Development of Biosensing System from Power to Analysis for GFAP and UCH-L
Lisandro Cunci-lisandro.cunci@upr.edu (Principal Investigator)
Emmanuel Arzuaga- emmanuel.arzuaga@upr.edu )(Co-Principal Investigator))
Project Description
This project seeks to start the development of a complete biosensing system to gather preliminary results to prepare a proposal for the development of this system. This project focuses on the development of a biosensing system to measure glial fibrillary acidic protein (GFAP) and ubiquitin C-terminal hydrolase L1 (UCH-L1) as biomarkers for traumatic brain injury (TBI), which is known to increase the risk of cardiovascular diseases (CVD) like coronary artery disease, strokes, amongst others. This will also increase the number of analytes that can be measured in sweat. The measurement of peptides on sweat has not been possible until now; therefore, we are proposing the detection of GFAP and UCH-L1 on sweat using surface-modified flexible ink-jetted microelectrodes. This work describes the approaches that will be necessary to prepare and validate the aptamers for GFAP and UCH-L1 modified with a redox-label (methylene blue) for detection and monitoring of GFAP and UCH-L1. The wearable potentiostat will be powered with our developed flexible Zn-air battery using a guar-gum quasi-solid state safe electrolyte. The data gathered from our biosensor will be analyzed using machine learning techniques. Machine Learning (ML) based models have been recently designed for different scientific applications, including time series data analysis in various fields, such as biomedical applications (e.g. EEG signals) and analytical chemistry (voltammetry), using data at unprecedented scales and speeds1,2. In this work, we propose to develop a ML-based Prototype for square wave voltammetry (SWV) data analysis. SWV is an electroanalytical technique based on frequency (Hz) and step height (mV). It generates a peak-shaped symmetrical voltammogram. The current is sampled twice during each square wave cycle, once at the end of the forward pulse and once at the end of the reverse pulse. The concentration of the studied substance is proportional to the peak of the current. Since the current peak is larger than the oxidation or reduction of the separated signals, this method has high sensitivity. In this work, we are interested in developing ML-based methods that can automatically process voltammograms and normalize them to speed up the analysis of this data (Figure 1).
Background
Traumatic Brain Injuries (TBI) alter the brain’s functionality permanently, and it is one of the main causes of death. They are typically classified as mild, moderate, and severe, with mild injuries particularly challenging because of their post-injury symptoms3. As science evolves, the methods for diagnosing, prognosis, and monitoring TBI’s have stayed the same for about half a century. Currently, neuroimaging and the Glasgow Coma Scale (GCS) are the gold standards. The problem with these methods is that patients are exposed to harmful ionizing radiation, misreads of CT (Computed Tomographic) scans and MRIs, and healthcare costs increase4 . Additionally, some clinical test results for TBIs typically require up to five days to be analyzed, which can be fatal for patients with these injuries because specific post-injury symptoms can cause irreversible damage or death. According to Steward et al. 5 ,it has been suggested that people suffering from TBI present a higher risk factor for cardiovascular diseases (CVD) like coronary artery disease and strokes, among others. The rates were higher for patients like veterans in Iraq or Afghanistan in the post-9/11 era5.
That is why a non-invasive alternative for the early detection of TBIs is of paramount importance. An electrochemical biosensor can help optimize the diagnosis and prognosis of these injuries and diseases, helping to improve already existing treatment options. In addition, it can help reduce the amount of neuroimaging that patients receive and healthcare costs. For those reasons, our study focuses on developing an electrochemical biosensor utilizing electrochemistry-based bioanalytical techniques for the detection and real-time monitoring of biomarkers. These molecules are of the utmost importance for TBIs and cardiovascular diseases because they can be found in bodily fluids and help us determine the efficacy of a treatment for an illness or condition. Specifically, cell-specific proteins like S100B, NSE, Tau, GFAP, and UCH-L1 are the most studied for TBIs and CVD. Our study focuses on the combination of GFAP and UCH-L1 working in tandem. Because there is evidence that suggests that these two can help detect and determine the severity of the injury, even if the typical methods provide a negative result6 . In addition, it has been suggested that these biomarkers are also related to cardiac arrest, predicting the outcome
For the development of our biosensor, we will employ a series of electrochemical techniques like Electrochemical Impedance Spectroscopy (EIS), Cyclic Voltammetry (CV), and Square Wave Voltammetry (SWV). There are many advantages to using electrochemistry over other techniques that use light-based detection, but the ones mentioned are the most important. First, we can avoid using light measurements, which allow us to detect biomolecules in pure fluids without performing purification. Also, we can modify the surface of our microelectrodes to provide specificity towards our molecule of interest and make it more adaptable. Not only would it measure our molecule of interest but a wide range, instead of only those that oxidize or reduce, which makes them ideal for field applications in a wearable biosensor. Dr. Cunci’s lab has been developing biosensors for non-electroactive molecules for the last 8 years, working on that
NPY is located in different bodily fluids, including sweat, blood, and cerebrospinal fluid, among others. The physiological concentration of NPY in sweat and cerebrospinal fluid is very low in the pg/ml and ng/ml ranges, respectively. We have been able to measure concentrations as low as 10 ng/ml in our last publication using implantable modified platinum microelectrodes without labels. Moreover, adding a redox label, planar bigger modified microelectrodes, and slower voltammetric techniques allow us to improve this measurement method to lower concentrations. In preliminary measurements using platinum microelectrodes, we improved the measurement of NPY when using aptamers as the recognition molecule (Figure 2). In addition, using planar gold circular electrodes, we have measured NPY between 1 – 1000 pg/mL, as seen in Figure 3.
