| The advancements in high-throughput technology in the past decade have generated an enormous amount of data that enables the identification of various types of potential therapeutic targets in microbes and mammalian cells. Identification of therapeutic targets against viruses, however, has lagged far behind much due to the ability of viruses to tolerate a large number of mutations with little to no fitness cost. The most obvious case that exemplifies this difficulty is Human Immunodeficiency Virus, which is capable of escaping all available therapeutics through mutations. Any therapeutic target that is to be effective must remain stable over time and be resistant to accumulating mutations.; This work addresses the problem of therapeutic target stability in viruses, and proposes two computational frameworks for identifying stable ones. We consider two classes of therapeutic targets: conserved viral peptides presented by host class I Major Histocompatibility Complex (MHC) molecules and interfacial regions of protein complexes. We first develop a method that mines large databases of fully sequenced HIV genomes and MHC binding peptides to construct combinations of peptide cocktails that achieve group conservation, where no known HIV isolates have mutations against every component of the cocktail.; Based on the observation that the degree of inter-protein side-chain coevolution is higher than expected by independence, we derive a statistical method for measuring the degree of correlated mutations between proteins and develop a second method, which integrates sequential and structural analyses to identify sites of protein-protein interactions. Finally, using the HIV reverse transcriptase and integrase proteins as an example, we demonstrate how covariation analysis and rigid body docking can infer the protein interfaces and combining sites. |
