By Eric Martin
It is a trend that the World Health Organization predicts could see AMR become the leading cause of death in the next 20 years.
Perth researchers, led by Associate Professor Anthony Kicic from the Wal-yan Respiratory Research Centre at Telethon Kids Institute and Curtin School of Population Health, have been using AI to explore the medical potential of bacteriophages as an alternative means of treating bacterial infection, harnessing machine-learning to sift through the organisms’ unique genetic structure and find the best, individualised match for pathogens.
“We’re on the doorstep of a similar scenario to COVID now in terms of global need,” Professor Kicic said. “We all need to take this quite seriously because there’s dire consequences: in less than five years, infections we can currently treat might be killing people again because we don’t have effective treatment options for them anymore.
“It is a scary thing, yet it’s not on everyone’s priority list. Shining the light back onto bacteriophage, these superhero viruses, is fantastic. It may add another treatment option for clinicians.”
Ironically, bacteriophages were first identified as a potential treatment more than 100 years ago, in 1915, and were the subject of investigation until the sudden discovery of penicillin relegated ‘phages’ to the largely forgotten pages of medical history.
“It’s one of those pieces of history that has never found recognition because of everyone’s embracement of antibiotics. They were developed at the same time and still work but their voice has just been lost in time,” Professor Kicic said.
“However, their importance is clearly shown by the fact that they outnumber any other class of biological creatures on earth and are among the most numerous groups in the human virome.”
Which is where the computational power of AI is helping to identify the correct phage in a fraction of the current time.
“For example, we have about 3000 bacteriophages against a particular bacterial species called Pseudomonas aeruginosa, and when we have patient X come in with a particular infection type, we have to try and match them,” Professor Kicic said.
“At the moment it’s ad hoc and is not an efficient process. You start from phage number one and go right through and see which one might be active against this bacterial species. We can look at some characteristics to provide clues as to which ones we think will work, but it’s really a guessing game.
“So, the goal was to address the bottlenecks of time-impacting processes. We wanted to pull this information that we’re manually gathering now, feed it into a machine-learning algorithm and have it come up with a predictive list that will identify the most likely candidates that will be active against a bacterium.
“We could prioritise our screening process against those the machine would find and have this turned over in a very time-efficient manner. Currently, it takes two to three weeks to screen about 100 phages. So, if you’ve got a thousand, that can take a lot of time that a patient may not have.”
Professor Kicic said with philanthropic and State government funding, a specific machine-learning tool was being developed that would also continue to build on the whole genome sequencing data they were amassing, both locally and nationally.
“The system is always improving itself and uses the genetic sequences of both the bacteriophage and the bacteria as well, very much like bioinformatics, to search for the answers in that genetic code,” he said.
“The screening and the genetic sequencing of the phage will identify whether it’s a lytic versus a temperate phage, and whether the bacterial host has any prophage as well. We have also screened our library to only look at those that will be therapeutically applicable, as well as contributing to a nationwide network collective called Phage Australia.
“All the data will then feed into this algorithm to increase its accuracy and diagnostic capability, and be rolled out across Australia, and hopefully internationally as well.
“We already have colleagues at Westmead who have started successfully treating patients. But while they’re a little bit more advanced than us in the application, their genome sequencing and matching is still conducted on a manual basis.
“One of the main benefits is we’ve all had our own specific research interests, and we all bring repositories of bacteriophage here in Australia to the table. We also hope to have the ability to share manufactured phage and treat everyone on a national level.”
Professor Kicic said Phage Australia’s international connections were also very diverse, enabling them to bring new phage into Australia, which was otherwise challenging as live biologics.
“What we would like to do in countries that don’t have the ability to generate their own phage libraries and algorithms is to provide that matching service for them.
“If they supply us either with the genome sequences of the bacteria and the phage that they’re trying to test, or in fact, ask us, ‘here’s the bug that we’ve got a problem with, do you have any phage with it?’ Then we can actually get that manufactured to a medicinal level.”