With the rapid technological advancement and the need to develop flexible devices such as our proposed biosensor, a great need has arisen to build batteries or power sources for these flexible devices. In addition to the energy crisis and environmental pollution due to lithium mining and use, it is quite urgent to find a new sustainable power source7 . Although Li-ion rechargeable batteries have been the standard for energy storage due to their excellent power density and
energy, they have drawbacks such as toxicity, flammability, environmental issues, and limited natural abundance8 . Metal ion batteries have gained attention recently as a safer and more environmentally friendly alternative to lithium-based batteries. Among metals ion batteries, zincair batteries (ZABs) have been of great interest due to their safety of using non-flammable electrolytes and non-toxic, low cost, the excellent abundance of Zn metals, and high theoretical specific energy density (1218 Wh kg-1 or 6136 Wh L-19,10. Huang et al. developed a flexible quasisolid zinc-ion battery with a conductive guar gum electrolyte11. The guar gum-based electrolyte developed by this group obtained a conductivity of 1.07 X 10-2 S cm-1 . As a result, the flexible quasi-solid ZIBs deliver a high specific capacity of 308.2 mAh g-1 at 0.3 A g-1 and a charge and discharge capacity of 131.6 mAg g-1 at 6.0 A g-111. More recently, Bhardwaj et al. developed a guar gum-based electrolyte employing LaMnO3 (LMO) perovskite as a bi-functional electrocatalyst for zinc-air batteries12. The group reports a power density of 9.77 mW cm-2 at the current density of 11.12 mA cm-2 with a specific capacity of 2624 mAh g-112.
With the rapid technological advancement and the need to develop flexible devices such as our proposed biosensor, a great need has arisen to build batteries or power sources for these flexible devices. In addition to the energy crisis and environmental pollution due to lithium mining and use, it is quite urgent to find a new sustainable power source7 . Although Li-ion rechargeable batteries have been the standard for energy storage due to their excellent power density and
energy, they have drawbacks such as toxicity, flammability, environmental issues, and limited natural abundance8 . Metal ion batteries have gained attention recently as a safer and more environmentally friendly alternative to lithium-based batteries. Among metals ion batteries, zincair batteries (ZABs) have been of great interest due to their safety of using non-flammable electrolytes and non-toxic, low cost, the excellent abundance of Zn metals, and high theoretical specific energy density (1218 Wh kg-1 or 6136 Wh L-19,10. Huang et al. developed a flexible quasisolid zinc-ion battery with a conductive guar gum electrolyte11. The guar gum-based electrolyte developed by this group obtained a conductivity of 1.07 X 10-2 S cm-1 . As a result, the flexible quasi-solid ZIBs deliver a high specific capacity of 308.2 mAh g-1 at 0.3 A g-1 and a charge and discharge capacity of 131.6 mAg g-1 at 6.0 A g-111. More recently, Bhardwaj et al. developed a guar gum-based electrolyte employing LaMnO3 (LMO) perovskite as a bi-functional electrocatalyst for zinc-air batteries12. The group reports a power density of 9.77 mW cm-2 at the current density of 11.12 mA cm-2 with a specific capacity of 2624 mAh g-112.
