarcom anad other cancers. Note that several of the notebooks perform best on a computer cluster with sufficient memory.

All proteomic data for the Sarcoma cohort is deposited in UCSD’s Massive with ID MSV000092188. Reviewers can use the username MSV000092188_reviewer and password VanTine_001_0616 to anonymously access the data. The FTP download link is ftp://[email protected]. Accessing data for the Hoshino et al. dataset is detailed in their Cell manuscript.

File details

• SampleMetadata.xlsx: Sample metadata for Sarcoma cohort
• expt_smty.xlsx: Annotation file needed for ProteoQ processing of Sarcoma cohort
• ProteinQuantitationWithSampleMetadata.xlsx: Final protein-level quantitation of the Sarcoma cohort with sample metadata added
• uniprotAnnotationsForFigure2.xlxs: Uniprot annotations for Figure 2
• dfHarmonizer.txt: file to harmonize sarcoma and Hoshino datasets
• dfMerged_description.csv: file to harmonize sarcoma and Hoshino datasets
• _selectedFeaturesRFECV_target-dfHarmonizedCancer_nSampleFilter60_trainFrac-0.75_wholeDataset.json: selected features for KNN pan-cancer classification
• studyKNeighborsClassifier_dfHarmonizedCancer_nSampleFilter60_trainFrac-0.75_wholeDataset.db: KNN pan-cancer classifier as a db file
• KNeighborsClassifierTuned_target-dfHarmonizedCancer_nSampleFilter60_trainFrac-0.75_wholeDataset.pkl: KNN pan-cancer classifier as a pkl file

Jupyter notebook details

• for all notebooks, use Python v3.10, R v4.2.1, jupyterlab v4.1.1, pandas v2.0.0, numpy v1.23.4, scikit-learn v1.3.2, joblib v1.3.2, statsmodels v0.14.0, scipy v1.10.1, pingoin v0.5.3, openpyxl v3.0.10, optuna v3.5.0, xgboost v2.0.3, catboost v1.2.3, lightgbm v4.3.0, rpy2 v3.5.9, and pip 24.0.
• Notebook VanTineEtAl_ProteoQ.ipynb: used for ProteoQ processing of the Sarcoma cohort and to generate some panels of Figure 2.
• Notebook VanTineEtAl_Fig1_LinePlot.ipynb: used to generate the line plot in Figure 1.
• Notebook VanTineEtAl_Fig2_VolcanoPlot.ipynb: used to generate the volcano plot (sarcoma vs. healthy) in Figure 2.
• Notebook VanTineEtAl_Fig2_Catplot.ipynb: used to generate the catplot in Figure 2.
• Notebook VanTineEtAl_Fig3_panCancerClassification.ipynb: used for the pan-cancer classification in Figure 3.
• Notebook VanTineEtAl_Figs3&4_AllButClassification.ipynb: used for the non-classification plots in Figures 3 & 4.
• Notebook VanTineEtAl_Harmonizer.ipynb: used to harmonize the sarcoma and Hoshino datasets. Uses R v4.2.1.
• Notebook VanTineEtAl_Fig5_VolcanoPlot_resistantPatients_PreVsPost.ipynb.ipynb: used to generate the volcano plot (pre vs. post treatment in drug resistant sarcoma patients) in Figure 5.
• Notebook VanTineEtAl_Fig5_VolcanoPlot_sensitivePatients_PreVsPost.ipynb.ipynb: used to generate the volcano plot (pre vs. post treatment in drug sensitive sarcoma patients) in Figure 5.
• Notebook VanTineEtAl_Fig5_VolcanoPlot_SvR.ipynb: used to generate the sensitive:resistant volcano plot in Fig 5.
• Notebook VanTineEtAl_Fig5_LDA.ipynb: used to generate the LDA-related plots in Figure 5.
• Notebook VanTineEtAl_Fig5_Classification.ipynb: used for the sarcoma vs healthy and doxorubicin sensitivity classification in Figure 5.
• Notebook VanTineEtAl_Fig5_Catplot.ipynb: used to generate the catplots in Figure 5.