nightknight3000 / contact-mutation-regression

Contact-Mutation-Regression

Setup

1. git clone https://github.com/Nightknight3000/Contact-Mutation-Regression
2. pip install -r requirements.txt

Usage

The CLI - Command Line Interface

> python start_contact_mutation_regression.py
usage: start_contact_mutation_regression.py -c <learning algorithm> -i <inputfilepath> -p <inputfilepath> -o <outputpath> [-s/-v]
-c <arg> either 'xgb', 'sklearn', or 'ridge' to specify learning algorithm
-i <arg> csv/txt-file for classifier training
-p <arg> csv/txt-file to be predicted by classifier (optional)
-o <arg> output path for classified inputfile (optional)
-s/-v, --silent/--verbose provide -v or --verbose to see full classification process and training (default -s)
(currently only works with xgb)

Examples

Using the example file 'contact_map_blomap_6A.csv' as first inputfile in the data folder: data.

Now we can run the project using our shell of choice:

python start_contact_mutation_regression.py -c xgb -i data/inputfiles/contact_map_blomap_6A.csv -p data/inputfiles/contact_map_blomap_6A.csv -o data/outputfiles/contact_map_blomap_6A_xgb_predicted.csv

Authors

Team iGEM 2018 Tübingen
Lukas Heumos
Steffen Lemke
Alexander Röhl

contact-mutation-regression's Issues

Please add hyperparameter optimization

Hey,

hyperparameter optimization is an easy way to improve your regression performance. Furthermore, we can write lots of stuff about it on our wiki.

I suggest that you implement grid-search as well as randomized-search.
They're super simple and you can steal most of the code from here:
https://github.com/Zethson/MHC-1-Binding-Predictor-iGEM2018-Tuebingen/tree/development/src/algorithms/ml/gbtrees

Randomized search is a loooooot faster than grid-search but the results are likely a little bit worse.

Cheers