This month, we hosted two PHAVES events; here’s your recap of both!

PHAVES 2: Dr. Elyse Stachler describes her work combining disinfectants with phage

Youtube video of PHAVES 2 (34:23)

Elyse Stachler began her talk by explaining the scope of the problem of hospital-acquired infections. She focused on Pseudomonas aeruginosa and the surface-associated biofilms it often creates, especially on plastics and other surfaces found in clinical settings (like catheters, trachial tubes, etc). She noted that due to the COVID crisis, hospital-acquired bacterial infections are becoming even more common and dangerous.

Elyse set out to pursue the question of how the microbiome of the built environment might be controlled, with a focus on phages. Her unique background as an engineer, and current affiliation at Switzerland’s national water research agency, Eawag, have undoubtedly put her in a great position to delve into these questions.

Elyse’s main questions were whether phages could be effectively combined with disinfectants (like bleach), which are already used to decontaminate hospital surfaces, whether the combination helped, and whether the order mattered. She surmised that if bleach could inactivate phages, this could have the added benefit of letting phages do their work, while also having them be conveniently inactivated after they’d finished (thus limiting the release of active phages into the environment).

She studied two phages, one model phage and one freshly isolated from wastewater, and looked at how they were able to affect P. aeruginosa biofilms in combination with sodium hypochlorite (bleach) in vitro, and she varied the order of application. She found that the two were slightly synergistic, and that biofilms were most effectively reduced if phages were added first. She hypothesized that this might be because the sodium hypochlorite negatively impacted the host bacteria, making the phage infection process less efficient due to sickly hosts. She went on to show that her phages could be partly inactivated by chlorine, suggesting that this too could be an explanation for why phages worked better when added before the disinfectant instead of after.

In the future, Elyse wants to study multispecies biofilms and look at more realistic biofilm conditions. She also made a note that phages do not all react the same way to disinfectants, so many more phage-disinfectant combinations should be examined in the future.

More about Elyse

To keep up with Elyse, follow her on Twitter at @ElysianTweets, or find her on Phage Directory.

PHAVES 3: Piotr Tynecki talks phage AI

Youtube video of PHAVES 3 (57:48)

Piotr Tynecki started his talk by introducing the recent paper by Jean-Paul Pirnay that laid out an imagined future for phage therapy by 2035, where AI might one day predict phage-host interactions and help power intelligent phage design to treat any infection. Piotr, a PhD student studying machine learning who stumbled into the world of phage, feels that we’re already moving quickly toward this future. However, he emphasized that its success depends on a strong foundation of data that is shared by the phage community in a standardized, structured way.

Piotr first talked about what AI actually is, how it’s not new (e.g. Neflix, Spotify, etc. already rely on AI). He said it’s not even new to the phage world, because many bioinformatics tools already exist that let phage researchers make predictions based on AI. He walked through the basics about how AI works, and we learned about which parts of AI actually involve the input of human domain experts.

Piotr then described a platform that he co-created with Proteon Pharmaceuticals, and how it can already be used to predict phage lifestyle. He talked about how the algorithm initially worked great for lytic and temperate phages, until he heard from the community that it was failing for phages with lifestyles outside these two canonical types. When his team later updated the tool to account for chronic lifestyles, this vastly improved predictions. He impressed upon us that phage researcher feedback was clearly key to developing algorithms that work.

The second part of Piotr’s talk covered how his team’s platform makes phage taxonomy predictions. He said a recent switch from nucleotide to amino acid-based predictions was incredibly useful, and since there are so many more amino acid sequences matched to phage taxa (37 million!) compared to phage sequences with known lifestyles (low thousands), predicting phage taxonomy has been much more accurate than predicting lifestyle. However, predictions are not so reliable at the level of the genus yet (often there is only one phage genome per genus, as opposed to at least 20 examples of each at the family level). Again, the importance of researchers contributing phage data to central platforms that aggregate and structure it was emphasized.

The third section of Piotr’s talk covered phage.ai, the platform he and his company, SLAVIC AI, co-created with Proteon Pharmaceuticals, and how it can be used today. They built the free, open source platform based on 10,000 publicly available phage genomes. For instance, a user can input a phage genome and get a prediction of its lifestyle, or get a prediction on how different their phage is from all other phages.

Piotr concluded by saying that “the future of phage AI will be unsupervised”, meaning that with enough reliable, aggregated data, we’ll be able to make predictions about phage features without relying on expert-defined rules. He shared what he’s particularly excited about: eventually developing algorithms that can predict what kind of genetic engineering would make a phage more stable, more efficient at infecting, or otherwise “boost” its activity. First step: getting our data in order!

Want more?

• If you’d be interested in participating in a series of practical data science webinars dedicated to phage research, please join Phage Directory’s Slack, and let us know in the #phage-ai channel!

Papers to check out

• PhageAI - Bacteriophage Life Cycle Recognition with Machine Learning and Natural Language Processing, by Piotr Tynecki et al. (BioRxiv, 2020)
• ProtTrans: Towards Cracking the Language of Life’s Code Through Self-Supervised Deep Learning and High Performance Computing, by Ahmed Elnaggar et al. (arXiv, 2020)
• Modeling aspects of the language of life through transfer-learning protein sequences, by Michael Heinzinger et al. (BMC Bioinformatics, 2019)

More about Piotr

To keep up with Piotr, follow him on Twitter at @ptynecki, or find him on Phage Directory.

Phage Directory is now on YouTube!

Subscribe to Phage Directory’s YouTube channel to keep up to date with future PHAVES recordings!