Guar gum is a natural polysaccharide extracted from endospermic beans usually used in food but also possesses high biocompatibility, biodegradability, hydrophilic, and non-toxicity13. In addition, guar gum is soluble in water, and its high acid resistance makes it ideal for developing solid-state electrolytes. It has previously been used as a binder in lithium-sulfur (LiS) batteries with excellent cycling performance and high rate capability11,14. Despite the reported advances related to guar gum as an electrolyte for flexible batteries, some challenges must be addressed. One of the challenges is to increase the conductivity, cyclic performance stability, and catalytic materials that can favor the application of rechargeable metal-air batteries. One of the materials of great interest
is onion-like carbon (OLC). OLCs are carbon allotropes used as anode and cathode material in lithium-ion batteries (LIB) and Metal-ion batteries due to their unique structure of concentric carbon shells, high electrical conductivity, and large surface area15,16. However, to our knowledge, OLC has not been studied in developing solid-state electrolytes for flexible batteries.
The intellectual merit of our research is separated into two parts. The intellectual merit of our biosensor research lies in the importance of developing an analytical technique and analysis that allows the measurement of essential biomarkers, such as GFAP and UCH-L1, that cannot be accomplished with existing techniques in sweat. Neuropeptides are very important to understand the constant dynamic changes that modulate behavior because, within the body, they work as biomarkers of different injuries as well as neurotransmitters and neuromodulators between neurons as well as a group of neurons. Other techniques have been used to measure proteins, but they all depend on invasive procedures while also providing low temporal resolution compared to electrochemical techniques. The main challenge is that peptides are not specifically oxidized or reduced to measure them using electrochemical techniques selectively. Therefore, the use of a novel strategy to monitor peptides is needed. Electrochemical methods such as cyclic voltammetry paired with a known and robust label such as methylene blue can detect small changes in the chemical and physical properties of the surface of the microelectrode. Combined with a novel strategy for continuous measurement and highly selective aptamers, the proposed biosensor will be used to detect neuropeptides in sweat. The intellectual merit of our battery research lies in developing a quasi-solid-state electrolyte based on guar gum and OLC as a safe strategy for wearable devices. The importance of using materials such as polysaccharides for developing electrolytes for flexible batteries is that they can be safely worn on a person’s body.
The broader impacts consist of three main parts. First, monitoring peptides in sweat will provide a novel strategy to detect and monitor TBI, which will directly increase the risk for cardiovascular diseases. Understanding the correlation between peptides and behavior in people suffering from these diseases will allow the research community to develop novel pharmacological treatments. Secondly, developing a flexible, low-cost, and eco-friendly non-lithium battery as a power source for wearable electrochemical sensors is essential. With the growth of portable and wearable devices, the demand for batteries increases, which is led by the market for lithium batteries, which are dangerous and toxic for the environment and highly cost since lithium is scarce in the Earth’s crust. Third, as part of a primarily undergraduate university (PUI) with a majority being firstgeneration university students, the integration of research and education in our laboratory will increase the number of Hispanic students that work in STEM disciplines in Puerto Rico. Our laboratory consists of 12 undergraduate students at any time, with the majority being women, first-generation university students. Moreover, two high school students participated two years ago and two this current year in our laboratory, where they developed their scientific fairs projects and obtained first place in their schools.
Innovation
This proposal will provide a novel technique that will allow IRG 1 to start measuring GFAP and UCH-L1 in sweat as a proof-of-concept to continue measuring different target peptides. It will allow IRG 2 to have a flexible safe battery to power the lab-made potentiostat for our measurement. Moreover, it will allow IRG 3 to have an algorithm that will analyze the data obtained.
This proof-of-concept is required in order to develop a novel analytical technique to measure biomarkers noninvasively. The lack of these methods hinders the advancement of wearable technologies for biomedical devices. Therefore, GFAP and UCH-L1 measurement in sweat using a robust electrochemical technique is imperative. The innovation of this proposal is the combination of aptamers and methylene blue on surface-modified ink-jetted electrodes together with microfluidics for sweat harvesting with the aim to detecting novel targets in sweat. These targets are highly important biomolecules to be monitored. GFAP and UCH-L1 have been related to diseases such as TBI with correlation on a higher risk on cardiovascular diseases5. Therefore, being able to measure these targets using non-invasive techniques will allow better diagnosing and treating these illnesses. In order to analyze the data obtained using our biosensor, we will use the computer cluster purchased by the Center for the Advancements of Wearable Technologies (CAWT)
Traditional ML methods proposed for voltammogram data include Decision Trees, Naïve Bayes, SVM and Artificial Neural Networks (ANN). Out of these strategies SVM have been found to provide better overall results 1 . Although ANN strategies have not proven to be definitively better than SVM, there is still work to be done to validate the use of Deep Learning (DL) Methodologies to analyze SWV data as an alternative to traditional ML strategies, avoiding the burden of manual data tagging or specialized knowledge of data arrangements required for better results in ML, as well as using less training data
In this project we will evaluate the performance of DL algorithms for SWV and compare them to traditional ML methods (DT, NB and SVM). Our hypothesis is that properly trained DL algorithms will prove equal or better in performance than traditional ML methods. In particular, we will evaluate residual neural network architectures such as LSTMs, U-Net, and merge-and-run (MnR) strategies.
