Your role
Despite the emergence of AlphaFold and related AI-based modeling tools, large macromolecular complexes still cannot be consistently modeled with high accuracy. Experimental data can guide the modeling of large complexes; however, methods for modeling based on experimental data are not sufficiently automated, computationally inefficient, and lack general applicability. To address this, you will:
1. Develop computational methods to model the atomic structure of large macromolecular complexes using tools such as AlphaFold, RoseTTAfold, and OpenFold, alongside your own innovative AI-based algorithms, and integrating experimental data from in-cell cryo-ET tomography and crosslinking mass spectrometry
2. Investigate and implement methods that combine statistical techniques like Bayesian inference and Monte Carlo sampling with non-linear optimization approaches and advanced deep learning (e.g., geometric deep learning and reinforcement learning).
3. Contribute to open-source codebases such as OpenFold, AlphaPulldown, and Assembline
4. Disseminate your code in well-organized, documented, and rigorously tested software packages
5. Use these new methods for research projects within the TransFORM consortium.
You will be affiliated with the research group of Jan Kosinski at EMBL Hamburg, but you will spend at least a year as a visiting postdoctoral fellow at the Fabian Theis group at Helmholtz Münich.
Closing date: 19 August 2024
6. Contract duration: 3 years with possibility of extension up to 5 years
7. Grading: Year 1 stipend - €4,010 per month after tax
8. Reference number: HH00227
Related
You have
9. Ph.D. in Computer Science, Computational Structural Biology, or a related field
10. Proficiency in numerical libraries such as NumPy and SciPy
11. Experience with machine learning techniques and frameworks, including Scikit-learn, TensorFlow, PyTorch, or JAX
12. Experience in the dissemination and management of software packages
13. Strong motivation to work in a highly collaborative and multidisciplinary environment of EMBL, Helmholtz Münich, and the TransFORM consortium
You might also have
14. Experience in computational structural modeling or deep learning. However, candidates with a strong background in either computer science (and a willingness to learn structural biology) or structural biology (with proven extensive programming expertise) are encouraged to apply.
15. Experience in deep learning techniques that can be applied for modeling or analysis of 3D macromolecular structures, such as geometric deep learning, graph neural networks, transformer and attention-based models, and reinforcement learning.