Bacterial infections are the leading cause of illness and death worldwide; a current public health problem exacerbated by slow or inaccurate diagnoses. Now, NIBIB-funded scientists have designed an inexpensive paper-based test that can quickly identify multiple types of bacteria.
The University of Nebraska research team used a complex combination of microbiology, chemistry, and artificial intelligence (AI) to create a deceptively low-tech-looking testing platform designed for use in remote, low-resource settings such as field hospitals and rural clinics.
“We have designed this technology to be extremely sensitive and accurate in identifying bacterial species while also being easy to manufacture,” explained Denis Svechkarev, Ph.D., research assistant professor in the department of pharmaceutical sciences and co-first author of the paper along with graduate student Aayushi Laliwala. “The test is also durable enough to survive shipping to remote locations and simple enough to be easily used by healthcare personnel with limited training and equipment.”
The work is being carried out in the laboratory of Aaron M. Mohs, Ph.D., associate professor in the department of pharmaceutical sciences and senior author of the publication, which appeared in the journal Analytical Chemistry1 on January 24.
The “simple” platform, which is in the development and testing stages in hopes of eventual use in the field, has a complex name: “Paper-based Ratiometric Fluorescent Sensor Array.” About the size of a 3 x 5 card, the paper sensor is “laid out” with a grid of small circles onto which the bacterial samples to be analyzed are applied. The “radiometric fluorescent” part of the name refers to the ingenious way bacteria are identified.
The research team designed and synthesized fluorescent dyes that can “sense” the subtle biochemical differences of each type of bacteria and convert those differences into different fluorescent signals. Four different fluorescent dyes are dried in four circles on the matrix comprising a single test. A bacterial sample, such as E coli, is placed in each of the four circles and the dyes are activated with ultraviolet light, causing each of the four dyes to send five fluorescent signals for a total of 20 fluorescent signals per test.
A fluorescent plate reader scans the 20 fluorescent signals, which vary depending on the interaction of the dyes with the outer membrane of the bacteria. A state-of-the-art artificial intelligence program, in the form of an artificial neural network, was trained to recognize the subtle but specific pattern of fluorescent intensities created by each type of bacteria. The result is a “signature” fluorescent pattern that is transferred from the reader to the artificial neural network program, which identifies the type of bacteria.
In collaboration with microbiologists, Drs. Marat R. Sadykov and Kenneth W. Bayles’ team tested the system using a collection of 16 bacterial species. The system correctly identified all 16 species more than 90% of the time, a level of accuracy that could provide a healthcare worker in the field with valuable information about the specific bacteria in an infected individual, allowing for accurate and rapid antibiotic treatment. The test also determined whether the bacteria was gram-positive or negative with 95% accuracy. The Gram test is a technique that further determines the composition of bacteria and is essential for knowing which types of antibiotics are most effective. The accuracy of the test was extremely promising considering that a few hours of delay in the diagnosis and treatment of an infectious disease dramatically worsens the patient’s prognosis.
Every aspect of the test was designed for potential use in even the most remote parts of the world, where current techniques requiring sophisticated equipment and expertise are not feasible. For example, drying fluorescent dyes on the paper card eliminated the need to use liquid fluorescent dyes that would require refrigeration, often not available in low-resource regions. Photolithography was used to “stamped” the grid of circles onto the paper card, a quick and inexpensive way to make thousands of cards. In tests, conducted by placing the cards in a box in the closet, the cards remained stable for up to 6 months, making them ideal for shipping and distribution to remote areas. The card pattern is identical to the 96-well plates used for many tests that use liquid components, allowing paper cards to be scanned and read using standard available machines.
“This project is an extraordinary example of how making something simple requires the use of multiple complex technologies,” said Tatjana Atanasijevic, Ph.D., (Scientific Program Manager) of the Bioanalytical Sensors program at the National Institute of Biomedical Imaging and Bioengineering (NIBIB), which co-funded the project along with several additional institutes at the National Institutes of Health.
The work is in the research and development stage and the team is testing and refining the system using samples that replicate what would be collected from patients in the field. Future technical feats in the team’s sights include working with engineers to create a system that would allow the 96-point paper card to be read with a simpler device, perhaps even a cell phone camera—admittedly a lofty goal, but doable, Svechkarev explained.
When asked about the work, Mohs credits the extraordinary efforts of Svechkarev and Laliwala. “The technology needed to create this bacteria detection system was devised during the pandemic, when we had limited access to the laboratory. Denis and Aayushi used this time to develop skills that included new computer coding methods, learning how to use different types of artificial intelligence, and finalizing the design of the best fluorescent dyes – all key elements that came together to build this promising diagnostic system.”
This work was supported by grant EB027662 from the National Institute of Biomedical Imaging and Bioengineering, the National Cancer Institute, and the National Institute of Allergy and Infectious Diseases. Additional funding was provided by the Bukey Memorial Fund, the Nebraska Research Initiative, and the University of Nebraska Foundation.
—Thomas Johnson, Ph.D., special to NIBIB
