Cover design proposal for Nature Chemistry, illustrating a machine-learned coarse-grained model for protein simulation developed by the group of Prof. Dr. Cecilia Clementi at FU Berlin.
Simulating proteins atom by atom is accurate but extremely expensive. Coarse-graining lowers the cost by collapsing groups of atoms into single beads. The hard part is building one model that also works on proteins it was never trained on.
The illustration shows the same protein chain in both representations as it crosses a free-energy landscape: all-atom on the left, with explicit water molecules, dissolving into coarse-grained beads on the right. The translucent spheres mark the atoms absorbed into each bead. The hills and valleys of the surface are the folding energy landscape the model has to navigate, the peaks are barriers between stable folded and unfolded states.
