Cancer Data Science Pulse

Modeling the Dynamics of Membrane-bound Mutant RAS to Accelerate Discovery of Novel Drug Targets

Photo of Dwight Nissley Ph.D., Director, Cancer Research Technology Program, Frederick National Laboratory for Cancer Research and Frederick Streitz, Ph.D., Chief Computational Scientist, Director, HPC Innovation Center, Lawrence Livermore National Laboratory.

This is the third in a series of posts that discuss the principles underlying the three-year collaborative program "Joint Design of Advanced Computing Solutions for Cancer (JDACS4C)." Investigators from the National Cancer Institute (NCI) and the Frederick National Laboratory for Cancer Research have been working collaboratively with computational and data scientists affiliated with several national laboratories supported by the Department of Energy (DOE): principally Argonne, Los Alamos, Lawrence Livermore, and Oak Ridge. Their aim is to develop and apply large-scale computational approaches to answer challenges in cancer biology, surveillance, screening, and pre-clinical development. This pilot builds on the accomplishments of the ongoing RAS National Mission Initiative launched by former NCI director Harold Varmus in 2013.

The goal of the NCI-DOE Collaborations Molecular Level Pilot for RAS Structure and Dynamics in Cellular Membranes is to better understand the mechanisms and dynamics of RAS interaction with the plasma membrane and subsequent signal transmission through binding and activation of RAF Kinase and, in parallel, develop machine learning-powered multiscale computational tools that enable molecular dynamic simulation of realistic biological systems. In oncogenic (mutant) RAS-driven cancers, signaling is constitutive and results in uncontrolled proliferation. Deepening our understanding of RAS/RAF/membrane biology may reveal unrealized therapeutic opportunities. The development and use of emerging high-resolution experimental approaches, predictive models, machine or deep learning, and new, more finely delineated simulations have been integrated into this biological investigation. Computational platforms are being pushed closer to the exascale level as the research team explores the potential of unsupervised deep learning to perform and validate hierarchical multi-scale modeling, quantify uncertainty, guide experiments, and abbreviate the time it takes to reach solutions.

Surface representation of the structure of oncogenic mutant of KRAS (colored violet) in complex with GTPase-activating proteins (colored light blue).

Mutant RAS proteins - most notably the isoforms HRAS, NRAS, KRAS4a, and KRAS4b - contribute to the pathogenesis of at least 30 percent of all human malignancies, including the most aggressive forms of pancreatic, lung, and colorectal cancers. RAS oncoproteins, and the signaling pathways they modulate, are tantalizing targets for therapeutic intervention since the potential clinical payoff would be immense. RAS acts as a molecular switch that can activate numerous signal transduction pathways. Oncogenic RAS signals constitutively as it has lost the ability to hydrolyze GTP to GDP and transition from the "on" GTP-bound state to the "off" GDP-bound state. From the cancer research point of view, the MAPK (mitogen-activated protein kinase) pathway that signals through RAF-MEK-ERK is of primary importance. Dysregulation of this pathway by mutant RAS proteins leads to uncontrolled cell growth, proliferation, dedifferentiation, and resistance to programmed cell death - all "hallmarks" of cancer. KRAS4b receives and propagates signal when associated with membranes and this localization is dependent on its C-terminal tail (HVR) which is farnesylated to achieve membrane attachment and further stabilized by electrostatic interactions between the HVR and lipid head groups. Since RAS activity takes place against a complex background of molecular trafficking within a dynamic lipid membrane where RAS undergoes conformational changes, it is plausible that exploring the molecular dynamics of mutant RAS and its interaction with RAF at multiple resolutions could uncover biological insights and potential targetable mechanisms.

Understanding the dynamics of normal and oncogenic RAS proteins in the context of the plasma membrane at the molecular level is a challenge that requires multidisciplinary insights from microscopy, molecular, and structural biologists, physicists, biochemists, bioengineer, data scientists, and computational experts. The RAS pilot has been designed, planned, and executed to leverage computational approaches to address experimental gaps and develop multi-scale and machine learning capabilities so that the interests of the National Cancer Institute (NCI) and those of the Department of Energy (DOE) not only co-exist in the same multidisciplinary research setting but also converge.

