Rather than wait for bacteria to develop resistance to new drugs, researchers have created a computer program that puts them on the offense. The algorithm, explained in a study published last month in the Proceedings of the National Academy of Sciences, can predict the next move of harmful bacteria — before it happens.
Talk about developing a program that can predict the next move of bacteria.
There are many mechanisms by which organisms become resistant [to antibiotics, for example] and many of them are very fancy, devilishly clever. One focus of our work has been trying to develop technology, particularly algorithms and software, that could help in the battle against resistance to antibiotics.
There are many ways that bacteria become resistant. One of the ways is as follows: The worker molecules in a cell are proteins. They might perform a step in a pathway to synthesizing important molecules for the bacteria. Many drugs work by trying to block that interaction in the pathogen. One way bacteria evade drugs is they evolve to make changes to the target proteins. [They can evolve mutations] that still allow the native function [of the bacteria] to happen and should block the association of the drug molecule.
How can we deal with these things? There are many approaches. If you have a new drug, you can perform experiments to see how [bacteria] might evolve in order to evade them. It’s time-consuming and expensive. [It's more common to] take your best guess, go through clinical trials, deploy the antibiotics and wait until the resistance mutations surface. You’ll find people resistant to these [drugs]. You have to find out why they’re resistant, isolate the proteins that have mutated in the bugs and the drug design starts again. It’s kind of like an arms race. We have a catalog of ways that proteins have become resistant in the past.
My lab specializes in protein design. We said, Why can’t we use the protein design algorithms to predict ahead of time — without any catalog? We do a positive design and a negative design. In the positive design, we redesign the protein in MRSA to still do its native function, but with different mutations. The negative design is to disable the inhibitor, so it doesn’t bind. You intersect those two lists. If you find mutations that are predicted to destabilize the inhibitor and make viable the native function of the protein, then those are good candidate mutants. If you made those proteins, [they] would exhibit the properties the pathogen wants — not binding the inhibitor and still doing its native function.
Our algorithm works by predicting the structures of the mutants [the bacteria is] thinking about. The structure of the mutant [in the lab] was very close to the structure predicted by our algorithm. We really didn’t know ahead of time how the proteins could change in order to evade a drug.
Drug design is exactly where we want to go. Resistance is everywhere. Because our algorithm is based on predicting interactions between a protein and an inhibitor and also stabilizing a native function, we think it could be applied for different kinds of bacterial resistance. Other possibilities would be viral resistance. Great examples would be HIV, influenza, herpes. It’s also possible, in principle, to model some of the types of resistance that cancer cells will evolve [to] anti-cancer therapeutics. We think we have a chance of predicting resistance ahead of time. Knowledge of that potential resistance could be used early in the drug design process.
What motivates you to do this work?
On a personal level, we see resistance to antibiotics emerging around us. I have a colleague whose wife acquired invasive MRSA. My kids are very healthy [but after taking antibiotics as babies], I’m often told that now we’ll have to give them something stronger. I have a collaborator who is allergic to a lot of antibiotics, so there’s a need to develop more. It seems very real when you know people who could benefit.
In general, there has been less of a rush to develop new antibiotics in the pharmaceutical industry, I think because of profit issues. People typically take antibiotics for a relatively short time, they go generic fairly quickly and it might be hard to make money off that. It’s an interesting niche for an academic lab [to be] a participant in the search for possible drug leads. Because we think this is important we’re releasing [our software] on open source for any researcher who wants to use it. This kind of capability is too important to be locked up in a laboratory.
Image, top: Visualization of the 3-D crystal structure of the top-scoring DHFR protein mutant, determined from experimental data. DHFR is a likely molecule for attack by a drug to stop MRSA bacteria infections. / Courtesy of Bruce Donald
Image, bottom: Bruce Donald