#### Faculty Sponsor

Christopher M. Summa

#### College(s)

College of Sciences

#### Submission Type

Oral Presentation

#### Description

Calculation of the relative binding free energies of drug compounds to their cognate proteins is an extremely computationally intensive process, with each calculation taking perhaps weeks of processor time. With large datasets, where free energies need to be calculated for hundreds of compounds within a database of drug targets, minimizing the number of free energy calculations becomes an important undertaking. Toward this end, we present a graph-theoretic approach to planning these calculations, modeled as the optimization of a fully connected graph where the nodes represent the drug compounds in the dataset, and the edges represent a free energy calculation that must be performed. The edges are weighted such that compounds that are structurally and chemically similar have a value close to 1.0, whereas compounds that are highly divergent have a weight approaching 0.0. We show that our technique , implemented in Python using the Networkx library from Los Alamos National Laboratory, significantly outperforms other approaches for optimizing the graph. Also, while our technique is well suited to this particular problem is it more generally applicable for optimizing a wide array of graph models in computational chemistry and bioinformatics.

#### Creative Commons License

This work is licensed under a Creative Commons Attribution-Noncommercial-Share Alike 3.0 License.

Binding Free Energy Calculation Planning Using Graph Theoretic Techniques

Calculation of the relative binding free energies of drug compounds to their cognate proteins is an extremely computationally intensive process, with each calculation taking perhaps weeks of processor time. With large datasets, where free energies need to be calculated for hundreds of compounds within a database of drug targets, minimizing the number of free energy calculations becomes an important undertaking. Toward this end, we present a graph-theoretic approach to planning these calculations, modeled as the optimization of a fully connected graph where the nodes represent the drug compounds in the dataset, and the edges represent a free energy calculation that must be performed. The edges are weighted such that compounds that are structurally and chemically similar have a value close to 1.0, whereas compounds that are highly divergent have a weight approaching 0.0. We show that our technique , implemented in Python using the Networkx library from Los Alamos National Laboratory, significantly outperforms other approaches for optimizing the graph. Also, while our technique is well suited to this particular problem is it more generally applicable for optimizing a wide array of graph models in computational chemistry and bioinformatics.

## Comments

1st place, Oral Presentation, College of Sciences