Ulrich Baxa (at left) with Natalia de Val Alda of the Center for Molecular Microscopy (CMM) with the Titan Cryo-EM at the Advanced Technology Research Facility (ATRF), Frederick National Laboratory for Cancer Research.  Photographed for the Insite newsletter.

New laboratory instrumentation - in particular, advances such as cryo-electron microscopy, stochastic optical reconstruction microscopy, photoactivation localization microscopy, and stimulated emission depletion microscopy - make it possible to visualize and interrogate biologic functions at the sub-molecular level, leading to the formulation of more and more finely tuned parameters with which to define algorithms and build models. This cutting-edge data generated by such advanced, highly refined instrumentation lead to advances in scientific understanding that inform advances in analytical algorithms and machine learning which then drive further advances in cancer research, forming a virtuous, iterative cycle of hypothesis generation, validation, and discovery.

The main research thrust of the RAS pilot is to develop and exploit a novel multiscale modeling framework that, after validation, enables exploration of RAS activity (e.g., RAS/RAF binding and activation) at multiple, biologically relevant time and length scales. The team is using machine learning both to help optimize execution parameters for the simulations and to examine and classify features in this complex space, including membrane structure and morphology as well as data describing activation of RAS complexes.

In the past year, the project has developed and tested this novel multiscale modeling framework, successfully connecting a macroscopic modeling technique (phase field) with a microscopic one (molecular dynamics). In building this model, the team defined an optimized lipid composition for investigating mammalian membranes and performed single-molecule tracking of RAS dynamics on live cell membranes to provide validation. In addition, the team at Frederick National Lab measured RAS poses (binding modes) on membranes with nuclear magnetic resonance and small angle neutron scattering. This information, corroborated by simulations, was instrumental in the creation and validation of the model.

Most recently, the RAS Team demonstrated their methodology on the 100+ Petaflop/s supercomputer being installed at Lawrence Livermore National Laboratory (part of the ongoing DOE Collaboration of Oak Ridge, Argonne and Livermore (CORAL) initiative). This expansion of computational capability will allow the researchers to build a validated model of interactions between RAS and the cellular membrane and to further investigate RAS biology, specifically its activation and its interactions with RAF structures. The ability to simulate RAS activity at unprecedented scale and fidelity will be used to accelerate the search for new therapeutic targets with the aim of developing a treatment for this most deadly of cancers.


This post was prepared as an account of work sponsored by an agency of the United States government. Neither the United States government nor Lawrence Livermore National Security, LLC, nor any of their employees makes any warranty, expressed or implied, or assumes any legal liability or responsibility for the accuracy, completeness, or usefulness of any information, apparatus, product, or process disclosed, or represents that its use would not infringe privately owned rights. Reference herein to any specific commercial product, process, or service by trade name, trademark, manufacturer, or otherwise does not necessarily constitute or imply its endorsement, recommendation, or favoring by the United States government or Lawrence Livermore National Security, LLC. The views and opinions of authors expressed herein do not necessarily state or reflect those of the United States government or Lawrence Livermore National Security, LLC, and shall not be used for advertising or product endorsement purposes.

This work has been supported in part by the Joint Design of Advanced Computing Solutions for Cancer (JDACS4C) program established by the U.S. Department of Energy (DOE) and the National Cancer Institute (NCI) of the National Institutes of Health. This work performed under the auspices of the U.S. Department of Energy by Lawrence Livermore National Laboratory under Contract DE-AC52-07NA27344.

Older Post
CANDLE: Scaling JDACS4C Machine Learning Algorithms to Unprecedented Magnitudes
Newer Post
Introducing the Data Commons Framework

Leave a Reply

Vote below about this page’s helpfulness.

Your email address will not be published.