Antimicrobial resistance (AMR) – the ability of microorganisms to resist antimicrobial treatments, especially antibiotics – has a direct impact on human health and carries a heavy economic burden due to higher costs of treatments and reduced productivity caused by sickness. AMR is responsible for an estimated 25,000 deaths per year in the EU and 700,000 worldwide. A report commissioned by the UK government in 2014 predict that by 2050, antimicrobial resistant infections will kill 10 million people across the world – more than the current toll from cancer. It is also estimated that AMR costs the EU 1.5 billion EUR per year in healthcare costs and productivity losses.
This situation calls for new and innovative approaches to treat multidrug-resistant infections. One of the most promising alternatives to antibiotics is phage therapy – i.e. the use of viruses (bacteriophages) to specifically infect and kill bacteria during their life cycle. Phages or bacteriophages have the advantage of being extremely specific, much more than most antibiotics. For instance they do not kill beneficial bacteria as found in Gut flora. A phage is be able to infect and kill only one bacterial strain. Therefore, the practice of phage therapy requires to know precisely the bacterial strain responsible for the infection and the phage or phages that can kill it.
Nevertheless, associate the right bacteriophage to a bacteria is a major challenge in clinic applications. Current solutions to discover which phage attacks a specific bacteria are performed in the lab by trial and error using cultures of the phage and bacteria in Petri dishes. This method have the following main drawbacks:
- It is a long and expensive process that can take several days; this time may be lethal for many patients with aggressive bacteria.
- Search for the right phage is not easy. Usually that research must be done through networks of scientists working with phages. And, in the more extreme case, some scientists use as last resource social media.
- In top of all that, sometimes is required to test not one but several phages (cocktails) at the same time, taking the problem to a combinatorial explosion of the number of experiments to be perform.
Our company Phages4A (Phages For All) has developed a unique workflow (including data selection, curation, feature selection and modeling) able to predict in an in-silico way the phage or phages that can specifically infect and kill a bacteria. Our technology harnesses the genome of phages and bacteria and relies on bacterial-phage interaction data to train artificial intelligence algorithms. The models thus developed make it possible to select the phage that can specifically infect and kill a given bacteria. Our predictions will help to create custom bacteriophage cocktails to treat patients with bacterial infections. With our solution, it will no longer be necessary to wait for a laboratory to test all possible phages.
We had published the following peer reviewed papers describing our approach :
“Computational Prediction of Host-Pathogen Interactions Through Omics Data Analysis and Machine
Learning”. Leite et al. IWBBIO (2) 2017: 360-371
“Computational prediction of inter-species relationships through omics data analysis and machine learning.”
Leite et al. BMC Bioinformatics 2018 19 (Suppl 14) :420 .
“Exploration of multiclass and one-class learning methods for prediction of phage-bacteria interaction at strain level.”
Leite et al. IEEE International Conference on Bioinformatics and Biomedicine (BIBM) 2018