Most automated approaches to drug discovery are based on testing thousands of compounds against a single disease target. The platform developed by Serenex, a discovery company in Durham, North Carolina, takes the opposite approach by testing drug candidates against thousands of protein targets simultaneously. This technique, which the firm calls proteome mining, promises to deliver unique information about interactions between compounds and targets.
“In contrast to profiling compounds against a single target, what we're doing is profiling one compound against all targets in an entire sub-proteome,” says Steven Hall, the company's senior vice-president of R&D. “We get extremely broad selectivity and information.”
Serenex is focusing on the purine-binding proteins within the human proteome. This group of some 2,000 proteins includes protein kinases and many other clinically important proteins that use purine nucleotides such as adenosine triphosphate as co-substrates or co-factors. A potential drug candidate can be tested against all 2,000 in a single step.
The approach depends on a patented affinity medium of small beads to which many molecules of a purine nucleotide such as ATP are tethered. When a cell or tissue extract is mixed with this medium, the tethered molecules ensnare its purine-binding proteins. Unbound proteins are then washed away, and a drug compound introduced. If the compound has an affinity for any of the bound proteins, it will displace them from the tethered purines. A selective drug will displace just one protein, whereas less-selective compounds will displace many different proteins. The displaced proteins can then be collected and identified by mass spectrometry and bioinformatics.
The aim is to help medicinal chemists make a more informed decision about which compounds to move forward to lead optimization. Larger-scale screening of large compound libraries against sub-proteome sets of proteins will provide a very useful and information-rich matrix of compound–target interactions, Hall notes.
“We feel that we will get to the point where we can develop some fundamental understandings of what kind of substructures will lead to effects on what type of target,” he says. “It's only by screening large libraries of targets against large numbers of compounds that the informatics community will be able to get the connection between structure and protein affinity.”