Official Implementation of

Squeeze and Learn: Compressing Long Sequences with Fourier Transformers for Gene Expression Prediction

Import the notebooks in your Colab Environment.

• In Workflow_GPU you can find two example notebook of FNetCompression and FNet_1_1 on Xpresso's dataset using GPU.
• In Workflow_TPU you can find two example notebook of FNetCompression and FNet_1_1 on CTB dataset using TPU.
• In Classes/Transformer.ipynb you can find the class that manage all the transformer-based solution of the paper.
• In Classes/DataManager.ipynb you can find the class that manage Xpresso's data.
• In Hyperparameters you can find the best hyperparameters found for each dataset.

The data relating to the promoters, half-life features and median gene expression levels are available thanks to the following publication:

• Agarwal V, Shendure J. Predicting mRNA abundance directly from genomic sequence using deep convolutional neural networks. 2020. Cell Reports 31 (7), 107663