Structural and Dynamical Determinants of Influenza Transmissibility

Rommie Amaro, University of California, San Diego

Usage Details

Both seasonal and pandemic influenza have been responsible for millions of deaths worldwide. The persistence of seasonal influenza strains costs between 3,000 and 49,000 lives annually in the United States alone. The project aims to use Blue Waters to run simulations of the influenza virus to explain flu transmissibility. These simulations will be of a very large scale, involving over 200 million atoms, and are expected to provide unprecedented insights into the mechanisms of influenza virulence and drug resistance. In addition, simulations will set the stage for the longer-term goal of developing an atomic-level understanding of the influenza infection process, leading to new pathways for pharmacological intervention.

The project has used molecular dynamics (MD) simulations of the individual influenza membrane glycoproteins hemagglutinin (HA) and neuraminidase (NA) to identify experimentally validated antiviral compounds that bind to new glycoprotein pockets never before captured by experimental techniques. However, influenza glycoproteins are components of a much larger virion surface that collectively participates in host recognition and infection processes. The surface glycoprotein distributions and their decoration with glycans poses many interesting unanswered questions related to the roles of HA, NA, glycans, and substrate (and substrate mimicks) play in the viral infection and assembly process. Additionally, the functional balance of HA, which binds to host-cell receptors, and NA, which abrogates receptor binding via cleavage of terminal sialic-acid residues, has not yet been fully characterized The project request time to run two whole virion simulations: (1) H1N1 with glycans attached to key glycolysation sites on HA and NA and (2) with sialic acid bound in the sialic acid binding sites (mimicking a substrate bound complex). With the addition of the glycans and substrates, both systems will consist of over 200 million atoms. The simulations will for the first time combine large-scale, subcellular / cellular level modeling at atomic detail with Markov state model theory; effectively aggregating the many copies of the individual glycoproteins into one statistics based network of states type model that enables it to extract long timescale dynamics from the short time scale simulations.