Go to content
SV På svenska

AI guided design for cyclic peptide drugs

Reference number
ID20-0109
Project leader
David, Florian
Start and end dates
210101-260228
Amount granted
2 500 000 SEK
Administrative organization
Chalmers University of Technology
Research area
Computational Sciences and Applied Mathematics

Summary

Cyclic peptides combine the advantages of protein therapeutics such as high target specificity with the target reach of small drug molecules. Current selection techniques for finding novel peptide leads only select for affinity to the target, but not for cell permeability. New techniques for screening and design of cell permeable peptides are of high interest to tap into the vast potential of intracellular drug targets. We will address this challenge by developing a machine learning based pipeline to design potent and cell permeable cyclic peptides. Experimental data sets are generated through a new drug discovery platform using yeast synthesising a large library of cyclic peptide variants. Big data sets on screening for potency and permeability will be used as basis for machine learning approaches to in silico design peptide sequences with improved properties. Predicted peptides will be tested in vivo and iterative optimization cycles will be conducted. We will use this pipeline on a relevant anti-cancer target, hypoxia inducible transcription factor HIF-1 and expect to find superior cyclic peptide lead structures combining both potency and permeability. This high impact project will leverage on world leading expertise in Computational Drug Discovery at AstraZeneca and Synthetic Biology at Chalmers University. It is a unique opportunity to exploit the synergies between both institutions to enable the development of a transformative novel drug discovery workflow.

Popular science description

Modern drug discovery is focused on understanding diseases on a molecular level to identify new and more specific drug targets. The common challenge we are facing is how to address and reach these targets, as 80% are inside the cell and inaccessible to current therapeutics. It is very challenging to develop new drugs which can penetrate into the cell and at the same time have a specific mode of action to avoid side effects. Within this project we are focusing on a new class of potential drugs, namely cyclic peptides. They combine the advantages of protein therapeutics such as high target specificity with target reach of small drug molecules. Current selection techniques are not able to identify peptides which can enter into cells. New tools for screening and design of cell permeable peptides are of high interest to tap into the vast potential of intracellular drug targets. In this project we will address this challenge by developing a machine learning based workflow to design potent and cell permeable cyclic peptides. Experimental data sets are generated through a new yeast-based drug discovery platform creating big data sets based on screening for potency and cell permeability. The created datasets will be fed into a computational machine learning based workflow that will optimize the cyclic peptides iteratively based on additional experimental input in each iteration. We will use this pipeline on a relevant anti-cancer target and expect to find superior drug candidates which can enter into cells and modulate the intracellular protein. This high impact project will leverage on world leading expertise in Computational Drug Discovery at Astra Zeneca and Synthetic Biology at Chalmers University. It is a unique opportunity to exploit the synergies between both institutions to enable the development of a transformative novel drug discovery workflow.