In order to power the biosensor, this proposal will provide a novel guar gum gel electrolyte for developing flexible zinc-ion and zinc-air batteries for our wearable potentiostat. We will manufacture a battery prototype to study the quasi-solid electrolyte, which will be a great addition to IRG2 as a strategy for developing wearable batteries. In turn, the IRG1 group will benefit from this project, where a flexible zinc battery can be used as an energy source for wearable sensors. This proposal’s innovation combines the polysaccharide guar gum and OLC to manufacture a novel quasi-solid gel electrolyte for flexible zinc batteries. The development of these flexible batteries that can be worn on the body is of great importance due to the safety concerns of lithiumbatteries. In addition, flexible zinc batteries are cheaper and safer in contrast to lithium batteries due to the use of electrolytes that are flammable and toxic(Cha et al., 2018; C. Wang et al., 2021).
Summary of Work from Previous Exploratory Funds
Our group has developed the use of continuous electrochemical impedance (CEI) to monitor the adsorption of biomolecules, such as neurotransmitters and neuropeptides, on microelectrodes with dynamic changes in concentration. CEI is a modification of EIS to measure a subset of frequencies with time18. In electrochemical impedance spectroscopy (EIS), a sine wave voltage with a sufficiently small amplitude that the current response continues to be linear is applied to the working electrode. The current measured at each frequency is used to calculate the impedance with different methods, such as the Lissajous curve, which plots voltage vs. current for each cycle, or Fast Fourier Transform (FFT)18. While EIS can only be done discreetly, CEI measurement applies a signal produced by the convolution of sine waves with different frequencies. Measuring rapid dynamic changes requires cutting low frequencies (that take longer to measure) and optimizing the frequencies to be measured with time. Therefore, very rapid changes in the concentration of molecules can be seen using between one single frequency depending on the environment and molecules present.
Figure 4 shows the results obtained at 30 Hz when measuring NPY using different aptamers. As can be seen, all the different aptamers shown have selectivity to NPY in different levels with selectivity against PPY, PYY, and somatostatin (commonly found with NPY) depending on the aptamer. Following the results shown, we can use an array of sensors with different aptamers on their surfaces that can provide selectivity for NPY and against different confounding molecules.
To have sensors that are supporting the rapid, convenient measurement of NPY irrespective of their chemical reactivity, we want to develop an electrochemical aptamer-based sensor, an approach that can eventually support measurements directly in sweat, clinical samples, among others. In this case, an aptamer selected to specifically bind NPY is attached to the surface of the electrode via a thiol group and modified with a redox reporter here methylene blue (MB) to support electrochemical signaling using square wave voltammetry (SWV). Because the signaling is predicated on a binding-induced current change, the sensor shouldbe sensitive to NPY through MB. These results are great and confirmed with electrochemical response using SWV for different aptamer for NPY. Exploring this idea, we can have a convenient measurement of the NPY using MB as a redox probe.
We have used silver inkjet-printed flexible biosensors to detect NPY using SWV. We have characterized the biosensors using artificial sweat with different pH following the range found in sweat between 4.0 and 8.0. Silver electrodes have shown to be very unstable at the potentials needed providing a very low resolution as shown in Figure 5. Due to the low resolution found at limit pH, we have started to deposit a small layer of gold on top of the inkjet-printed silver ink to provide stability to the electrodes.
Products of Work from Previous Exploratory Funds
Espinosa, A.; Diaz, J.; Vazquez, E.; Acosta, L.; Santiago, A.; Cunci, L. Fabrication of Paper-Based Microfluidic Devices Using a 3D Printer and a Commercially-Available Wax Filament. Talanta Open 2022, 6, 100142.
