Analysis notebooks and scripts for the first paper about vampire

This repository makes the plots for the paper Deep generative models for T cell receptor protein sequences by Kristian Davidsen, Branden J Olson, William S DeWitt III, Jean Feng, Elias Harkins, Philip Bradley and Frederick A Matsen IV.

To make the figures in the paper, download results files from https://zenodo.org/record/2619576#.XKElTrfYphE and place in an input directory in the root of this repository. Make an output directory as well.

Then run these notebooks in the vampire conda environment built as described in the main vampire repository. You will also need to install jupyter as follows:

conda install jupyter
conda install -c r r-irkernel

as well as install the R packages cowplot, latex2exp, and reshape2.

Reproducing results

The above instructions only concern making plots from output files. If you want to reproduce the analysis, you will want to modify the main vampire repository to work with your cluster scheduler and then follow the instructions below.

Comparative analyses

To build the results for plotting,

• download the relevant data from immuneACCESS
• run python util.py split-repertoires from the main repository (you can see an example call in _output_deneuter-2019-02-07/deneuter-2019-02-07.json)
• Run scons --data=/path/to/the/resulting/json/file.json

Cohort frequency analysis

This repo also includes a script prep-cohort-frequency.sh that prepares files for the cohort frequency analysis. If you want to reproduce this analysis,

• download the data from Adaptive
• preprocess it using preprocess_adaptive.py script in the vampire repository
• run the prep-cohort-frequency.sh script (editing paths)
• run the pipe_freq pipeline with scons --pipe=pipe_freq (editing the path in the SConstruct file)