Decoding the metabolic response of Escherichia coli for sensing trace heavy metals in water

Abstract

Heavy metal contamination due to industrial and agricultural waste represents a growing threat to water supplies. Frequent and widespread monitoring for toxic metals in drinking and agricultural water sources is necessary to prevent their accumulation in humans, plants, and animals, which results in disease and environmental damage. Here, the metabolic stress response of bacteria is used to report the presence of heavy metal ions in water by transducing ions into chemical signals that can be fingerprinted using machine learning analysis of vibrational spectra. Surface-enhanced Raman scattering surfaces amplify chemical signals from bacterial lysate and rapidly generate large, reproducible datasets needed for machine learning algorithms to decode the complex spectral data. Classification and regression algorithms achieve limits of detection of 0.5 pM for As 3+ and 6.8 pM for Cr 6+ , 100,000 times lower than the World Health Organization recommended limits, and accurately quantify concentrations of analytes across six orders of magnitude, enabling early warning of rising contaminant levels. Trained algorithms are generalizable across water samples with different impurities; water quality of tap water and wastewater was evaluated with 92% accuracy.

Artificial Intelligence for Surface‐Enhanced Raman Spectroscopy

“…ResNet, [ 115] ANN, [ 116] CNN, [117][118][119][120][121] PCA [ 122] Quantify the abundance of certain molecules RF, [44] PCA+LR, [123] ANN, [ 47,[123][124][125][126] CNN, [127][128][129][130] PCA+SVM, [ 131] PLS, [124,125] SVR, [124] PLS+GA, [132] SVM [ 133] Discover the multiplexed variation in the whole profile PCA, [ 134] Autoencoder, [ 135] CNN, [ 136,137] PCA+LDA [138] Early disease diagnosis ResNet, [ 115] RF, [139] KNN, [ 139] naïve Bayes, [ 139] PLS+SVM [140] SERS spectrum with microRNAs Early disease diagnosis RF, [ 141] LR, [141] naïve Bayes [ 141] Covariance matrices of SERS spectrum…”

“…As for the ML model is usually trained upon the dataset collected from one specific detection system, which can be extended by transfer learning to other systems without additional large dataset. As the result, the built model can be applied to a wider range of molecules [128] under different background conditions, [129,133] gearing toward broad applications (Figure 6b). Unfortunately, in the complex mixtures, the detectability and the discriminability of the target molecules may be varied (unpredictable increase or reduction) by other coexisting molecules due to competitive adsorption [44] and spectral feature overlapping, [47,283] causing unreliable quantification.…”

“…c) Quantification is to determine the detailed relative or absolute abundance of certain target molecules. Reproduced with permission [133]. Copyright 2023, PNAS.…”

Metabolite-Mediated Bacterial Antibiotic Resistance Revealed by Surface-Enhanced Raman Spectroscopy

“…Surface-enhanced Raman spectroscopy (SERS) has emerged as a promising technique for metabolite detection due to its high sensitivity, real-time response, fingerprint information, and minimal sample manipulation. − SERS spectra exhibit fingerprint peak features that enable metabolite identification and offer the potential for multiplex metabolomic analyses. , SERS peak intensities can be used for metabolite quantification and monitoring of bacterial metabolic activity. The monitoring of changes in metabolic SERS profiles allows the evaluation of metabolite transformation and the study of bacterial responses to environmental changes such as nutrient deprivation, oxidative pressure, heavy metal exposure, and viral infection. ,− Further, SERS substrates can be designed for on-site, field deployment. − SERS maps generated through area or depth scans enable the evaluation of metabolite spatial distributions and the study of interspecies interactions. − …”

Modernized Machine Learning Approach to Illuminate Enzyme Immobilization for Biocatalysis

“…Machine learning has exhibited remarkable performance in analyzing vast spectroscopic data from biological samples with complex backgrounds. − Specifically, non-negative matrix factorization (NMF) is an unsupervised machine learning technique that is frequently employed for unmixing pure, individual chemical species present within mixtures. − NMF operates by factorizing an original non-negative data matrix into both non-negative basis vectors and corresponding weights via multiplicative algorithms that minimize the norm of the difference matrix between the original data matrix and its approximate reconstruction. − NMF is advantageous over other decomposition methods because the resolved components reflect true spectra and provide direct chemical information about a mixture rather than variance-based indirect loadings. However, selecting the appropriate number of components to construct an NMF model while avoiding over- or underfitting remains a challenge.…”