Taking the leap between analytical chemistry and artificial intelligence: A tutorial review

Tan C.S., Leow S.Y., Ying C., Tan C.J., Yoon T.L., Jingying C., Yam M.F.

• Tri-step FT-IR fingerprints of Chuan-Mutong and Guan-Mutong are presented. • Machine learning classifiers were developed to distinguish between Chuan-Mutong and Guan-Mutong. • Comparison of machine learning classifier with PCA and PLS-DA was performed. • The proposed FT-IR with PLS-DA and machine learning classifier method was characterized by being simple, fast and reliable. Chuan-Mutong ( Clemetis spp.) is a precious medicinal herb in traditional Chinese medicine that possesses various therapeutic effects especially well known for its diuretic effect and widely used in Malaysia. However, there were several reported Chinese herb nephropathy cases due to the adulteration of Aristolochia spp. found in combinational herbal regimen. Guan-Mutong ( Aristolochia manshuriensis ), which looks similar in appearance and has similar therapeutic effects as Chuan-Mutong, has the possibility to substitute the Chuan-Mutong. Therefore, there is a necessity to differentiate the types of Mutong using analytical authentication methods. In this paper, a rapid and accurate method is proposed to discriminate Chuan-Mutong from Guan-Mutong by using tri-step fourier transform infrared spectroscopy (FT-IR) identification approaches. The method involves the deployment of FT-IR, second derivative infrared spectra (SD-IR), and two-dimensional correlation infrared spectra (2D-IR). In our approach, FT-IR spectra of Chuan-Mutong and Guan-Mutong were subjected to discrimination using principal component analysis (PCA), partial least squares discriminant analysis (PLS-DA), and machine learning classifiers (ML). Chuan-Mutong and Guan-Mutong can be clearly classified or discriminated against each other by ML, PLS-DA and PCA. The sensitivity, accuracy and specificity of ML were >90%, while the sensitivity, accuracy and specificity of PLS-DA were 100%. It is hence demonstrated that the infrared spectroscopic identification approach using PCA, PLS-DA and ML can be effectively used to differentiate Chuan-Mutong and Guan-Mutong. PLS-DA and ML provide a simple, fast, and high accuracy prediction to differentiate Chuan-Mutong and Guan-Mutong.

Li H., Mehedi Hassan M., Wang J., Wei W., Zou M., Ouyang Q., Chen Q.

Thiabendazole (TBZ) is extensively used in agriculture to control molds; residue of TBZ may pose a threat to humans. Herein, surface-enhanced Raman spectroscopy (SERS) coupled variable selected regression methods have been proposed as simple and rapid TBZ quantification technique. The nonlinear correlation between the TBZ and SERS data was first diagnosed by augmented partial residual plots method and calculated by runs test. Au@Ag NPs with strong enhancement factor (EF = 4.07 × 106) of Raman signal was used as SERS active material to collect spectra from TBZ. Subsequently, three nonlinear regression models were comparatively investigated and the competitive adaptive reweighted sampling-extreme learning machine (CARS-ELM) achieved a higher correlation coefficient (Rp2 = 0.9406) and the lower root-mean-square-error of prediction (RMSEP = 0.5233 mg/L). Finally, recoveries of TBZ in apple samples were 83.02–93.54% with relative standard deviation (RSD) value

Li L., Rong S., Wang R., Yu S.