CAWT Relevancy Statement
• How this project catalyzes the CAWT efforts for the integration of all the three relevant areas: sensor, power and AI/ML? As part of CAWT, we need to increase the number of analytes that can be measured using wearable technologies such as sweat biosensors while providing enough power to reliably produce the measurements in time, while automatically analyzing the data obtained. In order to build final biosensors, different analytical strategies must be developed that are able to measure novel analytes such as peptides that cannot be measured until now. In this proposal, we are devising a novel strategy for measuring GFAP and UCH-L1 on sweat samples. The successful completion of this project will allow the researchers in IRG 1 and IRG 3 to measure and analyze two biomarkers, as well as researchers in IRG 2 to have a safe battery to power the potentiostat. We need to increase the development of flexible batteries for wearable devices that are economical, non-toxic, and safe to use in wearable devices. In this proposal, we develop a strategy for manufacturing a flexible electrolyte to be used in flexible wearable zin-ion and zincair batteries. We will also be manufacturing a flexible zinc battery prototype using the novel electrolyte we propose. The successful completion of this project will allow the research in IRG 1, IRG 2, and IRG 3 to collaborate in the development of a battery powering a small and light-weight wearable potentiostat measuring GFAP and UCH-L1 in sweat. The scope of this project is aligned to target one of this competition which is to Develop AI, data mining, and Big Data software tools to build a computational pipeline to guide the design, discovery or selection of materials for wearable technology.
• How will this project capitalize on existing CAWT resources? Dr. Cunci will use the acquired Bionavis SPR for the development of this biosensor as well as the Keyence 3D surface profiler (IRG 1). Dr. Arzuaga will use the acquired computer cluster for analysis of the data purchased by IRG 3. Moreover, Dr. Cunci will use the battery cyclers purchased by IRG 2. Dr. Cunci has Figure 5 – SWV in silver flexible electrodes. Continuous electrochemical impedance measured also established collaborations with Dr. Jose A. Rodriguez in the Biology Department at the University of Puerto Rico in Rio Piedras
To have sensors that are supporting the rapid, convenient measurement of NPY irrespective of their chemical reactivity, we want to develop an electrochemical aptamer-based sensor, an approach that can eventually support measurements directly in sweat, clinical samples, among others. In this case, an aptamer selected to specifically bind NPY is attached to the surface of the electrode via a thiol group and modified with a redox reporter here methylene blue (MB) to support electrochemical signaling using square wave voltammetry (SWV). Because the signaling is predicated on a binding-induced current change, the sensor shouldbe sensitive to NPY through MB. These results are great and confirmed with electrochemical response using SWV for different aptamer for NPY. Exploring this idea, we can have a convenient measurement of the NPY using MB as a redox probe.
We have used silver inkjet-printed flexible biosensors to detect NPY using SWV. We have characterized the biosensors using artificial sweat with different pH following the range found in sweat between 4.0 and 8.0. Silver electrodes have shown to be very unstable at the potentials needed providing a very low resolution as shown in Figure 5. Due to the low resolution found at limit pH, we have started to deposit a small layer of gold on top of the inkjet-printed silver ink to provide stability to the electrodes.
Research Plan
Wearable devices can be engineered containing biosensors to detect biomacromolecules related to specific diseases. Moreover, the electrochemical capacitive behavior of DNA anchored onto metallic and semi-metallic surfaces via self-assembly with various chemistries (e.g., Au-S, Au-Carbon Nanotubes), has shown great promise as the basis for detection in biomedical devices.2 Self-assembled ssDNA on a metallic interface such as gold3–5 or semi-metallic surfaces of carbon nanotubes6 and diamond7,8 has already shown potential for incorporation into DNA microarrays, which can monitor human health. Electrochemical biosensors complement optical measurements by taking advantage of interfacial changes in the chemical properties of the functionalized microelectrodes. Different biomarkers that we have successfully detected include enzymes, DNA, and small peptides. The first objective of this project is to develop a biosensor using aptamers as the biorecognition molecule for glial fibrillary acidic protein (GFAP) and ubiquitin C-terminal hydrolase L1 (UCH-L1) as biomarkers for TBI. The second objective of this project is the development of a quasi-solid electrolyte based on guar-gum for safe and flexible metal-air batteries. The developed sensor will use a small and light-weight potentiostat powered by our developed metal-air battery. As the third objective, we will develop a deep learning algorighm to analyze the data measured with our biosensor.
To achieve these goals, we have separated our project in three parts: (1) development of aptamers for GFAP and UCH-L1, (2) development of guar-gum electrolyte for Zn-air batteries, and (3) development of the algorithm of deep learning for the data obtained with the sensors.
