Add expression data, sample metadata, and recount2

[jaclyn-taroni]

This PR is the initial data download and set up, including git LFS, for this repository. This project utilizes a data repository (greenelab/rheum-plier-data). I obtain expression data and sample metadata from greenelab/rheum-plier-data. Due to size constraints, the recount2 data processed using source code from that repository is stored on figshare and that's what is used to wget the recount2 expression data and PLIER model.

[huqiwen0313]

Looks good to me

[gwaybio]

I am wondering why you're storing these files in this repo also. These files can be read in directly from the url (in both R and python)

url <- 'https://github.com/greenelab/rheum-plier-data/raw/4be547553f24fecac9e2f5c2b469a17f9df253f0/sle-wb/processed/aggregated_data/SLE_WB_all_microarray_QN_zto_before.pcl'
data <- readr::read_tsv(url)

[jaclyn-taroni]

In the case that I have to load this into R in multiple scripts/notebooks/what have you, I think it would be preferable to not have to read this from the URL (somewhere between 400-500MB in this particular case) each time. I can add these files to .gitignore, though, if you think that's better.

[gwaybio]

Is it faster to read from the file than the url? Are there other concerns too?

[cgreene]

It should be a few orders of magnitude faster unless this does some serious caching.

[gwaybio]

yeah, in python (did not test in R)

%%time
import pandas as pd
url = 'https://github.com/greenelab/rheum-plier-data/raw/4be547553f24fecac9e2f5c2b469a17f9df253f0/sle-wb/processed/aggregated_data/SLE_WB_all_microarray_QN_zto_before.pcl'
data = pd.read_table(url)

CPU times: user 6.61 s, sys: 1.51 s, total: 8.12 s
Wall time: 55.1 s

%%time
file_loc = 'SLE_WB_all_microarray_QN_zto_before.pcl'
data = pd.read_table(url)

CPU times: user 6.61 s, sys: 1.2 s, total: 7.81 s
Wall time: 12.4 s

[cgreene]

Fast internet connection there 👍

[gwaybio]

I wonder if this use case is suitable as a git submodule then 🤔 . Up to you @jaclyn-taroni - I will approve this PR. (just wondering about pros vs. cons for alternative solutions)

[jaclyn-taroni]

So I think I will keep downloading the data this way rather than using a submodule or reading directly from the url, but I will ignore the data from that GitHub repo (same as what I'm doing with the recount2 data from figshare). Will update in next commit.