A medicinal drug can be created from any combination of millions of chemical compounds. Experimenting with all of these combinations can be costly when trying to develop a new drug—but MechSE associate professor Srinivasa Salapaka in collaboration with ISE Professor Carolyn Beck have simplified the problem using tools and concepts from statistical physics and statistics in what is called combinatorial library design.
Written by Meredith Staub
A medicinal drug can be created from any combination of millions of chemical compounds. Experimenting with all of these combinations can be costly when trying to develop a new drug—but MechSE associate professor Srinivasa Salapaka in collaboration with ISE Professor Carolyn Beck have simplified the problem using tools and concepts from statistical physics and statistics in what is called combinatorial library design.
Combinatorial drug discovery is a computational approach to developing potential pharmaceuticals. In order to find a drug to treat a certain disease—to kill certain types of cancer cells, for example—researchers will try many different combinations of compounds against the disease and measure their effectiveness. If they are effective, they are selected as potential drug candidates and tested further.
Combinatorial library methods were first developed in the 1980s, and were originally only applied to peptides and oligonucleotides. With new computing technologies, libraries have been expanded to include proteins, synthetic oligomers, small molecules, and oligosaccharides. Larger, more diverse, and more complex libraries increase the chance that researchers will find compounds with significant therapeutic value.
The problem is that there are so many compounds, doing experiments with all of them is extremely costly. But some compounds are exceptionally similar, at least in several qualities, to groups of other compounds. In this way, a compound can be experimentally representative of its entire library. Selecting them is called lead compound selection.
"This is numerically a very hard problem," Salapaka said. "It takes a very large number of computations. So we developed some algorithms to do such a selection process. The compounds that we choose are representative of their entire library; that is, we are making sure that all of the properties of the compounds in the original library are represented."
Organizing compounds by library is dependent on the compounds' properties. All of the compounds in a certain library will have about the same value for all of the desired properties of the potential drug. So on a graph of two values, for example, the compounds in a certain library can be seen in a cluster around their values for those two properties.
"You choose one representative from each of these clusters to represent the entire library," Salapaka said, "because doing an experiment with one of these compounds is the same as doing the experiment with any of the compounds in that cluster. It saves a lot of time and money."
Considering that combinatorial libraries typically consist of several million compounds, this system is the only way that makes combinatorial chemistry techniques even remotely practical. But its ability to create molecules and test them en masse could revolutionize drug discovery.