• Artificial intelligence (AI) methods in drinking water treatment (DWT) are summarized. • The application potential of deep learning in DWT is highlighted. • The lack of powerful detection facilities and data limits the application of AI in DWT. • A combination of a new instrument and AI to identify unknown compounds is proposed. • The establishment of a macro model of DWT plants based on AI needs further research. Because of its robust autonomous learning and ability to address complex problems, artificial intelligence (AI) has increasingly demonstrated its potential to solve the challenges faced in drinking water treatment (DWT). AI technology provides technical support for the management and operation of DWT processes, which is more efficient than relying solely on human operations. AI-based data analysis and evolutionary learning mechanisms are capable of realizing water quality diagnosis, autonomous decision making and operation process optimization and have the potential to establish a universal process analysis and predictive model platform. This review briefly introduces AI technologies that are widely used in DWT. Moreover, this paper reviews in detail the mature applications and latest discoveries of AI and machine learning technologies in the fields of source water quality, coagulation/flocculation, disinfection and membrane filtration, including source water contaminant monitoring and identification, accurate and efficient prediction of coagulation dosage, analysis of the formation of disinfection by-products and advanced control of membrane fouling. Finally, the challenges facing AI technologies and the issues that need further study are discussed; these challenges can be briefly summarized as a) obtaining more effective characterization data to screen and identify targeted contaminants in the complex background with the assistance of AI technologies and b) establishing a macro intelligence model and decision scheme for entire drinking water treatment plants (DWTPs) to support the management of the water supply system.

Zhang X., Li Y., Tao Y., Wang Y., Xu C., Lu Y.

• IRN, SVM and RF models based on IR spectra of parvalbumin were constructed and compared. • IRN model had the greatest accuracy for recognizing fish pavalbumin (up to 97.3%). • IRN model was based on highly representative featured from IR spectra of the parvalbumin allergen. • IRN model could detect parvalbumin accurately in seafood matrices. • Infrared spectroscopic IRN method was rapid (~20 min) and effective. We have developed a novel approach that involves inception-resnet network (IRN) modeling based on infrared spectroscopy (IR) for rapid and specific detection of the fish allergen parvalbumin. SDS-PAGE and ELISA were used to validate the new method. Through training and learning with parvalbumin IR spectra from 16 fish species, IRN, support vector machine (SVM), and random forest (RF) models were successfully established and compared. The IRN model extracted highly representative features from the IR spectra, leading to high accuracy in recognizing parvalbumin (up to 97.3%) in a variety of seafood matrices. The proposed infrared spectroscopic IRN (IR-IRN) method was rapid (~20 min, cf. ELISA ~4 h) and required minimal expert knowledge for application. Thus, it could be extended for large-scale field screening and identification of parvalbumin or other potential allergens in complex food matrices.

Carter J.A., O'Brien L.M., Harville T., Jones B.T., Donati G.L.

Supervised and unsupervised machine learning methods are used to evaluate matrix effects caused by carbon and easily ionizable elements (EIEs) on analytical signals of inductively coupled plasma optical emission spectrometry (ICP OES). A simple experimental approach was used to produce a series of synthetic solutions with varying levels of matrix complexity. Analytical lines (n = 29), with total line energies ( E sum ) in the 5.0–15.5 eV range, and non-analyte signals (n = 24) were simultaneously monitored throughout the study. Labeled (supervised learning) and unlabeled (unsupervised learning) data on normalized non-analyte signals (from plasma species) were used to train machine learning models to characterize matrix effect severity and predict analyte recoveries. Dimension reduction techniques, including principal component analysis, uniform manifold approximation and projection and t -distributed stochastic neighborhood embedding, were able to provide visual and quantitative representations that correlated well with observed matrix effects on low-energy atomic and high-energy ionic emission lines. Predictive models, including partial least squares regression and generalized linear models fit with the elastic net penalty, were tuned to estimate analyte recovery error when using the external standard calibration method (EC). The best predictive results were found for high-energy ionic analytical lines, e.g. Zn II 202.548 nm ( E sum = 15.5 eV), with accuracy and R 2 of 0.970 and 0.856, respectively. Two certified reference materials (CRMs) were used for method validation. The strategy described here may be used for flagging compromising matrix effects, and complement method validation based on addition/recovery experiments and CRMs analyses. Because the data analysis workflows feature signals from plasma-based species, there is potential for developing instrument software capable of alerting users in real time ( i.e. before data processing) of inaccurate results when using EC. • Machine learning is used to identify plasma-related matrix effects. • Ar, H and O species and machine learning are used to predict analyte recoveries. • PCA, UMAP and t -SNE indicate matrix effect severity while performing calibration. • Supervised learning models are effective for high-energy ionic analytical lines. • Internal standardization with Ar I minimizes matrix effects on low-energy atomic lines.

