string(13) "dissertations"
  • Python programming; Probabilities; Linear algebra
  • Machine learning toolkits (e.g. Scikit learn, PyTorch); Data analysis; Embedding; Understanding of basic molecular dynamics concepts
  • George Giannakopoulos
  • ggianna [at]
  • The AI Lab

Molecular simulation is a powerful technique to investigate the properties of complex chemical systems and understand the microscopic mechanisms that determine macroscopic materials properties. However, it is an extremely computationally intensive method, and this limits the number of atoms that can be simulated, even with state-of-the-art supercomputing resources.

In order to apply these methods to large molecules, such as polymers, it is essential to develop a representation of the systems with a reduced level of detail (Coarse Graining), in order to carry out large scale molecular simulations. After the simulation, it is necessary to reconstruct the atomistic detail, to perform the relevant properties calculation (Back-Mapping). These processes are currently carried out in a system-specific manner and rely on chemical intuition.

In the overarching project into which the work will be integrated, we aim to develop a systematic approach to Coarse Graining/Back-mapping, dimensionality reduction techniques, such as Autoencoder Neural Networks, will be applied to the efficient generation of coarse grained representations of polymer molecules, which preserves physical consistency and captures the relevant features of the system. Subsequently, Back-mapping strategies will be implemented, leveraging Machine Learning methods and investigating the possibility to employ a graph-based representation of the molecular system, thus formulating the task as a graph layout problem.

The project is conducted at the National Centre of Scientific Research Demokritos and involves the synergy of Software and Knowledge Engineering Lab (SKEL | The AI lab)  of the Institute of Informatics Telecommunications and the Molecular Thermodynamics and Modeling of Materials Laboratory (MTMML) of the Institute of Nanoscience and Nanotechnology.

Since the project can be either an internship (at least 2 months full time), or a BSc/MSc thesis, the exact assignment will be decided on a case-by-case basis with the interested student.


Dr Georgios Giannakopoulos,

Dr Niki Vergadou,

Relevant literature

  1. Wang, W., Gómez-Bombarelli, R. Coarse-graining auto-encoders for molecular dynamics. npj Comput Mater 5, 125 (2019).
  2. An, Yaxin, and Sanket A. Deshmukh. “Machine learning approach for accurate backmapping of coarse-grained models to all-atom models.” Chemical Communications 56.65 (2020): 9312-9315.
  3. Theodorou, D.N., 2004. Understanding and predicting structure–property relations in polymeric materials through molecular simulations. Molecular Physics 102, 147–166.

Figure from: Wang, W., Gómez-Bombarelli, R. Coarse-graining auto-encoders for molecular dynamics. npj Comput Mater 5, 125 (2019). Licensed under CC BY 4.0 (


Skip to content