SindiLab/N-ACT-TrainingScripts

A repository to host the main scripts for training (and pre-training) N-ACT

Training Scripts for N-ACT

To make package development and maintaining more efficient, we have split NACT's main package, main training scripts and tutorials into different repositories. This repository will host the different main scripts for training N-ACT (e.g. supervised/unsupervised annotation or transfer learning), as well as the bash scripts for running the trainer scripts.

Installing N-ACT

N-ACT can be installed using PyPI:

$ pip install git+https://github.com/SindiLab/N-ACT.git

or can be first cloned and then installed as the following:

$ git clone https://github.com/SindiLab/N-ACT.git
$ pip install ./N-ACT

Training N-ACT

We have provided an example bash script called run_nact_trainer.sh which demonstrates which machine-specific parameters should be changed prior to execution. run_nact_trainer.sh also specifies parameters that are commonly tweaked based on user preference (such as learning rate). Please take a look at each main script for all the command line parameters that can be specified.

Once N-ACT is installed and the trainer main script is downloaded, NACT can be trained using the command line in the following form:

$ python train_nact_with_scanpy_object.py --dataset "<Nickname for Dataset>" --epochs <Number of Epochs> --data_path "<Path_To_Scanpy_Object/object.h5ad>" --annotation_key "<AnnData Key That Stores Labels for Training> " --branches <Number of Projection Branches> --where_to_save_model "<Path_To_Save_Trained_Model>/"

We provide an example of these parameters in the companion bash script run_nact_trainer.sh. Once the parameters are specified in run_nact_trainer.sh, run:

$ bash run_nact_trainer.sh

Post Training Analyses and Tutorials

We have compiled a set of notebooks as tutorials to showcase N-ACT's capabilities and interptretability. These notebooks located here.

Please feel free to open issues for any questions or requests for additional tutorials!

Citation

If you found our work useful for your ressearch, please cite our preprint:

@article {Heydari2022.05.12.491682,
	author = {Heydari, A. Ali and Davalos, Oscar A. and Hoyer, Katrina K. and Sindi, Suzanne S.},
	title = {N-ACT: An Interpretable Deep Learning Model for Automatic Cell Type and Salient Gene Identification},
	elocation-id = {2022.05.12.491682},
	year = {2022},
	doi = {10.1101/2022.05.12.491682},
	journal = {The 2022 International Conference on Machine Learning (ICML) Workshop on Computational Biology Proceedings.},
	URL = {https://www.biorxiv.org/content/early/2022/05/13/2022.05.12.491682},
	eprint = {https://www.biorxiv.org/content/early/2022/05/13/2022.05.12.491682.full.pdf},
}