Zhong S., Zhang K., Wang D., Zhang H.

• MF-ML assisted-QSAR model was developed for 1089 compounds toward HO reactivity. • An ensemble model that combined XGBoost and DNN was developed. • The SHAP method was used to interpret all the obtained models. • The model made predictions based on the chemical knowledge correctly “learned”. Developing quantitative structure-activity relationships (QSARs) is an important approach to predicting the reactivity of HO radicals toward newly emerged organic compounds. As compared with molecular descriptors-based and the group contribution method-based QSARs, a combined molecular fingerprint-machine learning (ML) method can more quickly and accurately develop such models for a growing number of contaminants. However, it is yet unknown whether this method makes predictions by choosing meaningful structural features rather than spurious ones, which is vital for trusting the models. In this study, we developed QSAR models for the log k HO values of 1089 organic compounds in the aqueous phase by two ML algorithms—deep neural networks (DNN) and eXtreme Gradient Boosting (XGBoost), and interpreted the built models by the SHapley Additive exPlanations (SHAP) method. The results showed that for the contribution of a given structural feature to log k HO for different compounds, DNN and XGBoost treated it as a fixed and variable value, respectively. We then developed an ensemble model combining the DNN with XGBoost, which achieved satisfactory predictive performance for all three datasets: Training dataset: R-square ( R 2 ) 0.89–0.91, root-mean-squared-error ( RMSE ) 0.21–0.23, and mean absolute error ( MAE ) 0.15–0.17; Validation dataset: R 2 0.63–0.78, RMSE 0.29–0.32, and MAE 0.21–0.25; and Test dataset: R 2 0.60–0.71, RMSE 0.30–0.35, and MAE 0.23–0.25. The SHAP method was further used to unveil that this ensemble model made predictions on log k HO based on a correct ‘understanding’ of the impact of electron-withdrawing and -donating groups and of the reactive sites in the compounds that can be attacked by HO . This study offered some much-needed mechanistic insights into a ML-assisted environmental task, which are important for evaluating the trustworthiness of the ML-based models, further improving the models for specific applications, and leveraging the implicit knowledge the models carry.

Zhu J., Sharma A.S., Xu J., Xu Y., Jiao T., Ouyang Q., Li H., Chen Q.

In this study, a novel analytical approach is proposed for the identification of pesticide residues in tea by combining surface-enhanced Raman scattering (SERS) with a deep learning method one-dimensional convolutional neural network (1D CNN). First, a handheld Raman spectrometer was used for rapid on-site collection of SERS spectra. Second, the collected SERS spectra were augmented by a data augmentation strategy. Third, based on the augmented SERS spectra, the 1D CNN models were established on the cloud server, and then the trained 1D CNN models were used for subsequent pesticide residue identification analysis. In addition, to investigate the identification performance of the 1D CNN method, four conventional identification methods, including partial least square-discriminant analysis (PLS-DA), k -nearest neighbour ( k −NN), support vector machine (SVM) and random forest (RF), were also developed on the basis of the augmented SERS spectra and applied for pesticide residue identification analysis. The comparative studies show that the 1D CNN method possesses better identification accuracy, stability and sensitivity than the other four conventional identification methods. In conclusion, the proposed novel analytical approach that exploits the advantages of SERS and a deep learning method (1D CNN) is a promising method for rapid on-site identification of pesticide residues in tea. • Rapid identification of pesticide residues in tea by 1D CNN coupled with SERS. • The SERS spectra were rapid on-site collected by a handheld Raman spectrometer. • PLS-DA, k -NN, SVM, RF and 1D CNN were comparatively studied in identification. • 1D CNN showed superior performance in terms of accuracy, stability and sensitivity.

Vrzal T., Malečková M., Olšovská J.

