Opendata, web and dolomites

Report

Teaser, summary, work performed and final results

Periodic Reporting for period 1 - ScaleCell (Scalable Kinetic Models: From Molecular Dynamics to Cellular Signaling)

Teaser

Proteins can dynamically switch between metastable structures and associate into signaling machineries that give rise to cellular function. Molecular dynamics (MD) simulations can simultaneously probe structure and dynamics of such processes at atomistic resolution. Recently...

Summary

Proteins can dynamically switch between metastable structures and associate into signaling machineries that give rise to cellular function. Molecular dynamics (MD) simulations can simultaneously probe structure and dynamics of such processes at atomistic resolution. Recently, breakthroughs have been achieved in the long-standing problem to sample rare transition events in unbiased MD. The PI is a pioneer in the development and application of Markov state models (MSMs), which, combined with MD simulations on graphical processing units, make millisecond-timescale kinetics broadly accessible. With multi-ensemble techniques, timescales of seconds and beyond can be reached. Recently, we have demonstrated protein-protein association and dissociation with atomistic MD. Long-timescale simulation of small to medium-sized protein systems is thus now possible in atomistic MD.

However, these methods have fundamental scaling limitations that prevents long-timescale simulation to be employed in the modeling of large protein systems and whole cells. To address these limitations, the present ERC project will develop, and implement in open-source software:

1. A model structure that describes protein kinetics as a network of local switches and will overcome scaling limitations of MSMs that suffer from an exponential increase in parameters for large systems.
2. An “effective force field for cells” that predicts structure and kinetics of multi-body protein interactions based on simulations of relatively few protein interactions.
3. A multi-scale method to embed atomistic kinetic models in whole-cell reaction-diffusion simulations.

These methods may be transformative for molecular and cellular biophysics. In collaboration with leading experimentalists, we will employ the methods to obtain multi-scale models of the mechanochemical protein dynamin, its oligomers and its function in membrane constriction and fission, which may address long-standing questions in endocytosis.

Work performed

Work package 1: Scaling to large molecules: From MSMs to Markov Random Fields
We have made a significant number of advances in the robust and scalable kinetic modeling for molecules of increasing size. The core publications in this area are a new deflation method for selecting kinetic processes that are of particular interest for the function of a large protein (Husic and Noe, JCP 2019) and a first implementation of the proposed Markov Random Field approach (Olsson and Noe, PNAS 2019).

Work Package 2: Scaling to many interactions: multi-body Equivalent Energy Models
We have worked out basic theory and methodology for this work package. It became clear that the multiscale simulations planned here rely on both mechanisms for coarse-graining, but also mechanisms for sampling atomistic configurations from coarse-grained representations. For the latter, we have developed new methodology, demonstrated its usefulness and publised it (Wu et al, NeurIPS 2018, Noe et al, Science 2019)

Work Package 3: Scaling to cells: a hybrid MSM - reaction diffusion simulation scheme
We have developed a general hybrid MSM-reaction-diffusion simulation scheme (to be published) and continued to develop interacting-particle reaction dynamics methodology and the ReaDDy simulation package (Hoffmann, Fröhner, Noé, PLOS One 2019, Fröhner and Noe, JPCB 2018)

Final results

The most noteworthy progress beyond the state of the art are:
1. The development of dynamic graphical models for modeling the kinetics of large-scale molecules with many uncoupled or loosely coupled domains. This is a completely novel way of describing kinetics whose relevance is becoming more and more clear in the community (Olsson and Noe, PNAS 2019).
2. Boltzmann Generators: The one-shot sampling of configurations of many-body systems such as proteins with deep learning (Noe et al, Science 2019). This development was not planned in the beginning of the project, but rather came out of a necessity in Work Package 2.

We expect that especially the Boltzmann Generators will lead to a significant amount of follow-up work, as it represents a new and powerful way to address the sampling problem of many-body physics systems.

Website & more info

More info: https://www.mi.fu-berlin.de/en/math/groups/comp-mol-bio/projects/erc_con/index.html.