Dresden 2026 – scientific programme
Parts | Days | Selection | Search | Updates | Downloads | Help
BP: Fachverband Biologische Physik
BP 14: Poster Session II
BP 14.71: Poster
Tuesday, March 10, 2026, 18:00–21:00, P2
Applying a topology sensitive metric for RNA contact prediction — •Christian Faber1, Utkarsh Upadhyay1, Oskar Taubert2, and Alexander Schug1,2 — 1Jülich Supercomputing Centre, Forschungszentrum Jülich, 52428, Jülich, — 2Scientific Computing Centre, Karlsruhe Institute of Technology, 76344, Eggenstein-Leopoldshafen
Predicting the spatial structure of non-coding RNA (ncRNA) is an important task for understanding fundamental processes in living nature. Physical force fields are used to infer the structure from a sequence using simulations on high-performance computers. However, the best results are obtained by incorporating probable contacts as additional restraints. These can be derived from evolutionary data using statistical methods or from more recent artificial intelligence (AI) algorithms.
In the past, the focus was on achieving the highest possible proportion of correctly predicted contacts, while the distribution of these contacts on the contact map was overlooked. We have demonstrated the importance of this distribution for structure prediction and have therefore introduced a measure of it.
In our current work, we apply our new metric to a state-of-the-art algorithm Barnacle. To achieve this, the algorithms must undergo complete retraining and a new dataset must be generated that avoids data leakage. While our results demonstrate the practical application of such a procedure, they also underscore the challenges posed by the limited availability of data for RNA molecules, a problem which becomes particularly apparent when modelling AI networks.
Keywords: Contact Maps; Artificial Intelligence; RNA