Retention index in gas chromatographic analyses is an essential tool for appropriate analyte identification. Currently, many libraries providing retention indices for a huge number of compounds on distinct stationary phase chemistries are available. However, situation could be complicated in the case of unknown unknowns not present in such libraries. The importance of identification of these compounds have risen together with a rapidly expanding interest in non-targeted analyses in the last decade. Therefore, precise in silico computation/prediction of retention indices based on a suggested molecular structure will be highly appreciated in such situations. On this basis, a predictive model based on deep learning was developed and presented in this paper. It is designed for user-friendly and accurate prediction of retention indices of compounds in gas chromatography with the semi-standard non-polar stationary phase. Simplified Molecular Input Entry System (SMILES) is used as the model’s input. Architecture of the model consists of 2D-convolutional layers, together with batch normalization, max pooling, dropout, and three residual connections. The model reaches median absolute error of prediction of the retention index for validation and test set at 16.4 and 16.0 units, respectively. Median percentage error is lower than or equal to 0.81% in the case of all mentioned data sets. Finally, the DeepReI model is presented in R package, and is available on https://github.com/TomasVrzal/DeepReI together with a user-friendly graphical user interface. • Advanced model for retention indices prediction of compounds in GC was developed. • The model is based on a convolutional neural network and advanced approaches. • Median percentage error of prediction is ≤ 0.81%. • The model is publicly available in the R package - DeepReI.

Duan Q., Xu Z., Zheng S., Chen J., Feng Y., Run L., Lee J.

Determination of complex pollutants often involves many high-cost and laborious operations. Today's pop machine-learning (ML) technology has exhibited their amazing successes in image recognition, drug designing, disease detection, natural language understanding, etc. ML-driven samples testing will inevitably promote the development of related subjects and fields, but the biggest challenge ahead for this process is how to provide some intelligible and sufficient data for various algorithms. In this work, we present a full strategy for rapid detecting mixed pollutants through the synergistic application of holographic spectrum and convolutional neural network (CNN). The results have shown that a well-trained CNN model could realize quantitative analysis of the mixed pollutants by extracting spectral information of matters, suggesting the strategy's value in facilitating the study of complex chemical systems.

Cady N.C., Tokranova N., Minor A., Nikvand N., Strle K., Lee W.T., Page W., Guignon E., Pilar A., Gibson G.N.

The 2019 SARS CoV-2 (COVID-19) pandemic has illustrated the need for rapid and accurate diagnostic tests. In this work, a multiplexed grating-coupled fluorescent plasmonics (GC-FP) biosensor platform was used to rapidly and accurately measure antibodies against COVID-19 in human blood serum and dried blood spot samples. The GC-FP platform measures antibody-antigen binding interactions for multiple targets in a single sample, and has 100% selectivity and sensitivity (n = 23) when measuring serum IgG levels against three COVID-19 antigens (spike S1, spike S1S2, and the nucleocapsid protein). The GC-FP platform yielded a quantitative, linear response for serum samples diluted to as low as 1:1600 dilution. Test results were highly correlated with two commercial COVID-19 antibody tests, including an enzyme linked immunosorbent assay (ELISA) and a Luminex-based microsphere immunoassay. To demonstrate test efficacy with other sample matrices, dried blood spot samples (n = 63) were obtained and evaluated with GC-FP, yielding 100% selectivity and 86.7% sensitivity for diagnosing prior COVID-19 infection. The test was also evaluated for detection of multiple immunoglobulin isotypes, with successful detection of IgM, IgG and IgA antibody-antigen interactions. Last, a machine learning approach was developed to accurately score patient samples for prior COVID-19 infection, using antibody binding data for all three COVID-19 antigens used in the test. • GC-FP biosensor yields rapid (30 min) detection of antibodies against COVID-19 infection. • Antibody detection is quantitative and can provide antibody titer information. • Diagnosis of prior COVID-19 infection using serum with 100% selectivity and sensitivity. • Multiplexed approach yields information on multiple antibody-antigen interactions. • Dried blood spot testing significantly simplifies sample collection and yields high sensitivity/specificity.

Ghaffari M., Omidikia N., Ruckebusch C.

