MareoRaft/tcr-analysis

About

A single blood sample taken from a person has many T-cells in it. Each T-cell has RNA in it. Each RNA strand has a CDR3 sequence in it. From the sample we record which CDR3 sequences occur and the frequency of each one. This data is called the TCR repertoire (T-cell receptor repertoire) of the sample.

Given two separate samples, we can define a distance between them. In the file sample_distances.py you can see various distance functions that we are experimenting with (so far we have the l2 distance (Euclidean), the jaccard distance, and the weighted_jaccard distance).

The goal of this repo is to use metrics to evaluate how good these distance functions are, as well as perform other TCR analysis.

Install

Install Docker. That's it!

Run dev environment

docker-compose up --build

then visit localhost:4001 or http://localhost:4001/lab/tree/analysis.ipynb. You should see a Jupyter frontend.

Test dev

To run tests, run the dev environment (see above) and then (in a separate terminal window):

docker exec -it tcr-analysis-container bash
pytest

build, test, and deploy

The server is a dev server

todo: once the below is reliable, move it into a script so it can be fully automated

last update took just 5 minutes. there was no docker update though.

git push docker push mvlancellotti/tcr-analysis:dev ssh date

if it says 'System restart required', restart the machine

sudo reboot su matt tmux

reattach to 'tcr-analysis' tmux session

cd tcr-analysis

commit changes if biologist made any meaningful changes, otherwise just throw them out

git checkout -- . git pull sudo docker pull mvlancellotti/tcr-analysis:dev

consider deleting old images if the droplet runs out of space

use ctrl-C to exit current container

remove the container if necessary

sudo docker rm tcr-analysis-container

ALL docker commands including this one are not necessary if you didn't make docker changes

sudo docker-compose up --build

then visit date:4001 or http://date:4001/lab/tree/analysis.ipynb.