**4. Applications of Electrochemical Biosensors**

#### *4.1. Food Industry*

The maintenance of quality and safety of food is one of the major issues of the food industry. Traditionally, spectroscopic and chemical methods have been used to test the safety and quality of food. The traditional methods are laborious, time-consuming, and costly. The biosensors act as an excellent alternative to the traditional methods of food monitoring. The biosensors are efficient, selective, have a fast response time, and are cost-effective [90]. A biosensor made of cobalt phthalocyanine was successfully used to monitor the ageing of beer during storage [91]. E. coli is a pathogen whose presence in the vegetable is an indication of food contamination [92,93]. The electrochemical biosensor (potentiometric) can be used to detect its presence in the vegetables by simply monitoring the pH variation caused by its presence [94]. To detect the presence of organophosphate pesticides in milk, an enzymatic biosensor made of a screen-printed carbon electrode can be used effectively [95]. Artificial sweeteners are used widely in food products and are the reason for various diseases such as diabetes, dental problems, heart diseases, etc. The traditional methods used to detect sweeteners in food products require expertise and a lot of time. The electrochemical biosensors are the new alternatives that can effectively detect artificial sweeteners in food. The signal produced during the electrochemical analysis is analyzed for the presence of artificial sweeteners such as cyclamate and saccharin via

MATLAB. The biosensors find huge applications in the food industries where fermentation processes are used. In China, about 90% of the industries use biosensors for monitoring the fermentation processes. Maintaining the quality and safety of the food product during fermentation is extremely important. The biosensors can monitor the condition of the process and the presence of enzymes and biomass in the product. The glutaminase-based electrochemical biosensor chip has proved to be effective in monitoring the fermentation process in various industries and factories. The biosensors can be controlled automatically and are again cost-effective and highly efficient. These can effectively monitor lactate, ethanol, glucose, etc. in the food products during fermentation and act as an indicator for stopping and resuming the fermentation process. The biosensors have gained huge interest in recent years due to their quick and effective monitoring of the fermentation processes [96]. The glucose content in a food product can get altered during the storage process and it gives a direct indication of the quality of the food product [97]. Thus, monitoring the glucose content helps in determining the safety and best-before conditions of a food product. The electrochemical biosensors are the most effective and commonly used biosensors for the monitoring of glucose levels in food products as well as in the human body [98]. The electrochemical biosensors can effectively detect even minute concentrations of harmful metals such as arsenic, lead, cadmium, etc. These have been successfully tested to detect paraoxon, aldicarb, carbaryl, and pesticides in food products [99–102].

#### *4.2. Medical Sciences*

The biosensors and their application in detecting the glucose level of diabetic patients are growing very rapidly as they cover 80% of the global biosensors used at home [103,104]. The electrochemical biosensors have been used in the detection of various infectious diseases in the human body such as urinary tract infection, identifying pathogens, microbial bodies, etc. An electrochemical biosensor based on hafnium oxide (HfO2) has been developed to detect the infection that occurs after implantation in the human body [105,106]. Cardiovascular diseases and heart failure are the growing diseases in the world leading to the death of millions of people globally [107–111]. The electrochemical biosensors that are effective in detecting such heart failures are also incorporated into digital watches and bands. These watches and bands are efficient and cost-effective and are very useful in saving lives. The electrochemical biosensors based on fluorescence have been used to monitor the level of enzymes in cancer patients. Such biosensors detect the presence of a particular analyte and produce the fluorescent signal which can be detected and measured [112,113]. Such biosensors prove very effective in the earlier detection of the diseases such as inflammation, arthritis, cancer, viral infections, heart-related problems, and metastasis. The electrochemical biosensors are an important part of the drug discovery program. They are used to monitor the working of a drug and are effective in both preand post-clinical evaluation [114–118]. Recently, electrochemical biosensors have been successfully used for guiding surgery using imaging techniques and also monitoring the impact of a drug on the disease [119].

#### *4.3. Defence*

In the current era, biological attacks or warfare is something that we all are aware of and biological warfare agents (BWAs) such as bacteria, virus, toxins, etc. are used in such warfare. The electrochemical biosensors can be effectively used to detect BWAs with high sensitivity and selectivity. The DNA sequencing, monitoring their activity and metabolism, enzymatic action, etc. are some of the principles of such electrochemical biosensor. A potentiometric biosensor was successfully developed to detect botulinum toxin, having an LOD of 10 ng/mL [120]. Gold nanoparticles were combined with magnetic nanoparticles to develop an electrode that can be used to detect the mecA gene, which is a biomarker for methicillin-resistant S. aureus (MRSA). Quantum dots of CdTe were coupled with nanoparticles of silica, and were then used to detect the Epstein-Barr virus via electrochemical methods with an LOD of 1 pg/mL [121]. For detecting Listeria monocytogenes in

food samples, a screen-printed carbon electrode was modified using gold nanoparticles. The amperometry result showed an LOD of 2 log CFU/mL [122]. Similarly, *salmonella* spp. [123] and M. tuberculosis [124] have been detected using electrochemical biosensors with high accuracy and sensitivity. The human papillomavirus (HPV), which is related to cervical cancer in humans, has been detected using an electrochemical nucleic acid biosensor [125]. Currently, viruses are being detected using enzyme-linked immunosorbent assay (ELISA) related to viral antigens.