Development of Aptamers for GFAP and UCH-L1
Currently, no aptamer is developed for GFAP and UCH-L1 in the research literature to the best of our knowledge. These two biomarkers are typically measured with laboratory testing methods that require expensive instrumentation and trained personnel. They have recently been approved as biomarkers for TBI in serum by the FDA. The development of aptamers for their detection using wearable biosensors will help us monitor them within minutes and hours of their release. As can be seen in Figure 6 , GFAP and UCH-L1 are released during the first minute-to-hours of TBI. Therefore, continuous monitoring of these biomarkers in specific situations is of utmost importance because they may have decreased when a serum sample can be obtained.
Aptamers for GFAP and UCH-L1 will be developed in collaboration with Dr. Jose Rodriguez Martinez from the Department of Biology at the University of Puerto Rico in Rio Piedras. Aptamers will be functionalized to attach them to the surface of the electrodes and a redox label. Cyclic voltammetry will be employed to measure the redox peak of a modified methylene blue (MB’) at the end of GFAP-selective and UCH-L1- selective aptamers. MB’-modified aptamers will be tethered to silver ink-jetted electrodes on a flexible substrate (Figure 7A). This technique allows biomarkers to be accurately measured using specific electrochemical interactions without the need for sensitive enzymes anchored to the surface of the electrodes. Silver-based ink-jetted flexible electrodes were fabricated on a flexible substrate as part of our NSF-EPSCoR project, an upgrade from a silicon and gold-based microchip already developed by Dr. Cunci in which two interdigitated microelectrodes were used to detect an enzyme.2
In this proposal, GFAP and UCH-L1 will be tested in artificial sweat by fabricating four devices separated into two three-electrode systems and two two-electrode systems, as seen in Figure 7B-E. These prototype patches will allow us to add a biosensing system on the main channel close to the sweat input or at the input in order to have better time resolution. Figure 7B-E show the different designs that will be tested for sweat harvesting. They will be tested using a flow cell that has been in development following Koh et al.9 , an artificial sweat pore system in collaboration with Dr. Pedro Resto from the University of Puerto
Rico at Mayaguez. This flow cell uses a polymer film with pores fabricated using a laser. 100 pores per cm2 are used, comparable to the eccrine sweat gland density in humans. A sweat harvesting area of 10 mm2 will recover from 1.2 to 12 µl/hour of sweat10; therefore, channels must be small enough to convert that volume into a sufficient linear velocity through the channel to have enough time resolution. To have a linear speed of 1 mm/min through the microchannel at the peak of 12 µl/hour, the area of the channel must be 0.2 mm2 with a square area of 0.45 mm2.
Two different biosensor strategies have been devised: (1) One-time measurement and (2) continuous measurement. The requirements for both are very different. One-time measurement allows for sweat to be collected and concentrated before or during measurement to obtain the concentration of a specific analyte. On the other hand, continuous measurement requires using a reservoir to work as a waste area so that sweat continues flowing within the measurement channel. In this project, while our long-term goal is to work on continuous measurement, a single measurement will be tested first for non-invasive one-time measurement of a specific analyte. Once single-measurement testing and a calibration curve can be obtained, the continuous measurement will be tested.
Whether the research will focus on eccrine or apocrine sweat, many formulations can be found. 11 Given that our work will focus on sweat released by eccrine glands, we will perform the tests using the alkaline and acid formulation stated in ISO 105-E01:2013 to test the in-vitro sweat delivery system. Two materials will be tested for these microelectrodes: bare silver ink-jetted electrodes and gold-plated silver ink-jetted electrodes. Electroactive amino acids will be added in the range found in sweat mixed with the artificial sweat (i.e. L-methionine, L-tyrosine, L-tryptophan, Lcysteine, and L-Histidine)
Fabricated devices will be connected to an electrochemical HPLC with PDA detector to corroborate the correct measurement of the target molecule. Most importantly, the HPLC will allow the calibration of the linear velocity in the microchannels for the different patch versions. The optimization of the developed patch will be done by using the flow, linear velocity of the solution in the microchannels, and time resolution of the measured target molecule together.
The addition of a redox active molecule (i.e. modified methylene blue) at the end of the aptamer allows the use of cyclic voltammetry. Aptamers with an electroactive molecule, a MB’ molecule, at the 5′ end will be covalently attached to the microelectrodes. The adsorption of GFAP and UCH-L1 causes the redox reaction of MB to change due to steric effects, which is seen in the oxidation peak.12 The analytical signal is the change in redox activity due to GFAP and UCH-L1 adsorption on the aptamer. The Bionavis SPR acquired by IRG 1 will be used as a control experiment to demonstrate an interaction between GFAP and UCH-L1 and the MB’-aptamers.
