DeepGlioma
============

.. sidebar:: Model Features

   - **Genome**: *hg38*
   - **Type**: Topic classification
   - **Parameters**: 34.9M
   - **Size**: 78MB
   - **Input shape**: (500, 4)
   - **Output shape**: (24,)

The **DeepGlioma** model is a topic classification model trained on a scATAC-seq dataset of human gliomas obtained from Wang et al., 2019 (Cancer discovery). The dataset comprises 6 patient samples, 4 IDH–wild-type GBMs, 2 IDH-mutant grade II astrocytomas, and 2 IDH-mutant oligodendrogliomas.

Using pycisTopic, binarized topics per region were extracted for 24 target topics, where topics 8/21 and topics 20/25 represent patient-specific and patient-mixed MES-like states, while topics 14/18/19 represent an OPC/NPC-like state.

The model is a CNN multiclass classifier that uses the :func:`~crested.tl.zoo.deeptopic_cnn` architecture.

Details of the data and the model can be found in the original publication.

-------------------

.. admonition:: Citation

   Kempynck, N., De Winter, S. et al. CREsted: modeling genomic and synthetic cell-type-specific enhancers across tissues and species. Nature Methods (2026). https://doi.org/10.1038/s41592-026-03057-2

.. admonition:: Data source

   Wang, L., Babikir, H., Müller, S., et al. The Phenotypes of Proliferating Glioblastoma Cells Reside on a Single Axis of Variation. Cancer Discovery (2019). https://doi.org/10.1158/2159-8290.CD-19-0329


Usage
-------------------

.. code-block:: python
    :linenos:

    import crested
    import keras

    # download model
    model_path, output_names = crested.get_model("DeepGlioma")

    # load model
    model = crested.utils.load_model(model_path)

    # make predictions
    sequence = "A" * 500
    predictions = crested.tl.predict(sequence, model)
    print(predictions.shape)