#### *4.4. Metabolic Engineering and Plant Biology*

The protection of the environment and less dependence on petroleum-driven products are the latest concern on the global scale to cope with global warming. Researchers around the world are searching for products that will be efficient and eco-friendly. Metabolic engineering is one such field where microorganisms are used to produce chemicals, fuels, and pharmaceuticals. Metabolic engineering is an important step towards sustainable development. The biosensors are important in metabolic engineering as they can monitor the metabolism process and help in the controlled production of chemicals and fuels. The imaging and sequencing of DNA have revolutionized the field of plant science. Traditionally, spectroscopic methods were used to monitor the enzymes, receptors, transporters, and substrates. The development of biosensors helps in monitoring such process and is also fast and effective. To monitor and control the level of calcium in live cells, a protein sensor was developed by Roger Tsien's lab [126–128]. FRET biosensors can monitor sucrose and sugar levels during phloem loading-sucrose efflux and the effect of glucose on yeast cells [129,130]. The electrochemical biosensors helped to monitor the effect of pH level on a plant species and play a vital role in genetic engineering [131,132].

## **5. Machine Learning for Biosensors**

#### *5.1. Improvement in Biosensor by ML*

Firstly, for specimen or complicated matrices, large sensing data can be efficiently processed by machine learning. Secondly, the gain of ML in biosensors comprises the probability of getting sensible analytical results from disorderly and low-resolution sensing data which could closely overlie on one another. Furthermore, appropriate use of ML methods may find unseen relationships between signals of sensing and parameters of specimen via the visualization of data and interrelations between bioagents and signals. Particularly, raw sensing data can be analyzed by using ML from a biosensor in different ways: Categorization, anomaly detection, noise reduction, and pattern recognition. Based on the target analyte, the algorithms aids to categorize the sensing signals in different manners. It is also observed that the operating conditions inevitably affects the performance of a biosensor. On-site usage of biosensor generally interferes with contamination. In that case, ML plays a very important role in checking the quality of the signal. Because of interferences and biofouling in real samples, the variations in sensor performance can be improved by using ML. It is also observed that sensing signals always contain noise. Hence, it is very important to train to develop the model of ML which extracts the good quality signal from the signal containing noise. Finally, the interpretation of sensing data occurs effectively and easily by developing the patterns and latent objects using ML algorithms [13]. For on-site diagnosis or detection, the ML can be significantly important to aid biosensors that can read out rapidly, accurately, automatically, and directly. Instead of predicting the model for electrochemical biosensor, the optical imaging method assisted by a convolutional neural network (CNN) was also developed to calculate the diagnostic consequences [14]. On the other hand, the pathology workforce takes thirty seconds to interpret the image. Additionally, for designing the desirable biosensors nowadays, ML has been preferred. Metamaterials with negative permittivity and permeability are used to enhance the ability to detect the signals of biosensors based on the surface plasmon resonance (SPR) [15]. To ensure that the resonance is beneficial for SPR biosensors, the process of preparation of metamaterials with different reflectance characteristics is crucial. For predicting the reflectance characteristics

of the metamaterial, SPR biosensors like multilayer perceptron (MLP) and Autoencoder (AE) are used. Afterwards, k-means clustering of the metamaterials was introduced for the dimensional reduction with the help of AE and t-StochasticNeighbor embedding (t-SNE). Hence, without experimenting extensively, the designing of the optimized sensing devices can be boosted up with the clustering of the metamaterials. ML plays a crucial role in predicting the mathematical model for the experimental results. Xiaoyu Zhu et al. [132] measured the voltametric behaviors, i.e., differential pulse voltammetry (DPV), using a fabricated electrode at different concentrations of carbendazim (CBZ). The cyclic voltametery (Figure 10A) and DPV (Figure 10B) of different electrodes from the range of 0.4–1.2 V is also calculated. In Figure 10C, it is observed that with increase in the concentration of CBZ, the value of the peak current starts increasing. Meanwhile, a good linear relationship (Figure 10D) between CBZ concentrations and currents (Ipa) is displayed in the range of 0.5–9.8 µM and 0.006–0.1 µM. The Relevance Vector Machine (RVM) model developed with the input of concentrations is represented in Figure 10E. The results obtained from the RVM model of Root mean square error (RMSE) and R-squared were 0.0143 and 0.9993, indicating that the model could be used for detection CBZ in real samples as it shows excellent performance. Figure 10F diagrammatically illustrates the developed RVM model for estimating CBZ concentration using the electrochemical biosensor. The RVM models possess robustness and generalization ability better than the traditional linear regression. Figure 10G depicts the comparisons of the RVM predicted values and experimental values. It is observed that both the predicted concentrations by RVM models and experimental values are in good agreement.
