Related Experiment Video
Updated: Oct 25, 2025
10:21
Author Spotlight: Streamlining Protein Target Prediction and Validation via Molecular Docking and CETSA
Published on: February 23, 2024
3.1K
Deep learning to design nuclear-targeting abiotic miniproteins
Carly K Schissel1, Somesh Mohapatra2, Justin M Wolfe1,3
1Department of Chemistry, Massachusetts Institute of Technology, Cambridge, MA, USA.
Nature Chemistry
|August 10, 2021
Summary
Machine learning designs novel nuclear-targeting miniproteins for delivering genetic material. These
Area of Science:
- Biochemistry
- Molecular Biology
- Artificial Intelligence
Background:
- Designing functional polymers is challenging due to the vast sequence space.
- Human learning is insufficient for exploring this chemical complexity.
Purpose of the Study:
- To develop a machine learning approach for de novo design of functional polymers.
- To create abiotic nuclear-targeting miniproteins for delivering antisense oligomers.
Main Methods:
- Combined high-throughput experimentation with a deep-learning model.
- Represented molecular structures as topological fingerprints.
- Employed a directed evolution-inspired approach.
Main Results:
- Developed 'Mach' miniproteins (average mass 10 kDa) capable of nuclear targeting.
- Achieved higher efficacy in cells compared to previous variants.
- Demonstrated successful protein delivery into the cytosol and non-toxic delivery of antisense cargo in mice.
Conclusions:
- Deep learning can decipher design principles for biomolecules.
- Enables generation of highly active biomolecules beyond empirical discovery.
- Shows potential for therapeutic applications in gene delivery.
