About this research groupOur research interests focus on bridging data-driven artificial intelligence (AI) and physics-based 3D modelling to better understand the molecular basis of diseases and to rationalize drug design. read more
About this research group
Our research interests focus on bridging data-driven artificial intelligence (AI) and physics-based 3D modelling to better understand the molecular basis of diseases and to rationalize drug design. We use information obtained from available structures, homology models, and other data-sources in order to, for example, interpret variations, improve experimental design, improve protein-protein docking, etc.
Currently one of our main projects focus on developing AI methods for better cancer vaccine design. Cancer vaccine is making clinical breakthroughs in eliminating tumors. The patient’s immune defenses are unleashed against tumor cells, by their T-cell receptor (TCR) specifically binding to tumor-specific mutations (neoantigens) that are presented by MHC proteins on the surface of the tumor, forming the TCR:peptide:MHC (TCR:pMHC) complex.
Despite the encouraging clinical successes that cancer vaccines have demonstrated in some advanced-stage patients, but translation into a generally applicable therapy has remained problematic. To tackle this challenge, we are working in two directions: 1) using 3D models to improve the accuracy of AI on predicting cancer vaccine candidates, and 2) using AI to improve the 3D modelling of TCR:pMHC complexes, providing 3D interpretation of TCR specificity (i.e., the ability of TCR to selectively recognize and respond to specific pMHC complexes).
We have built a deep learning platform for data mining protein-protein complexes, DeepRank (under review). Currently, DeepRank uses 3D Convolutional network. We are also working on extending it to graph networks. We are also extending DeepRank on interpreting genetic variance.
Other projects in the field of structural bioinformatics
Other projects in our lab evolve around the use and interpretation of 3D protein structures. We collaborate with many other departments to study the molecular effects of genetic diversity. When necessary, we can build, visualize and interpret homology models as well. Information obtained from these structures and models can be used for, for example, variant analysis, drug docking, and intelligent experimental design.
We welcome talented students to join our endeavor. We offer internships for Chem/MLS/Biol students (both Ma and Ba) either with a clear interest and understanding of protein folding or with an enthusiasm of AI. An internship without programming experience is possible in the field of 3D modelling and mutant analysis. The AI project requires at least a decent basic level of programming skills. Internships, also in collaboration with other departments, are possible in many different fields.
Aims of this research group
- to understand molecular mechanisms of health and disease
- to collect and combine structural data to predict new information
Aims of this research groupThis research group aims:
- to understand molecular mechanisms of health and disease;
- to collect and combine structural data to predict new information;
- to use this knowledge/data in collaborative projects;
- to teach Structural Bioinformatics to students in the fields of Molecular and Biomedical Sciences.
Discoveries of this research group
- DeepRank - a deep learning framework for data mining very large sets of protein-protein structures
- iScore - a novel graph kernel based scoring method for ranking protein-protein docking models
- pdb2sql - a handy python tool for playing with PDB files
- PSSMgen - a handy tool to generate PSSM files and mapping them to PDB files
- Automatic Mutation server HOPE
- Tolerance plots based on homologous domains
- Search Engine for biological and medical databases
Discoveries of this research group
News published items on Radboudumc website
Here you can find all items available on teaching we do in our goup
Our Youtube channel
Here you can learn more on what we do.
Like: Integrative modelling of peptide MHC class I complexes