SciDAC Review - Computational Molecular Modeling for Biofuel Research

News

New SciDAC Project

Computational Molecular Modeling for Biofuel Research

The SciDAC portfolio has again expanded to include a new project, "Understanding the Processivity of Cellobiohydrolase Cel7a (CBH I)." The project is led by Dr. Michael Himmel of the National Renewable Energy Laboratory and is supported by researchers from leading institutions across the country including Cornell University, the Scripps Research Institute, ORNL, the Forest Products Research Lab, and the University of California–San Diego. Using molecular dynamics (MD) simulations, this project will investigate the mechanisms by which enzymes hydrolyze and degrade cellulose. The result of this multidisciplinary effort will be fundamental advances in our understanding of the action of these critically important microbial enzymes, called cellulases. An additional benefit of this work will be the development of MD codes that can efficiently utilize terascale and petascale computer systems. The improvements in modeling software will include both straight MD to include more scientifically important and demanding techniques, including enhanced sampling, and quantum mechanical approaches. Ultimately, this work will lead to molecular simulations of systems the size of the entire cellulose-degrading enzymes acting on cellulose.

Figure 1. The MD results for modeling the entire CBH I protein docked on a cellulose microfibril.

To reduce our nation's dependence on foreign oil, bioethanol has garnered considerable attention as a viable alternative fuel. Its production from biomass such as agricultural residues and energy crops represents an enormous renewable resource for the production of ethanol and other liquid fuels. However, producing ethanol from biomass depends on a deconstruction step to form fermentable sugars—a step that requires costly enzymes. By increasing the efficiency and thermal tolerance of these expensive enzymes, the ultimate cost of ethanol production can be reduced. A detailed understanding of cellulase structure and function at the molecular level will require significant computational effort and expansion of the capabilities of existing tools.

The class of cellulases called cellobiohydrolases (CBH) are thought to decrystallize and processively depolymerize cellulose, producing glucose and cellobiose. The dominant cellobiohydrolase found in the fungus Trichoderma reesei, CBH I, is notable among cellulases. These enzymes are truly protein machines in that they translate along a single strand of crystalline cellulose (a cellodextrin) found in the microfibrils of plant cell walls and hydrolyze alternate beta (1-4) glucosidic linkages. Thus, they provide a direct link between the important biopolymer found in plant mater (cellulose) and sugars that can form the basis of many renewable fermentation technologies using biochemical processes. Understanding the mechanisms involved and improving the efficiency of this hydrolysis process through computational models and protein engineering presents a compelling grand challenge.

A molecular-level understanding of CBH I structure and function is required to direct protein engineers to the right modifications or, indeed, to understand if natural thermodynamic or kinetic limits are in play. This new project will leverage existing computational tools and expand the capabilities of these tools to investigate the complex mechanism of CBH I. Much can be learned by conducting careful modeling of the binding and catalytic domains of CBH I with biologically relevant cellulose surfaces. Some simulations are within the range of existing MD software, such as CHARMM and Amber, running on existing terascale computational facilities. Other critical simulations, however, will require significant modification of existing programs in order to efficiently utilize current (terascale) and future (petascale) computers to simulate systems the size of CBH I acting on microfibrils.

Early Results
Some preliminary studies using MD simulations have been performed to study individual components of CBH I and the cellulose substrate (figure 1). Such simulations provide a method by which to model the structure of cellulose in plant cell walls and the water layers above it. MD simulations have also provided a better understanding of the critical aspects of the binding of CBH I on crystalline cellulose surfaces. A preliminary MD simulation of a complete three-component cellulase has already been conducted by the team. The new project will include detailed studies of the domains of CBH I, isolated cellulose substrates, and the entire enzyme/cellulose ensemble, which will be modeled through this cross-disciplinary, five-year effort.

Plan for Discovery
An integrated plan for this effort will coordinate the production of cellulase and cellulose models, their simulation using improved MD codes, and dissemination of outcomes. Part of the team will form a biological application task force, which will conduct simulations, analyze results, provide guidance for construction of biological models, and compare results to experimental observables. Others will form the program optimization task force that will focus on the challenges of scaling MD simulation codes to make efficient use of available computing resources with the objective of producing one code that efficiently scales to greater than 100,000 processors. Moreover, this task force will focus on producing highly scalable versions of the essential simulation methods that are not currently present in existing, scalable, MD programs. In other words, rather than re-inventing the wheel, this team will focus on taking what already exists and improving it dramatically for this important biological application. Together, both tasks will ensure that the code is numerically correct, has the necessary features to run MD on the enzyme problems of interest to DOE, and is sufficiently general to address other molecular systems.