Fighting cancer: simulations at unprecedented length- and time-scales
A new method shows how mutant RAS proteins – implicated in 30% of human cancers – behave in cell membranes. RAS function depends on the clustering of these proteins in the membrane. This clustering depends on the distribution of lipids in the membrane. Clustering is a macroscale phenomenon, beyond the reach of near-atomistic simulation; the parameters that govern clustering are a function, however, of near-atomistic detail. The first image on the left represents a macroscale “continuum” simulation that can show RAS proteins co-localizing. Each layer indicates the distribution of a lipid type. A machine-learning algorithm selects novel small membrane patches (shown in "latent space" at top right) that are simulated in detail (bottom right). Parameters for the macroscale simulation are derived from these microscale computations. New microscale simulations refine the macroscale parameters as the overall simulation proceeds.
See Machine learning–driven multiscale modeling reveals lipid-dependent dynamics of RAS signaling proteins. More info.
See Machine learning–driven multiscale modeling reveals lipid-dependent dynamics of RAS signaling proteins. More info.
