At this colloquium, we are happy to welcome:
Giovanni Bussi (SISSA - Scuola Internazionale Superiore di Studi Avanzati)
RNA dynamics from combining molecular simulations and experimental data
Molecular dynamics (MD) simulations can be used to reconstruct RNA dynamics [1,2]. However, MD simulations suffer from poor accuracy due to incorrect force field parametrizations. I will show how it is possible to reconcile MD simulations with experiments following different possible routes, namely: force field refinement, ensemble refinement, forward model refinement, or arbitrary combinations of these approaches [3,4]. I will then use applications to biologically relevant systems to showcase the pros and the cons of the different approaches [5,6]. Finally, I will share our recent experience with CASP, showing how one can address the precision of MD simulations by introducing suitably tuned enhanced sampling methods that enable the characterization of RNA solvation in the difficult case of buffers containing divalent cations [7].
[1] Bussi et al, Structure https://doi.org/10.1016/j.str.2024.07.019 (2024)
[2] Bernetti and Bussi, COSB https://doi.org/10.1016/j.sbi.2022.102503 (2023)
[2] Gilardoni et al, JPCL https://doi.org/10.1021/acs.jpclett.3c03423 (2024)
[3] Gilardoni et al, arXiv https://arxiv.org/abs/2411.07798 (2024)
[5] Piomponi at al JPCB https://doi.org/10.1021/acs.jpcb.4c03397 (2024)
[6] Posani et al, bioRvix https://doi.org/10.1101/2024.07.24.604258 (2024)
[7] Languin-Cattoën et al, in preparation. See results at https://www.predictioncenter.org/casp16
Luca Ghiringhelli (FAU Friedrich-Alexander-Universität)
Bridging scales in materials modeling with sparse inference
The modeling of macroscopic properties of materials often require to accurately evaluate physical quantities at several time and length scales.Here we show how symbolic inference, i.e., the machine learning of simple analytical expressions that explain and generalize the available the data, can effectively bridge physical scales. The focus is on learning models that are as simple as possible (but not simpler…), with as few as possible data points. I will demonstrate the application of the methods to the modeling of catalytic properties of materials, thermal conductivity, and more.