DeepMEL1

Contents

DeepMEL1#

The DeepMEL1 model is a topic classification model trained on 339K ATAC-seq peaks from 16 human samples with the goal of investigating MEL- and MES-state enhancer logic.

Using pycistopic, binarized topics per region were extracted for 24 target topics, where topic 4 and topic 7 represent MEL- and MES-specific enhancers. In addition, two topics have regions that are generally accessible accross all cell lines (topic 1 and topic 19).

These sets of regions were used as input for a DL model, where 500bp one-hot encoded (ACGT) DNA sequences were used to predict the topic set to which the region belongs.

The model is a hybrid CNN-RNN multiclass classifier which is very similar to deeptopic_lstm() with addition of a reverse complement layer in the first layer of the model.

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


Citation

Minoye, L., Taskiran, I.I. et al. Cross-species analysis of enhancer logic using deep learning. Genome Res. 30, 1815–1834 (2020). https://doi.org/10.1101/gr.260844.120

Usage#

 1import crested
 2import keras
 3
 4# download model
 5model_path, output_names = crested.get_model("DeepMEL1")
 6
 7# load model
 8model = keras.models.load_model(model_path)
 9
10# make predictions
11sequence = "A" * 500
12predictions = crested.tl.predict(sequence, model)
13print(predictions.shape)