A key to understanding KRAS-driven cancers and the role of mutant RAS proteins and to discover new therapeutic possibilities is to better understand how RAS behaves on the cell membrane.
Investigating RAS in the context of membranes is somewhat challenging using conventional computational or experimental techniques. In a partnership with researchers from three Department of Energy national laboratories and others, scientists from the Frederick National Laboratory for Cancer Research (FNL) developed a detailed, machine learning-enabled multiscale model that simulated KRAS on a complex model membrane. The model revealed the importance of KRAS-lipid interactions for dynamics of RAS and the assembly of local domains enriched for signaling-competent KRAS clusters.
Published by the Proceedings of the National Academy of Sciences, the paper details the methodology behind the Multiscale Machine-Learned Modeling Infrastructure (MuMMI), which simulates the behavior of RAS proteins on a realistic cell membrane, their interactions with each other and with lipids — organic compounds that help make up cell membranes — and the activation of signaling through the RAS interaction with RAF proteins, on a macro- and molecular level.
Answers to central questions
Mutated RAS is implicated in 30% of all cancers, including lung, colorectal and nearly all pancreatic cancers. Normally, RAS receives and follows signals to switch between active and inactive states, but as the proteins move along the cell membrane — like balls of string tumbling along a fluid ground — they associate with other proteins that collectively affect signaling. Mutated RAS proteins become stuck in an “always on” growth state, leading to the deregulated tumor growth characteristic of cancer.
RAS researchers seek to understand how RAS behaves on the cell membrane and if there is a specific membrane environment that promotes RAS molecules coming together. Evaluating more than 100,000 co-related simulations, study findings shed light on how RAS binds to other proteins and how different kinds of lipids dictate how RAS collects and positions itself on the cell membrane. The team captured prior known protein interactions and much more showing that RAS associates without a specific interface. These data indicate that lipids rather than protein interfaces govern both RAS orientation and accumulation of RAS proteins.
From laboratory to simulation and back
Discoveries made at FNL, hub of the National Cancer Institute’s RAS Initiative, initialized and informed the MuMMI model. Parameters such as RAS structures, how quickly RAS moves across the cell membrane, how tightly it binds to it and which lipids RAS associates with provided a baseline for the model which in turn generated hypotheses that could be validated through experiments. Knowledge gained from the experiments will be fed back into the refinement of the MuMMI model, creating a validation loop that will make it more accurate through iterative cycles. “By tying together insights, machine learning and experimentation, the information loop improves over time,” said Dwight Nissley, Ph.D., head of the FNL’s Cancer Research Technology Program and corresponding author on the study.
Previous simulations contained insufficient quantity of simulations and they did not last long enough to draw robust conclusions. MUMMI employs a macro model to explore interactions between lipids and proteins, conducted over a sufficiently long time period to observe statistically relevant findings.
“The scale of simulations is a hundred-fold greater than has been done previously, representing a substantial advance in the ability to characterize RAS-membrane biology,” Nissley said.
The model is being extended to understand how the RAF protein interacts with RAS on the cell membrane, causing the signaling cascade.
The paper is part of an ongoing pilot project of the Joint Design of Advanced Computing Solutions for Cancer (JDACS4C) collaboration between the Department of Energy, the National Cancer Institute, FNL and other organizations. Additional FNL authors include Frank McCormick, Ph.D., Dhirhendra Simanchu, Ph.D., and Andrew Stephen, Ph.D.
Co-authors from outside organizations include researchers from Lawrence Livermore National Laboratory, Los Alamos National Laboratory, Argonne National Laboratory, the University of California, San Francisco, IBM’s Thomas J. Watson Research Center and San Jose State University.