ikarus is a stepwise machine learning pipeline that tries to cope with a task of distinguishing tumor cells from normal cells. Leveraging multiple annotated single cell datasets it can be used to define a gene set specific to tumor cells. First, the latter gene set is used to rank cells and then to train a logistic classifier for the robust classification of tumor and normal cells. Finally, sensitivity is increased by propagating the cell labels based on a custom cell-cell network. ikarus is tested on multiple single cell datasets to ascertain that it achieves high sensitivity and specificity in multiple experimental contexts. Please find more information in the corresponding publication.

ikarus currently supports python>=3.8, and can be installed from PyPI:

pip install ikarus

Alterantively, one can install ikarus' master branch directly from github:

python -m pip install git+https://github.com/BIMSBbioinfo/ikarus.git

The easiest option to get started is to use the provided Tumor/Normal gene lists and the pretrained model:

from ikarus import classifier

model = classifier.Ikarus(signatures_gmt=signatures_path)
model.load_core_model(model_path)
predictions = model.predict(test_adata, 'test_name')

More information on how to train a model or how to create own gene lists is provided in the tutorial notebook.