Biomedical engineers at Duke University have shown that different strains of the same bacterial pathogen can be distinguished by machine learning analysis of their growth dynamics only, which in turn can also accurately predict other characteristics such as antibiotic resistance. The demonstration may refer to methods for identifying diseases and predicting their behavior that are faster, simpler, less expensive and more accurate than current standard techniques.
The results will be published on the internet on the 3rd August, 30th April, 2009 Proceedings of the National Academy of Sciences.
For most of the history of microbiology, bacterial identification has relied on growing cultures and analyzes the characteristics and physical behavior of the resulting bacterial colony. It wasn’t until recently that scientists could simply run a genetic test.
Genetic sequencing, however, is not universally available and can often be time consuming. And even with the ability to sequence entire genomes, it can be difficult to link specific genetic variations to different behaviors in the real world.
For example, even though researchers know the genetic mutations that help protect bacteria from beta-lactam antibiotics ̵1; the most commonly used antibiotic in the world – sometimes DNA is not the whole story. While one resistant bacterium usually cannot survive a dose of antibiotics alone, large populations can often.
Lingchong Int, a professor of biomedical engineering at Duke, and his graduate student, Carolyn Zhang, wondered if a new twist on older methods might work better. Perhaps they can magnify one specific physical trait and use it not only to identify the pathogen, but to make educated testimony about other traits such as antibiotic resistance.
“We thought that small variation in genes between strains of bacteria could have a subtle effect on their metabolism,” you said. “But because bacterial growth is exponential, that subtle effect can be amplified enough for us to take advantage of. To me, that concept is somewhat intuitive, but I was surprised at how well it worked.”
How fast a bacterial culture grows in a laboratory depends on the richness of the medium growing in it and in its chemical environment. But as the population grows, the culture consumes nutrients and produces chemical by-products. Even though different breeds start with the exact same environmental conditions, subtle differences in how they grow and influence the surrounding environment accumulate over time.
In the study, Int and Zhang took more than 200 strains of bacterial pathogens, many of which were variations of E. coli, placed them in identical growth environments, and carefully measured their population density so as it increased. Due to their slight genetic differences, cultures have grown in attention and in stages, each having a unique pattern of temporal shifts. The researchers then entered the growth dynamics data into a machine learning program, which taught itself to identify and compare growth profiles with different varieties.
To their surprise, it worked really well.
“Using growth data from only one initial condition, the model was able to identify a particular strain with more than 92 percent accuracy,” you said. “And when we used four different starting environments instead of one, that accuracy went up to about 98 percent.”
Taking this idea one step further, You and Zhang then looked to see if they could use dynamic growth profiles to predict another phenotype – antibiotic resistance.
The researchers again loaded a machine learning program with dynamic growth profiles from all varieties except one of the various strains, along with data on their resilience to four different antibiotics. They then tested to see if the resulting model could predict the resistance of the final strain antibiotics from its growth profile. To pass on their data collection, they repeated this process for all other breeds.
The results showed that the dynamic growth profile alone can successfully predict strain resistance to antibiotics 60 to 75 percent of the time.
“This is actually the same or better than some of the current techniques in the literature, including many using genetic sequencing data,” Int said. “And that was just proof of principle. We believe that with higher-resolution data of growth dynamics, we can do an even better job in the long run.”
The researchers also looked to see if the breeds showing similar growth curves had similar genetic profiles. As it turns out, the two are completely uncorrelated, showing again how difficult it is to identify cellular characteristics and behavior for specific DNA bands.
Moving forward, You are planning to optimize the growth curve procedure to reduce the time it takes to identify a strain from 2 to 3 days to maybe 12 hours. He also plans to use high-definition cameras to see if mapping how colonies of bacteria grow in space in a petri dish can help make the process more accurate.
A quarter of bacterial pathogens can spread antibiotic resistance directly to their peers
Carolyn Zhang et al, Temporal coding of bacterial identity and characteristics in growth dynamics, Proceedings of the National Academy of Sciences (2020). DOI: 10.1073 / pnas.2008807117
Provided by Duke University
Citation: AI can offer better way for ID of drug-resistant superbugs (2020, 4 August) obtained on 6 August 2020 from https://phys.org/news/2020-08- ai-id-drug-resistant-superbugs.html
This document is subject to copyright. Apart from any fair business for the purpose of private study or research, no part may be reproduced without written permission. The content is provided for information purposes only.