Metabolic Footprinting-Based DNA-AuNP Encoders for Extracellular Metabolic Response Profiling.

Metabolic footprinting as a convenient and non-invasive cell metabolomics strategy relies on monitoring the whole extracellular metabolic process. It covers nutrient consumption and metabolite secretion of in vitro cell culture, which is hindered by low universality owing to pre-treatment of the cell medium and special equipment. Here, we report the design and a variety of applicability, for quantifying extracellular metabolism, of fluorescently labeled single-stranded DNA (ssDNA)-AuNP encoders, whose multi-modal signal response is triggered by extracellular metabolites. We constructed metabolic response profiling of cells by detecting extracellular metabolites in different tumor cells and drug-induced extracellular metabolites. We further assessed the extracellular metabolism differences using a machine learning algorithm. This metabolic response profiling based on the DNA-AuNP encoder strategy is a powerful complement to metabolic footprinting, which significantly applies potential non-invasive identification of tumor cell heterogeneity.

2D nanomaterial sensing array using machine learning for differential profiling of pathogenic microbial taxonomic identification.

An integrated custom cross-response sensing array has been developed combining the algorithm module's visible machine learning approach for rapid and accurate pathogenic microbial taxonomic identification. The diversified cross-response sensing array consists of two-dimensional nanomaterial (2D-n) with fluorescently labeled single-stranded DNA (ssDNA) as sensing elements to extract a set of differential response profiles for each pathogenic microorganism. By altering the 2D-n and different ssDNA with different sequences, we can form multiple sensing elements. While interacting with microorganisms, the competition between ssDNA and 2D-n leads to the release of ssDNA from 2D-n. The signals are generated from binding force driven by the exfoliation of either ssDNA or 2D-n from the microorganisms. Thus, the signal is distinguished from different ssDNA and 2D-n combinations, differentiating the extracted information and visualizing the recognition process. Fluorescent signals collected from each sensing element at the wavelength around 520 nm are applied to generate a fingerprint. As a proof of concept, we demonstrate that a six-sensing array enables rapid and accurate pathogenic microbial taxonomic identification, including the drug-resistant microorganisms, under a data size of n = 288. We precisely identify microbial with an overall accuracy of 97.9%, which overcomes the big data dependence for identifying recurrent patterns in conventional methods. For each microorganism, the detection concentration is 105 ~ 108 CFU/mL for Escherichia coli, 102 ~ 107 CFU/mL for E. coli-ß, 103 ~ 108 CFU/mL for Staphylococcus aureus, 103 ~ 107 CFU/mL for MRSA, 102 ~ 108 CFU/mL for Pseudomonas aeruginosa, 103 ~ 108 CFU/mL for Enterococcus faecalis, 102 ~ 108 CFU/mL for Klebsiella pneumoniae, and 103 ~ 108 CFU/mL for Candida albicans. Combining the visible machine learning approach, this sensing array provides strategies for precision pathogenic microbial taxonomic identification. • A molecular response differential profiling (MRDP) was established based on custom cross-response sensor array for rapid and accurate recognition and phenotyping common pathogenic microorganism. • Differential response profiling of pathogenic microorganism is derived from the competitive response capacity of 6 sensing elements of the sensor array. Each of these sensing elements' performance has competitive reaction with the microorganism. • MRDP was applied to LDA algorithm and resulted in the classification of 8 microorganisms.

An explainable machine-learning approach for revealing the complex synthesis path-property relationships of nanomaterials.

Machine learning (ML) models have recently shown important advantages in predicting nanomaterial properties, which avoids many trial-and-error explorations. However, complex variables that control the formation of nanomaterials exhibiting the desired properties still need to be better understood owing to the low interpretability of ML models and the lack of detailed mechanism information on nanomaterial properties. In this study, we developed a methodology for accurately predicting multiple synthesis parameter-property relationships of nanomaterials to improve the interpretability of the nanomaterial property mechanism. As a proof-of-concept, we designed glutathione-gold nanoclusters (GSH-AuNCs) exhibiting an appropriate fluorescence quantum yield (QY). First, we conducted 189 experiments and synthesized different GSH-AuNCs by varying the thiol-to-metal molar ratio and reaction temperature and time in reasonable ranges. The fluorescence QY of GSH-AuNCs could be systematically and independently programmed using different experimental parameters. We used limited GSH-AuNC synthesis parameter data to train an extreme gradient boosting regressor model. Moreover, we improved the interpretability of the ML model by combining individual conditional expectation, double-variable partial dependence, and feature interaction network analyses. The interpretability analyses established the relationship between multiple synthesis parameters and fluorescence QYs of GSH-AuNCs. The results represent an essential step towards revealing the complex fluorescence mechanism of thiolated AuNCs. Finally, we constructed a synthesis phase diagram exceeding 6.0 × 104 prediction variables for accurately predicting the fluorescence QY of GSH-AuNCs. A multidimensional synthesis phase diagram was obtained for the fluorescence QY of GSH-AuNCs by searching the synthesis parameter space in the trained ML model. Our methodology is a general and powerful complementary strategy for application in material informatics.

Correction: A machine learning approach-based array sensor for rapidly predicting the mechanisms of action of antibacterial compounds.

Correction for 'A machine learning approach-based array sensor for rapidly predicting the mechanisms of action of antibacterial compounds' by Zhijun Li et al., Nanoscale, 2022, 14, 3087-3096, DOI: 10.1039/D1NR07452K.

A machine learning approach-based array sensor for rapidly predicting the mechanisms of action of antibacterial compounds.

Rapid and accurate identification of the mechanisms of action (MoAs) of antibacterial compounds remains a challenge for the development of antibacterial compounds. Computational inference methods for determining the MoAs of antibacterial compounds have been developed in recent years. In particular, approaches combining machine learning technology enable precisely recognizing the MoA of antibacterial compounds. However, these methods heavily rely on the big data resulting from multiplexed experiments. As such, these approaches tend to produce minimal throughput and are not comprehensive enough to be adapted to widespread industrial applications. Here, we present a machine learning approach based on a customized array sensor for directly identifying the MoAs of antibacterial compounds. The array sensor consists of different two-dimensional nanomaterial fluorescence quenchers with different fluorescence-labeled single-stranded DNAs (ssDNAs). By mapping the subtle difference of the physicochemical properties on the bacterial surface treated with different antibacterial compound stimuli, the array sensor ensures visualizing the recognition process. Moreover, the customized array sensor produces a high volume of the MoA database, overcoming the dependence on big data. We further use the array sensor to build a chemical-response unique "fingerprint" database of MoAs. By combining a neural network-based genetic algorithm (NNGA), we rapidly discriminate the MoAs of four antibiotics with an overall accuracy of 100%. Furthermore, a new screening antibacterial peptide has been discovered and evaluated by our approach for determining the MoA with high accuracy proven by other techniques.