Water Quality Monitoring by Bacterial Biosensors

 

News Article Link: Water Quality Monitoring by Bacterial Biosensors - Science Connected Magazine

Research Paper Link: Decoding the metabolic response of Escherichia coli for sensing trace heavy metals in water | PNAS

The news article I chose for my blog post was titled “Water Quality Monitoring by Bacterial Biosensors”. It was published in the Science Connected Magazine on March 20, 2023. The author of the article is Rachel Calder, who has a master’s degree focused on ecology. The primary focus of the article is on a novel approach of using Escherichia Coli’s metabolic stress response to detect the concentration of heavy metals in water samples. While this concept in general isn’t particularly novel, researchers have now utilized a combination of surface-enhanced Raman spectroscopy (SERS) and machine learning to accurately measure the concentration of metals such as Arsenic (As3+) and chromium (Cr6+) in wastewater and tap water. Specifically, the news article draws a large portion of its content from the research paper “Decoding the metabolic response of Escherichia coli for sensing trace heavy metals in water”, published on December 28, 2022, by Regina Ragan et al.

Overall, I would say that the article from the magazine was very faithful to the findings discussed in the original research. Particularly, I found the strongest point of the article to be its concise way of describing the methodology used to observe tiny spectral changes for various concentrations of As3+ and Cr6+. Furthermore, I appreciated how the article also emphasized that—while the metabolite concentration changes from E. Coli is the basis for heavy metal detection—Professor Ragan and her team dedicated much effort into coming up with an algorithm that can detect spectroscopy differences too small for the human eye or brain to comprehend. I also appreciated how the news article gave context to why it was important to focus on heavy metal pollution in water, as they are abundant due to fossil fuel burning and the use of agrochemicals. The article does not exactly go too much into how machine learning was used to tell apart one concentration from the other. However, I found this is to the article’s benefit, as the jargon used in the research paper could have turned some readers away from learning more.

To give a more detailed look at the research, Professor Ragan and her team worked off the idea that concentrations of metabolites and nucleotide molecules such as ATP will vary depending on environmental stress. After exposure to the contaminated water samples, these E. Coli cultures would be lysed and their metabolites would be analyzed on SERS surface, comprised of gold nanoparticles. This was conducted for both bank control samples and varying concentrations of both As3+ and Cr6+. Ragan et al. was also able to use the technique of principal component analysis to highlight the features of different spectral datasets because of different metal concentrations. From there, data augmentation was performed to balance the datasets. Specifically, the algorithm developed was also capable of differentiating As3+ from Cr6+. Finally, a Convolutional Neural Network (CNN) regression model was developed to fit varying spectral datasets to their corresponding concentrations. The more datasets they included, the better the regression model was at predicting the concentrations. The algorithm was also fine-tuned to predict concentrations of heavy metals for water from unknown sources. This was accomplished by spiking contaminants into water samples.

Unfortunately, I was more mixed on how the news article conveyed the results of the study. The author accurately stated that—using this new method—the limit of detection for As3+ and Cr6+ was 100,000 lower than the World Health Organization’s recommended limits. She even succeeds in conveying why it's important to have lower limits of detection, as it gives us a chance to detect significant changes in metal concentration before adverse side effects arise. However, I was rather confused about the author’s decision to write a segment about E. Coli’s safety. For something that was barely covered in the study, the safety of using E. Coli takes up a rather substantial portion of the article. Furthermore, I found it somewhat misleading for the article to say that this method achieved greater than 98% accuracy at the limit of detection. In the study, it is said that the limit of detection was determined with an accuracy of >98%. However—even in the abstract—Ragan et al. makes it clear that the accuracy of determining heavy metal concentrations in drinking water is 96%. That value decreases once a more complex water sample is used, with the exact accuracy being about 92%. The 98% value only applies to the accuracy at which the model can distinguish the sample from the control dataset.

In the end, I would rate the news article an 8/10. Calder does give additional context and generally avoids distorting the original study for a more sensational story. However, there are some areas where I feel more focus should have been devoted to. Overall, there is nothing major that’s holding the article down. It even does a surprisingly good job of explaining the methodologies of the original study without getting bogged down by technical jargon—something that other news articles often fail at.