MultiGPS is a multi-task deep learning framework for informativetarget panel selection. It combines stochastic gates with multi-task optimization, enabling end-to-end differentiable gene panel selection with multiple objectives covering comprehensive aspects regarding ST datasets.
The requirements of MultiGPS can be installed by:
pip install -r requirements.txt
then use pip install -e . or directly import the code.
MultiGPS provides some tutorials about the selection and evaluation. It can be used on various kinds of data including scRNA and ST dataset.
preprocess_ALM.ipynb and select_eval_ALM.ipynb provides how to preprocess, select and evaluate the scRNA dataset.
preprocess_MERFISH.ipynb and select_eval_MERFISH.ipynb provides how to preprocess, select and evaluate the MERFISH dataset.
imputation_analysis.ipynb and SVG_analysis.ipynb are two downstream analysis of ST data.