DEEPrior: a deep learning tool for the prioritization of gene fusions

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Gene fusions are increasingly studied, but distinguishing cancer drivers from passenger mutations remains challenging.
• Identifying oncogenic gene fusions is crucial for targeted cancer therapies.

Purpose of the Study:

• To introduce DEEPrior, a flexible deep learning tool for predicting the oncogenic potential of gene fusions.
• To provide a tool that analyzes amino acid sequences of fused proteins for cancer driver identification.

Main Methods:

• DEEPrior utilizes deep learning to predict the probability of a gene fusion being a cancer driver.
• The tool has two modes: Inference for direct prediction and Retraining for custom model development.
• It directly exploits the amino acid sequence of the fused protein for analysis.

Main Results:

• DEEPrior can predict the probability of a gene fusion's involvement in oncogenesis.
• The tool offers flexibility through its Inference and Retraining modes.
• It provides a novel approach to analyzing gene fusions using protein sequence data.

Conclusions:

• DEEPrior offers a powerful and flexible deep learning solution for identifying cancer-driving gene fusions.
• The tool's ability to incorporate user-provided data enhances its applicability in diverse research settings.
• This work contributes to a better understanding of gene fusions in cancer biology.