adapt-seq-design

Methods for predicting sequence activity.

This repository contains training data and models used by ADAPT, as well as analyses of their performance.

For more information on ADAPT and on how to run it, please see the ADAPT repository on GitHub.

If you find work in this repository useful, please cite ADAPT as described in the ADAPT repository.

Summary of contents

Below is a summary of this repository's contents:

• data/: Library design and training/test data for CRISPR-Cas13a.
• models/: Saved classification and regression models, as determined by a hyperparameter search.
• out/: Results of Cas13a library data analyses, model hyperparameter searches, nested cross-validation, and learning curves.
• plotting_scripts/: Scripts to produce plots on the data in this repository.
• model.sh: Wrapper script for training and evaluating models; includes calls to the Python modules.
• predictor.py, predictor_baseline.py: Methods to train and evaluate classification and regression models.
• parse_data.py: Methods to parse our data and prepare them for training/testing (encoding, splits, sampling, etc.).
• predictor_hyperparam_search.py: Methods to perform a hyperparameter search over models.
• predictor_learning_curve.py: Methods to produce a learning curve for the models.