An approach is proposed and illustrated for the joint selection of essential samples and essential variables of a data matrix in the frame of spectral unmixing. These essential features carry the signals required to linearly recover all the information available in the rows and columns of a data set. Working with hyperspectral images, this approach translates into the selection of essential spectral pixels (ESPs) and essential spatial variables (ESVs). This results in a highly-reduced data set, the benefits of which can be minimized computational effort, meticulous data mining, easier model building as well as better problem understanding or interpretation. Working with both simulated and real data, we show that (i) reduction rates of over 99% can be typically obtained, (ii) multivariate curve resolution - alternating least squares (MCR-ALS) can be easily applied on the reduced data sets and (iii) the full distribution maps and spectral profiles can be readily obtained from the reduced profiles and the reduced data sets (without using the full data matrix).

Rodrigues V.C., Soares J.C., Soares A.C., Braz D.C., Melendez M.E., Ribas L.C., Scabini L.F., Bruno O.M., Carvalho A.L., Reis R.M., Sanfelice R.C., Oliveira O.N.

The development of simple detection methods aimed at widespread screening and testing is crucial for many infections and diseases, including prostate cancer where early diagnosis increases the chances of cure considerably. In this paper, we report on genosensors with different detection principles for a prostate cancer specific DNA sequence (PCA3). The genosensors were made with carbon printed electrodes or quartz coated with layer-by-layer (LbL) films containing gold nanoparticles and chondroitin sulfate and a layer of a complementary DNA sequence (PCA3 probe). The highest sensitivity was reached with electrochemical impedance spectroscopy with the detection limit of 83 pM in solutions of PCA3, while the limits of detection were 2000 pM and 900 pM for cyclic voltammetry and UV-vis spectroscopy, respectively. That detection could be performed with an optical method is encouraging, as one may envisage extending it to colorimetric tests. Since the morphology of sensing units is known to be affected in detection experiments, we applied machine learning algorithms to classify scanning electron microscopy images of the genosensors and managed to distinguish those exposed to PCA3-containing solutions from control measurements with an accuracy of 99.9%. The performance in distinguishing each individual PCA3 concentration in a multiclass task was lower, with an accuracy of 88.3%, which means that further developments in image analysis are required for this innovative approach.

Gomes J.C., Souza L.C., Oliveira L.C.

Surface plasmon resonance (SPR) based sensors allow the evaluation of aqueous and gaseous solutions from real-time measurements of molecular interactions. The reliability of the response generated by a SPR sensor must be guaranteed, especially in substance detection, diagnoses, and other routine applications since poorly handled samples, instrumentation noise features, or even molecular tampering manipulations can lead to wrong interpretations. This work investigates the use of different machine learning (ML) techniques to deal with these issues, and aim to improve and attest to the quality of the real-time SPR responses so-called sensorgrams. A new strategy to describe a SPR-sensorgram is shown. The results of the proposed ML-approach allow the creation of intelligent SPR sensors to give a safe, reliable, and auditable analysis of sensorgram responses. Our arrangement can be embedded in an Intelligence Module that can classify sensorgrams and identify the substances presents in it. Also made it possible to order and analyze interest areas of sensorgrams, standardizing data, and supporting eventual audit procedures. With those intelligence features, the new generation of SPR-intelligent biosensors is qualifying to perform automated testing. A properly protocol for Leishmaniasis diagnosis with SPR was used to verify this new feature.

Draz M.S., Vasan A., Muthupandian A., Kanakasabapathy M.K., Thirumalaraju P., Sreeram A., Krishnakumar S., Yogesh V., Lin W., Yu X.G., Chung R.T., Shafiee H.

A virus detection method using deep learning–based analysis of smartphone-recorded microchip images without any optical hardware.

Callaway E.

Google’s deep-learning program for determining the 3D shapes of proteins stands to transform biology, say scientists. Google’s deep-learning program for determining the 3D shapes of proteins stands to transform biology, say scientists.

Meher A.K., Zarouri A.

