Training and Evaluating a Multiclass Classifier for Multiple Myeloma
Gregory Way and Casey Greene 2018
In the following analysis, we train a multiclass classifier on RNAseq data of patients with Multiple Myeloma. The classifier is trained to distinguish KRAS and NRAS mutations from wild-type tumors.
The analysis is presented in the following preprint:
Yu-Hsiu Tony Lin, Gregory P. Way, Benjamin G. Barwick, Margarette C. Mariano, Makeba Marcoulis, Ian D. Ferguson, Christoph Driessen, Lawrence H. Boise, Casey S. Greene, Arun P. Wiita. Integrated Phosphoproteomics and Transcriptional Classifiers Reveal Hidden RAS Signaling Dynamics in Multiple Myeloma. February 29th, 2019. biorXiv https://doi.org/10.1101/563312.
The data was provided by Arun Wiita and Tony Lin (UCSF) as part of the MMRF CoMMpass Study. The data was accessed in a Box link provided by Arun and Tony. Before performing the analysis, the appropriate data must be deposited into the appropriate folders. The data are not included in this repository.
The analysis expects the following data in the
If the data are not deposited in the
data/raw folder, the analysis will not run.
We will provide a reproducible solution by depositing data in an archived and versioned database at a later date.
We use conda (
version 4.5.2) to manage our computational environment.
All scripts must be run in this environment.
After installing conda, run the following:
# Initialize the environment conda env create --force --file environment.yml # Activate the environment conda activate multiple-myeloma-classifier
The analysis consists of a series of notebooks that are designed to be run in order.
run_analysis.sh for more details.
Notebooks are converted into python
.py files and stored in the
scipts/ folder with:
jupyter nbconvert --to=script --FilesWriter.build_directory=scripts *.ipynb