The research interests of the Livesay Lab are in the broad areas of bioinformatics and computational biology. Specifically, we use a combination of the two approaches to investigate complex molecular phenomena. Currently, our three main research efforts are: (a) protein functional site prediction, (b) elucidating stability/flexibility relationships within protein families, and (c) development of new bioinformatic algorithms. Taken together, these synergistic investigations are providing us the technical and scientific background for our broad goal of deciphering sequence/structure/function relationships with protein families.

Protein functional site prediction. A significant portion of our research is focused on development and testing of protein functional site prediction algorithms. One of the key outcomes of this work was development of a new Phylogenetic Motif-based functional site prediction approach. Current work is focused on extending the PM method to improve both sensitivity and specificity. In addition, we are very interested in rigorously benchmarking all (both alignment and structure-based) commonly used functional site prediction algorithms. A critical aspect of this undertaking is development of a robust Catalog of Important Sites protein benchmark that includes descriptions of many different types of functional sites (i.e., catalytic sites, active sites, ligand-binding sites, structure sites, allosteric sites, etc.). Once the CIS is complete, we will use it to finally quantitatively assess how well (or how poorly!) current functional site prediction algorithms preferentially predict specific functional site classes.

Improved multiple sequence alignment strategies. Along with Usman Roshan (New Jersey Institute of Technology), we are working on several methods to improve current multiple sequence alignment strategies. One alignment approach that we are investigating is the generalized tree alignment problem, which attempts to indentify the most parsimonious alignment/phylogeny combination. A second approach is to combine the maximum expected accuracy approach of Probcons with a statistical mechanical approach of determining posterior probabilities. The resulting approach, called Probalign, performs statistically significantly better than leading alignment programs (Probcons, MAFFT and MUSCLE) on the BAliBASE, HOMSTRAD and OXBENCH benchmarks.

Quantified stability/flexibility relationships in proteins. Along with Don Jacobs (UNC-Charlotte), we are developing a powerful Distance Constraint Model (DCM) to harmoniously calculate protein stability and flexibility metrics. The DCM is based on a rigorous free energy decomposition scheme representing structure as fluctuating constraint topologies. Entropy non-additivity is problematic for naive decompositions, limiting the success of heat capacity predictions. The DCM resolves non-additivity by summing over independent entropic components determined by an efficient network-rigidity algorithm. Additionally, free energy landscapes and quantitative stability/flexibility relationships (QSFR) are obtained within tractable compute timescales. Currently, we are using the calculated QSFR metrics to assess how well stability, flexibility, and various mechanical linkage properties (allostery) are conserved across protein families. Future work will attempt to design proteins with specific allosteric responses based on targeted QSFR characteristics.

Molecular dynamics simulations. Recently, we (the labs of Dr. Livesay and Dr. Jacobs) have undertaken several investigations based upon molecular dynamics simulations. The first is using MD simulations to help decipher intrinsic fluorescence experiments on the ligand binding domain of the androgen receptor. A second application is focused on using multiple snapshots from an MD simulation as a way of improving the input into the DCM. Finally, in collaboration with Andriy Baumketner (UNC-Charlotte) we are currently initiating a direct comparison of the predictions from the DCM and Go-like models.

Poisson-Boltzmann electrostatics. In the past, much of our work has been in the broad area of protein electrostatics, specifically we have tried to determined how protein surface electrostatics are related to thermostability. For example, we have demonstrated that conserved electrostatic properties is often used by evolution to maintain function across protein family and superfamily divergence. Similarly, we have used calculated electrostatic properties to better understand the functional roles of sites predicted from alternate bioinformatic methods. Future work will assess the ability of electrostatic calculations to reliably predict protein functional sites.