From Viral Sequences to Fitness Landscapes: A New Paradigm for in Silico Vaccine Design

Of the 34 million people living with HIV/AIDS, 23 million reside in the poor nations of sub-Saharan Africa that are least able to counter this epidemic. Hepatitis C virus (HCV) afflicts around 150 million people worldwide, and is the leading cause of liver transplantation in the developed world. No effective HIV or HCV vaccines exist. An inexpensive prophylactic vaccine offers the best hope to curb these epidemics, but systematic means to guide the design of effective vaccine immunogens for these, and other, infectious diseases are unavailable.

Appealing to spin glass models in statistical physics, we present a novel approach to translate viral sequence databases into landscapes of viral fitness described by Ising and Potts Hamiltonians. These landscapes chart the peaks and valleys of viral fitness as a function of amino acid sequence, and can be used to rationally design vaccines to eject the virus from the high fitness peaks, and drive it into the valleys where its compromised fitness impairs its ability to replicate and inflict damage to the host.

We illustrate this approach in the development of fitness landscapes for particular HIV and HCV proteins known to be particularly susceptible to immune pressure and drug therapies. In comparisons to experimental and clinical data, our inferred landscapes demonstrate good agreement with: 1) in vitro replicative fitness measurements, 2) clinically observed high­fitness circulating viral strains, 3) documented HLA associated CTL escape mutations, and 4) intra­host viral mutation pathways revealed by deep sequencing. We demonstrate the value of data driven fitness landscapes in the in silico design of cytotoxic T-cell vaccine immunogens against HIV-1 clade B. We are currently working with experimental and clinical collaborators to test our designed immunogen in mice models of HIV infection.