Figure 9 shows the custom-made potentiostat that is able to use the techniques proposed in this project, cyclic voltammetry and SWV. The data is gathered using the bipotentiostat in the middle of the picture (Figure 9B), EmStat pico, and information is recorded in the micro-SD card (Figure 9A) when the button (Figure 9D) is pushed. The information recorded in the micro-SD card is copied to the computer using the serial-to-USB connector (Figure 9C) and PSTrace software from PalmSens. A smaller flexible circuit will be developed in this proposal to be attached to the sensor for wearable measurements. The potentiostat shown in Figure 9 will be powered using a fllexible battery developed by our team. The complete proposed system is shown in Figure 10.
Development of Quasi-Solid State Electrolyte for Zn-air Battery
A quasi-solid state and flexible electrolyte can be fabricated using polymers or polysaccharides capable of forming a gel structure. The use of materials such as guar gum is an excellent option and is of great interest in the development of solid-state electrolytes because they are biocompatible, non-toxic, and do not represent a hazard that can cause harm to a person’s body Figure 8 – Flexible biosensor prototype with interdigital silver microelectrodes inkjet-printed on PET. Figure 9 – Small and low weight potentiostat prototype for wearable biosensors. (A) Micro-SD card for data storage, (B) EmStat pico potentiostat, (C) serial-to-usb converter to obtain the data stored at the micro-SD card, and (D) button to trigger recording. when used. Guar gum is soluble in water and forms a type of gel. Our goal is to develop a flexible guar gum gel electrolyte with onion-like carbon (OLC) for zinc-ion and zinc-air batteries. To achieve this goal, zinc will be electrochemically deposited on a piece of carbon cloth that works as the anode, a gel electrolyte based on guar gum and OLC to increase conductivity, and a piece of carbon cloth modified with metal doped OLC particles for the oxygen reduction reaction (ORR) will be used as cathode. The flexible electrolyte we propose will be made by diluting a quantity of guar gum, ZnSO4, and OLC in distilled water. Once the solution has a gel-like consistency, we will test the ionic conductivity, charge/discharge for circuit protection, power density, and cycle efficiency. To study the intrinsic properties of the electrolyte, the use of Raman spectra will be carried out to study the structure of the electrolyte, as well as the use of scanning electron microscopy (SEM) and energy-dispersive X-ray spectroscopy (EDS). Electrochemical techniques such as electrochemical impedance spectroscopy (EIS), Linear sweep voltammetry, cyclic voltammetry, and chronoamperometry will be carried out to study ionic conductivity, power density, corrosion rate, overpotential, and ion diffusion capacity. Additionally, cycle curve studies for lifespan and cycle time efficiency, bend and stretch tests to study the flexibility of the electrolyte, and charge/discharge curves for energy density and circuit protection analysis will be carried out. OLC will be doped with different metal and nitrogen ions to obtain an electrolyte with excellent conductivity and energy density that can surpass previously published works. The performance of each OLC doped with different metals or a combination of metals and nitrogen will be studied and compared to determine the optimal flexible electrolyte for zinc-air batteries.
Two strategies have been devised for flexible non-lithium batteries: (1) single-use disposable batteries and (2) long-use and rechargeable batteries. A disposable flexible battery allows powering a wearable sensor to carry out a measurement or monitoring of an analyte of interest. Once the battery is completely discharged, the materials are discarded or recycled to make a new battery. On the other hand, a rechargeable flexible battery requires materials that can carry out ORR and OER reactions. A rechargeable, non-lithium, nonrechargeable flexible battery will be tested first to provide power to a wearable electrochemical sensor. Our laboratory group started working on preliminary data related to the development of the flexible electrolyte based on guar gum and OLC. Figure 11 shows an image of the prototype of a flexible electrolyte based on guar gum and OLC. Preliminary electrochemical experiments were also performed to study the performance of the electrolyte gel. Figure 12 shows the results obtained from linear sweep voltammetry performed on a guar gum/OLC gel electrolyte. Electrochemical experiments were performed using two stainless steel plates as working and counter/reference electrodes. The results show that around 0.7 volts, the electrolyte begins to break and carries out an oxidation reaction.
Development of Machine Learning Algorithm for Biosensor Data Analysis
In this work, we plan on designing a DL algorithm for SWV. As part of the CAWT IRG3 activities, we have developed a prototype Big Data Pipeline using Hadoop and Spark19. We will leverage Figure 11 – Guar-gum electrolyte using OLC to increase conductivity. Figure 12 – Linear sweep voltammetry of guar-gum electrolyte with OLC using two stainless steel plates as working and reference electrodes. this prototype to build DL models that can execute in this platform. In particular we will evaluate (1) three different DL based architectures, abstract their respective learning block and evaluate the respective combination of the different block to (2) propose new DL algorithms that generate better normalized voltammogram.
