DeepPROTACs is a deep learning-based targeted degradation predictor for PROTACs

Enlisting deep learning for targeting ‘undruggable’ proteins

28 February 2023

Deep learning has been used to aid the design of drugs capable of targeting ‘undruggable’ proteins.

Because they employ a different mechanism from conventional drugs, novel drugs known as proteolysis targeting chimeras (PROTACs) are showing promise for tackling proteins that are not amenable to treatment by traditional strategies.

But it has been tricky to design them rationally because of the complex nature between their structure and activity.

Now, a team led by researchers from ShanghaiTech University in China has developed a deep neural network model that can predict the efficacy that a given PROTAC will take out a target protein.

In experiments, the model demonstrated an average prediction accuracy of 78%. 

The researchers envision that the model will accelerate the design of effective PROTAC and have made it freely available online.

Supported content

References

1. Nature Communications 13, 7133 (2022). doi: 10.1038/s41467-022-34807-3