Predicting gene and protein expression levels from DNA and protein sequences with Perceiver

Matteo Stefanini, Marta Lovino, Rita Cucchiara, Elisa Ficarra

Abstract: Highlights•To understand dysregulated biological processes, predicting mRNA and protein levels is crucial in clinical applications.•A transformer-based architecture with asymmetric attention (Perceiver) is exploited for mRNA and protein level prediction.•The Perceiver architecture attends to longer range interactions compared to Transformer, CNN, and LSTM.•The proposed model achieves state-of-the-art performance for mRNA level prediction on glioblastoma and lung cancer tissues.•To the best of our knowledge, the protein level prediction task is addressed.