PULSE Challenge: ML/AI approaches for designing proximity-inducing small molecules – PROTAC cell permeability and bioavailability
February 25, 2025 @ 12:00 – 13:00 CET
The seminar series PULSE Challenge is connected to the MSCA co-funded* postdoctoral program SciLifeLab PULSE, that will train 48 future leaders in life sciences. The program focuses on innovative, fundamental and translational research carried out in supportive and diverse academic and industrial environments, preparing postdocs with necessary skills for long-term career sustainability. Click HERE to find out more about SciLifeLab PULSE
On-line event via Zoom
Presenters:
Prof. Jan Kihlberg & Assoc. prof. Vasanthanathan Poongavanam
Uppsala University, Sweden
Abstract:
Most PROTACs which have a “druglike” oral bioavailability (>5%) are based on a CRBN E3-ligase ligand, but a few VHL PROTACs also display >5% bioavailability. Use of 2D descriptors places oral drugs, CRBN and VHL PROTACs in different parts of chemical space, while the use of 3D descriptors indicates that the chemical spaces overlap. Use of 3D descriptors may therefore enhance the probability of discovering PROTACs with satisfactory oral bioavailability
We have studied CRBN and VHL PROTACs by NMR spectroscopy and MD simulations. These studies revealed that the propensity of the PROTACs to adopt folded and semi-folded conformations with low solvent – accessible 3D polar surface area correlated to higher cell permeability. The length, chemical nature and flexibility of the linker was found to be essential for allowing intramolecular hydrogen bonds, π–π interactions and van der Waals interactions to stabilize (semi)folded conformations for permeable PROTACs in a membrane-like environment.
Use of MD simulations for prediction of the conformations and properties of PROTACs is time consuming. However, classification models that were developed by machine learning using 2D descriptors allowed differentiation between VHL PROTACs that had high or low cell permeability. Such models can be integrated as effective filters to prioritize PROTACs for synthesis in the design process. Nevertheless, certain challenges remain, including the imbalanced nature of datasets, and model interpretability. Addressing these limitations in future studies will be crucial for maximizing the full potential of machine learning in PROTAC design.
Biographies:
Jan Kihlberg holds a chair in Organic Chemistry at Uppsala University, Sweden since 2013. His key research interests are to understand what properties convey cell permeability, aqueous solubility and target binding to drugs such as PROTACs and macrocycles in the beyond rule of 5 space and to translate this knowledge into guidelines for design. He has published more than 200 peer–reviewed publications and book chapters. He establishing his independent research group at Lund University in 1991, became full professor in Bioorganic Chemistry at Umeå University in 1996, then moved to AstraZeneca R&D in Gothenburg in 2003 while maintaining a research group at Umeå University. At AstraZeneca Prof. Kihlberg held the role as Director, Head of Medicinal Chemistry for seven years.
Vasanthanathan Poongavanam is a computational chemistry and AI expert at the DDD Platform, SciLifeLab, Uppsala University, Sweden. His research focuses on applying computational chemistry and AI methodologies to accelerate drug discovery projects. He has authored more than 60 scientific publications, including research articles, reviews, and book chapters.
Google Calendar 