Development of Machine Learning Algorithm for Biosensor Data Analysis
a) Merge and Run based on MnRCMNe20
The Merge and Run (MnR) based DNNs are a modification of the residual neural network architecture (ResNet)21. They rely on the repetition of a block that is a linear idempotent function, where the transformation matrix is idempotent (see Figure 13). This implies that the information from the early blocks can quickly flow to the later blocks, and the gradient can be quickly backpropagated to the early blocks from the later blocks requiring a smaller set of training data to converge.
The Merge and Run (MnR) based DNNs are a modification of the residual neural network architecture (ResNet)21. They rely on the repetition of a block that is a linear idempotent function, where the transformation matrix is idempotent (see Figure 13). This implies that the information from the early blocks can quickly flow to the later blocks, and the gradient can be quickly backpropagated to the early blocks from the later blocks requiring a smaller set of training data to converge.
b) Architecture based on IRNet (Figure 14)22
IRNet is a ResNet based architecture that provides continuous shortcut connections after each layer so that each layer learns the residual mapping between its output and input. Such organization although simple, provides a stable learning convergence as well as is well suited for deeper extensions
c) Architecture based on UNet23:
UNet is a DNN design proposed for biomedical image segmentation and cell tracking. It consists of contracting paths to capture context and a symmetric expanding path that enables precise localization (see Figure 15).
Project Outcomes
In this work, we plan on designing a DL algorithm for SWV. As part of the CAWT IRG3 activities, we have developed a prototype Big Data Pipeline using Hadoop and Spark19. We will leverage Figure 11 – Guar-gum electrolyte using OLC to increase conductivity. Figure 12 – Linear sweep voltammetry of guar-gum electrolyte with OLC using two stainless steel plates as working and reference electrodes. this prototype to build DL models that can execute in this platform. In particular we will evaluate (1) three different DL based architectures, abstract their respective learning block and evaluate the respective combination of the different block to (2) propose new DL algorithms that generate better normalized voltammogram.
Sustainability Statement
Results of this work will enable the PIs to pursue external funding from competitive programs. Opportunities within NSF and NIH will be explored as well as opportunities in agencies such as NASA and the DoD from which the Co-PI Arzuaga currently has funded projects will also be explored
Group Description
This project will be a multi IRG collaboration between Dr. Lisandro Cunci (IRG1 & IRG2) and Dr. Emmanuel Arzuaga (IRG3).
Dr. Lisandro Cunci is an Assistant Professor at the University of Puerto Rico in Rio Piedras, and an Adjunct Professor at Universidad Ana G. Mendez, Gurabo Campus. He graduated from his Ph.D. in Chemistry in December 2013 under the mentorship of Dr. Carlos Cabrera at University of Puerto Rico, Rio Piedras. After his doctorate, Dr. Cunci was a Postdoctoral Associate for a year at the Molecular Sciences Research Center at the University of Puerto Rico, and later was hired as Associate Researcher in charge of part of the scientific instrumentation. Dr. Cunci also has a degree in Materials Science Engineering where he worked in wireless sensing during his thesis in Dublin, Ohio. Since his position at UAGM, Dr. Cunci has developed his independent research Figure 15 – Network diagram of UNet based architecture. Orange rectangles with the number of units below them represent fully connected layers + Batch Normalization + ReLU. Blue rectangle represents a fully connected layer + Linear activation. in biosensors and energy materials publishing in top journals of his field such as Analytical Chemistry. He has also served as a panel reviewer at the National Science Foundation, Department of Energy, and within the Institution for UAGM funding, as well as being a reviewer for different scientific journals. Dr. Emmanuel Arzuaga is a Professor with a joint appointment at the Departments of Computer Science and Engineering and Electrical and Computer Engineering of the University of Puerto Rico at Mayaguez. He is currently the director of the Laboratory for Applied Remote Sensing, Imaging and Photonics (LARSIP) and the Co-Director of the Center for Aerospace and Unmanned Systems Engineering (CAUSE). His research interests include: Virtualization, Cloud Infrastructure, Pattern Recognition, Remote Sensing, and Computer Systems Architecture and Security. Dr. Arzuaga has more than 15 years of experience in the areas pattern recognition, machine learning and image analytics. His work in this area has more than 300 citations in research articles