Keith Baggerly 2018-03-18

• Overview
• Brief Results
• Running the Analysis
  • Required Libraries

We want to illustrate assembly of a reproducible analysis using a dataset we care about. Our workflow closely follows that of Jenny Bryan's packages-report-EXAMPLE on GitHub.

Several years ago, Potti et al claimed to have found a way to use microarray profiles of a specific panel of cell lines (the NCI60) to predict cancer patient response to chemotherapeutics from a similar profile of the patient's tumor. Using different subsets of cell lines, they made predictions for several different drugs. We wanted to apply their method, so we asked them to send us lists of which cell lines were used to make predictions for which drugs. The method doesn't work; we describe our full analyses here.

The first dataset we received from Potti et al didn't have the cell lines labeled. We want to see if we can identify where the numbers came from and see if there were any oddities that should have raised red flags early on.

• 01_gather_data downloads the raw datasets used from the web.
• 02_describing_raw_data both summarizes background information about the datasets in question and describes what we can see by looking at the structure of the data with minimal processing.
• 03_preprocessing_data_base_r_version reorganizes the datasets into matrices of expression values and data frames of annotation using consistent names.
• 04_check_sample_matches_and_corrs establishes the mappings desired.
• 05_report_matches_to_potti_columns

Roughly, our analyses involve running the R and Rmd files in R in the order they appear.

Run R/95_make_clean.R to clear out any downstream products.

Run R/99_make_all.R to re-run the analysis from beginning to end, including generating this README.

Raw data from the web is stored in [data][data].

Reports and interim results are stored in [results][results].

These analyses were performed in RStudio 1.1.414 using R version 3.4.3 (2017-11-30), and use (in alphabetical order):

• downloader 0.4
• GEOquery 2.42.0
• here 0.1
• lattice 0.20.35
• magrittr 1.5
• readr 1.1.1
• rmarkdown 1.8
• tidyr 0.8.0