Postdoc kinase affinity modelling using synthetic data

Postdoc Opportunity in Kinase Affinity Modelling Using Synthetic Data

Are you a highly motivated, ambitious Postdoc candidate interested in pursuing cutting-edge artificial intelligence research in the area of computational medicinal chemistry? Are you interested in building better models to predict the properties of the next cancer therapeutics? We have an opening for a Postdoctoral Researcher at the Molecular Physiology research group (MolPhys) with Dr. Anthe Janssen in the Leiden Institute of Chemistry (LIC).

Main Goal

Develop statistical AI models to predict binding affinities for as many kinases as possible based on docking experiments.

Previously, we have created a large dataset of experimentally determined kinase-ligand pairs, and generated 3D complexes for all measurements using open-source molecular docking software. In this project, we aim to expand on this idea by partnering with Dompé farmaceutici S.p.A. They have developed the Exscalate virtual screening platform, which allows molecular docking at an unprecedented throughput. This will enable us to dock all possible kinase-ligand combinations, around 108 complexes. We will further push to expand the coverage of the kinases included, using both Dompé’s in-house and yet-to-generate homology models to approach 100% kinome coverage (previous work reached around 40%).

The newly generated dataset of 3D poses, apart from being released to the community, will be used to develop AI models to predict binding affinities. These will be trained on the ± 350,000 literature affinity values aggregated previously. Pose selection by an in-house developed AI model of Dompé will be tested to assess the impact on performance. As pose encoding, various interaction fingerprints will be assessed, as well as other molecular descriptors such as DompeKeys. With validated models, we will generate the first-of-its-kind full matrix of predictions for all ligands against all kinase targets. If time and resources permit, other docking engines such as Uni-mol v2.0 will be used to broaden the benchmarks, and we will explore possible validation of this generated matrix by performing Absolute Free Energy (ABFE) calculations on a stringent selection of relevant new datapoints.

Work and Profile

Key Responsibilities

• Generating large scale molecular docking datasets for artificial intelligence training
• Experiment design & Data management
• Publication in high-quality journals
• Disseminating results in conferences

Essential Selection Criteria

• PhD degree in the area of medicinal computational chemistry
• Strong experience in at least one of the following: molecular docking; training and using artificial intelligence models
• Motivated to conduct independent computational research
• Good communication skills, ability and willingness to work in a team

About Us

The Molecular Physiology group is an interdisciplinary group at the Leiden Institute of Chemistry, Leiden University. The artificial intelligence research is led by Dr. Anthe Janssen. Please find more details here https://www.janssenlab.nl/.

What’s in it for you?

• Opportunity to grow your independent research vision
• Collaborative project with an industrial partner
• Infrastructural support to pursue ambitious research directions