Does the “fingerprint” technique identify different bacteria in seconds?

Schematics of the general Raman data collection and analysis process where a single spectrum is obtained from a single cell and classified via deep learning. 1 credit

Bacterial identification can take hours and often longer, valuable time to diagnose infections and select appropriate treatments. There may be a faster and more accurate process according to KAIST researchers. By teaching a deep learning algorithm to identify “fingerprint” spectra of the molecular components of various bacteria, researchers were able to classify various bacteria in different media with accuracies of up to 98%.

Their results have been published in Biosensors and bioelectronicsahead of publication in the April issue of the journal.

Bacteria-induced diseases, those caused by direct bacterial infection or exposure to bacterial toxins, can induce painful symptoms and even lead to death. The rapid detection of bacteria is therefore crucial for preventing the consumption of contaminated food and for diagnosing infections from clinical samples. , like urine.

“Using surface-enhanced Raman spectroscopy (SERS) analysis enhanced by a newly proposed deep learning model, we have demonstrated a particularly simple, fast, and efficient way to classify the signals of two common bacteria and their backgrounds. residents without any separation procedure,” said Sungho Jo from the computer school.

Raman spectroscopy sends light through a sample to see how it scatters. The results reveal structural information about the sample – the spectral fingerprint – allowing researchers to identify its molecules. The surface-enhanced version places sample cells on noble metal nanostructures that help amplify sample signals.

However, it is difficult to obtain consistent and clear spectra of bacteria due to many sources of overlapping peaks, such as proteins in cell walls.

“In addition, strong signals from surrounding media are also enhanced to overwhelm target signals, requiring time-consuming and tedious bacterial separation steps,” said Yeon Sik Jung from the Department of Materials Science and Engineering.

To analyze the noisy signals, the researchers implemented an artificial intelligence method called deep learning that can hierarchically extract certain features from the spectral information to classify the data. They specifically designed their model, named Dual Branch Wide Core Network (DualWKNet), to efficiently learn the correlation between spectral features. Such capability is essential for analyzing one-dimensional spectral data, according to Jo.

“Despite spurious signals or noise from the media, which render the general shapes of different bacterial spectra and their resident media signals similar, high classification accuracies of bacterial types and their media were obtained,” said Jo, explaining that DualWKNet allowed the team to identify key peaks in each class that were nearly indistinguishable in the individual spectra, thereby improving classification accuracy. “Ultimately, with the use of DualWKNet replacing the bacteria and media separation steps, our method significantly reduces analysis time.”

The researchers plan to use their platform to study more bacteria and media types, using the information to create a library of training data for various types of bacteria in additional media to reduce collection times and detection of new samples.

“We have developed a meaningful universal platform for rapid bacterial detection with the collaboration between SERS and deep learning,” Jo said. “We hope to expand the use of our deep learning-based SERS analysis platform to detect many types of bacteria in other media important for food or clinical analysis, such as blood.”

Republished with kind permission from KAIST.

Comments are closed.