Almost every biological process that occurs in the human body depends on the interaction of proteins with other molecules. To learn how the human body functions at a molecular level, structural biologists have determined the atomic structure of tens of thousands of proteins.

The active sites for many interactions between proteins and other molecules are often located within a cavity on the protein surface. The shape of a cavity enables proteins to selectively bind the molecules with which they interact while avoiding those they should avoid.

Precisely evolved structural and chemical complementarity enables a diversity of molecular signaling pathways to operate effectively without interfering with each other. For the same reasons, specially designed molecules capable of blocking a particular protein cavity are often used as drugs to combat disease by preventing unwanted interactions. Designing these molecules is a core effort in the pharmaceutical industry.

Brian Chen, assistant professor of computer science and engineering, is developing software that uses solid volumetric representations of protein cavities. His approach enables an automated deconstructive analysis of the similarities and variations in cavity shape that affect selective binding.

In a paper published recently in PLOS Computational Biology, Chen and coauthor Barry Honig of Columbia University showed that this software could isolate the amino acids and cavity regions that drive differences in binding preferences among serine proteases, a well-studied family of digestive proteins. The two researchers also demonstrated the success of their software with START domains, an important family of lipid transporters.

“This is just a very early proof of concept,” says Chen. “Hopefully, this technology will help us understand the causes of preferential binding in a wider range of molecular interactions and one day improve the efficiency of drug design.”