DescriptionCharacterizing the specificity of proteases is important to illuminate their role as signaling moieties in a range of diverse biological processes. Proteases often display multispecificity, which is the ability of a single receptor protein molecule to interact with multiple substrates. The ability to accurately recapitulate protease specificity profiles would aid in the design of custom proteases designed to cleave targets in biotechnology or therapeutic scenarios. Current specificity prediction methods use machine - learning techniques that are not generalizable and relatively slow, and thus limited in use for prediction and especially design of multispecificity. We tackle these challenges using a two - pronged approach - by increasing the accuracy of scoring for biophysical protease substrate models as well as by hastening the process of sampling. We develop a general approach for prediction of protease specificity through the construction of high - resolution atomic models, using protein structure modeling and biophysical energetic evaluation of enzyme substrate complexes. Specifically, we develop a discriminatory scoring function using enzyme design modules from Rosetta and Amber-MMPBSA. Analysis of structural models provides physical insight into the structural basis for the observed specificities. We further test the predictive capability of the model by designing and experimentally characterizing the cleavage of four novel substrate motifs for the Hepatitis C virus NS3/4A protease using an in vivo assay. The presented structure-based approach is generalizable to other protease enzymes with known or modeled structures, and complements existing experimental methods for specificity determination. To improve our sampling approach, we develop a rapid, flexible-backbone self-consistent mean field theory-based technique, MFPred, for multispecificity modeling at protein-peptide interfaces. We benchmark our method by predicting experimentally determined peptide specificity profiles for a range of receptors. Our approach should enable the design of a wide range of altered receptor proteins with programmed multispecificities. Viral systems encoding proteases are exemplars of multispecificity. Multispecific proteases mediate the precise cleavage of the polyprotein during replication and viral assembly. The HCV NS3/4A protease is a multispecific protease, which is likely a result of both positive selection pressure to maintain cleavability of its four native substrates, i.e. known sites on the polyprotein, and negative selection pressure to avoid cleavage of other sites in the polyprotein. We map the specificity landscape of the HCV NS3/4A protease to obtain a comprehensive understanding of the protease–substrate interaction network. Using an in vivo yeast surface display assay, Fluorescence Assisted Cell Sorting, Next Generation Sequencing technology and computational modeling using Rosetta and Amber packages, we were able to reconstruct the entire (3.2 million sequences) HCV NS3/4A substrate landscape learning from the sequences identified in our experiment, using an SVM based approach. The work discussed in this thesis gives us insight into the biophysical basis of protease specificity. This work can further be used in rational design of custom proteases and in understanding the mechanisms underlying co-evolution of protease substrate interactions in viral proteases, as well as robustness of the interaction.