Professor Kicic said the other big challenge was that the FDA had recognised phage therapy one way, while the European Therapeutics Council viewed it slightly differently. The TGA here in Australia still hasn’t decided because it’s so new, it’s still not recognised as a therapeutic.
“As such, the national network is providing the data that we’re generating to the TGA so they can make a recognised decision that will then determine how we prepare these phage for human use at a medicinal level,” he said.
“We also feel that being on the doorstep of Asia, which has the highest density population in the world, including numerous megacities, it is likely that they – and we – are going to see significant outbreaks of AMR in the near future.”
Professor Kicic explained that his background as a researcher into respiratory conditions, particularly cystic fibrosis where those impacted tend to contract more viral, fungal and bacterial lung infections, set him on the search to find effective alternatives to antibiotics.
“Maintaining lung health is a top priority for CF communities. Repeated lung infections lead them to being hospitalised and having an 8-12-week course of antibiotics to try and eradicate these bugs, and obviously, over a 30-40 year life span, you can often see that by their early 20s, they are starting to experience AMR,” he said.
“We work in close partnership with the community, and they’ve raised it as a continuing issue, noting certain problematic bacteria in their adult populations that are extremely difficult to treat. That opened my eyes to the AMR crisis that we are facing globally.
“Even pharmaceutical companies are not investing in traditional antibiotics anymore. It’s not beneficial. They’re all off patent. There’s no money to be made in it. We need to preserve the antibiotics that we’ve got available and look at new therapeutics, totally different to antibiotics, but which have the same effect.”
“There will be a 15-20-year translational pipeline to make them readily available for the general public,” he said. “That’s not going to help the current situation where we want to make an impact to this 20-year prediction of it being the leading cause of human death.
“I needed to look at something that was going to be able to be implemented in clinics and help our doctors treat patients in the next two to five years.
“Bacteriophage therapy was something that our CF community was aware of in terms of technology. Several Australian patients had been travelling overseas to be treated and the question was posed, ‘why aren’t we doing this here in Australia?’
“As a result, I received a small grant to start a phage repository here in Perth against most of the bacteria that cause real problems for people with CF and begin the exploration process using AI.
“AI has been adapted and implemented in many ways across medical research and assisting diagnostics, such as increasing the sensitivity and detection of cancer cells during screening. Even though it’s not mainstream yet in the clinic, the proof of principle is there.”
While the inclusion of pseudomonas, staphylococcus, and several other pathogens was respiratory related, Professor Kicic soon realised that his primary targets also caused infections at other sites.
“For example, pseudomonas and golden staph can cause lung infections, but they also can cause skin infections and numerous complications after surgery, so you can imagine the potential,” he said.
“However, we’ve got to be careful how we use it, it’s not a golden bullet all the time. Once you have the correct phage that is active against the bacteria, you need to hit it hard, fast, and repeatedly to counter the active ability of the bacteria to develop resistance.”
Horizontal gene transfer, a common trait associated with phage, can render an originally suitable bacteriophage suddenly ineffective for therapeutic treatment.
“Once they start tolerating the bacterial genome, they become what we call temperate phage which, if internalised by the bacteria, will just integrate with their genome host and grant the first signs of resistance,” he said.
“As such, where we really place bacteriophage therapy currently is only when a patient identifies themselves as starting to develop problematic AMR. But eventually, we want to implement further upstream. Bacteriophage therapy is going to be prescribed as opposed to a course of antibiotics when you see your GP.
“We’re now at a point, some five years later, where even post pandemic, we’ve got the largest repository of bacteriophage against several prominent bacteria – the escaped pathogens – that are on the WHO priority list in Australia.
“We have national ethics approval through our collaborators to start treating patients on a compassionate use basis where there is critical need, no therapeutic alternatives, and they don’t have the luxury of time.
“And finally, we have a small-scale manufacturing facility here in Perth where we can manufacture phage to the medicinal quality that we need.”
