Â鶹ÒùÔº

Every protein in the body is encased in a water shell that directs protein structure, provides vital stability and steers function. Because of this, water molecules represent a powerful but largely underappreciated foothold in drug binding studies. Yet structural data about these water networks, usually collected at freezing temperatures, often carry temperature-based structural artifacts.

St. Jude Children's Research Hospital scientists have unveiled a new computational tool called ColdBrew to address this problem. The tool leverages data on extensive protein water networks to predict the likelihood of water molecule positions within experimental protein structures, potentially aiding drug discovery efforts. ColdBrew was today in Nature Methods.

Proteins have evolved to fold precisely according to the repulsion and attraction of their amino acid building blocks to water. Water is also key to their activity since it helps guide other molecules, including drug molecules, to bind effectively.

Drug discovery efforts based on protein structures use techniques such as X-ray crystallography and cryo-electron microscopy, but these techniques use freezing, or "cryogenic" temperatures, which can distort how water molecules appear. Marcus Fischer, Ph.D., St. Jude Department of Chemical Biology & Therapeutics, recognized this as a missed opportunity.

"Water molecules in protein structures have so many degrees of freedom that drug discoverers typically throw them out. They're kind of inconvenient," Fischer, corresponding author on the study, said.

With ColdBrew, seeing is believing

To put this lost information to work, Fischer and first author Justin Seffernick, Ph.D., St. Jude Department of Chemical Biology & Therapeutics, developed ColdBrew. "Our goal was to make a tool that's easy to use and understand," said Seffernick. "For each water molecule, our method can tell us how likely water is to be present at higher temperatures. We also found that this same metric can give us clues about how ligands bind to proteins."

This is particularly important to drug discovery. "When ligands bind to proteins, they kick out water from binding sites, so we need to pay attention to them in ligand design," said Fischer. "Encouragingly, we've seen in our data that our predictions were best within these binding sites and around ligands."

Considering that cryogenic structure-solving techniques can artificially increase the number of water molecules present in a structure, a tool such as ColdBrew can assure researchers that seeing is believing. To this end, Fischer and Seffernick have amassed and made publicly available a comprehensive library based on ColdBrew calculations.

"To enable the wide use of ColdBrew, we performed calculations on every structure that fit our criteria in the entire Protein Data Bank. We have over 100,000 predictions, which is over 46 million water molecules," Fischer said. "Remarkably, our results show that drug designers unknowingly avoid tightly bound waters, so actually knowing which ones to avoid could guide the process."

The precalculated data sets are available for download .