Green analytical chemistry represents a transformative approach to analytical science, emphasizing sustainability and environmental stewardship while maintaining high standards of accuracy and precision. This review highlights recent innovations in green analytical chemistry, including the use of green solvents, such as water, supercritical carbon dioxide, ionic liquids, and bio-based alternatives, as well as energy-efficient techniques like microwave-assisted, ultrasound-assisted, and photo-induced processes. Advances in green instrumentation, including miniaturized and portable devices, and the integration of automation and chemometric tools, have further enhanced efficiency and reduced the environmental footprint of analytical workflows. Despite these advancements, challenges remain, including the need to balance analytical performance with eco-friendliness and the lack of global standards to measure and promote sustainable practices consistently. However, the future of green analytical chemistry looks promising, with emerging technologies like artificial intelligence and digital tools offering new ways to optimize workflows, minimize waste, and streamline analytical processes. By focusing on these areas, green analytical chemistry is transforming analytical methodologies into tools that not only achieve high performance but also align with global sustainability goals. This review underscores how green analytical chemistry is more than just a scientific discipline, but a pathway for reducing the ecological impact of analytical processes while driving innovation in science and industry. With the continued commitment to research, collaboration, and the adoption of cutting-edge technologies, green analytical chemistry has the potential to shape a greener and more sustainable future for analytical chemistry and its diverse applications.

Li D., Chen Q., Ouyang Q., Liu Z.

Cuadros-Rodríguez L., Arroyo-Cerezo A., Jiménez-Carvelo A.M.

Pervaiz W., Afzal M.H., Feng N., Peng X., Chen Y.

Zhong A., Li Z., Song Y.

Ziani I., Bouakline H., El Guerraf A., El Bachiri A., Marie-Laure Fauconnier, Sher F.

Fan M., Brolo A.G.

Ramos M.C., Collison C.J., White A.D.

This review examines the roles of large language models (LLMs) and autonomous agents in chemistry, exploring advancements in molecule design, property prediction, and synthesis automation.

Salamat Q., Gumus Z.P., Soylak M.

Jiménez-Carvelo A.M., Arroyo-Cerezo A., Cuadros-Rodríguez L.

Environmentally friendly analytical methods are being developed and implemented, avoiding the use of chemicals and solvents as far as possible, and reducing the time and cost of analysis. For an analytical method to be considered as part of green analytical chemistry, it must meet certain requirements. Therein lies the importance of having tools at one’s disposal with which to assess how sustainable an analytical method is according to its intended purpose, in this case food quality assurance. The scientific community is advancing by leaps and bounds in the development of these new analytical methods. The techniques used that can be considered as green analytical techniques are becoming more and more sophisticated. This is a challenge for both industry and analysts, as the results produced by these techniques require the application of artificial intelligence tools. This also makes it possible to generate multivariate analytical methods through the development of machine learning models. This chapter looks at sustainable practices in the field of analytical food chemistry from a generic perspective and highlights current trends in the field.

Alemayhu A.S., Ji R., Abdalla A.N., Bian H.

Höfler L.

Liu Y., Ping M., Han J., Cheng X., Qin H., Wang W.

As a kind of long-term favorable device, the microelectromechanical system (MEMS) sensor has become a powerful dominator in the detection applications of commercial and industrial areas. There have been a series of mature solutions to address the possible issues in device design, optimization, fabrication, and output processing. The recent involvement of neural networks (NNs) has provided a new paradigm for the development of MEMS sensors and greatly accelerated the research cycle of high-performance devices. In this paper, we present an overview of the progress, applications, and prospects of NN methods in the development of MEMS sensors. The superiority of leveraging NN methods in structural design, device fabrication, and output compensation/calibration is reviewed and discussed to illustrate how NNs have reformed the development of MEMS sensors. Relevant issues in the usage of NNs, such as available models, dataset construction, and parameter optimization, are presented. Many application scenarios have demonstrated that NN methods can enhance the speed of predicting device performance, rapidly generate device-on-demand solutions, and establish more accurate calibration and compensation models. Along with the improvement in research efficiency, there are also several critical challenges that need further exploration in this area.

Pandiselvam R., Aydar A.Y., Aksoylu Özbek Z., Sözeri Atik D., Süfer Ö., Taşkin B., Olum E., Ramniwas S., Rustagi S., Cozzolino D.

Santos J.L., Barrenechea Bueno G.M., Moraes Flores É.L., Ogava L.E., de Souza Dias F., Leite O.D.