Method and system for detecting per- and polyfluoroalkyl substances, lung cancer, and bacterial biofilms using bioelectronic sensor

Abstract

A method of detecting the presence or absence of a per- or polyfluoroalkyl substance (PFAS) or lung cancer is provided herein that includes exposing a biological chemosensory array to 1) at least one test gas-phase PFAS or volatile organic compound (VOC) mixture and 2) at least one control gas-phase PFAS or VOC mixture, at least one control gas-phase non-PFAS or non-VOC mixture, or a combination thereof; obtaining at least one test neuronal response and at least one control neuronal response from the biological chemosensory array; comparing the at least one test neuronal response to the at least one control response; and outputting a result based on the comparison. A system for detecting the presence or absence of a PFAS or lung cancer is also provided herein, which includes an odor stimulus delivery component for delivering a test gas-phase PFAS or VOC mixture and a control to a biological chemosensory array, a neuron probe for detecting a test neuronal response and at least one control neuronal response, and a processor for storing the test and control neuronal responses.

Description

TEC2025-0129 I TEC2025-0177 I TEC2026-00016550-000538-WO-POAMETHOD AND SYSTEM FOR DETECTING PER- AND POLYFLUOROALKYL SUBSTANCES, LUNG CANCER, AND BACTERIAL BIOFILMS USING BIOELECTRONIC SENSORCROSS-REFERENCE TO RELATED PATENT APPLICATIONS

[0001] This patent application claims the benefit of U.S. Provisional Patent Application No.63 / 845.714 filed on 17 July 2025. U.S. Provisional Patent Application No. 63 / 836,627 filed on 01 July 2025, and U.S. Provisional Patent Application No. 63 / 778,182 filed on 26 March 2025. The entire content of each application recited above is hereby incorporated by reference.GOVERNMENT SUPPORT STATEMENT

[0002] This invention was made with government support under 2238686 awarded by the National Science Foundation. The government has certain rights in the invention.FIELD

[0003] This disclosure generally relates to a method and system for detecting the presence or absence of per- and polyfluoroalkyl substances, lung cancer, and bacterial biofilms by exposing insect antennae to gaseous mixtures, detecting test neuronal responses, and comparing the test neuronal responses to control neuronal responses.BACKGROUND

[0004] This section provides background information related to the present disclosure, which is not necessarily prior art.

[0005] Per- and polyfluoroalkyl substances (PFAS) pose a significant environmental threat due to their widespread presence in consumer waste and resistance to degradation. These “forever chemicals” persist in various ecosystems and exhibit bio accumulative behavior. Increased human exposure to PFAS has been linked to numerous health issues. Despite their growing relevance, current detection methods often lack the sensitivity and efficiency needed for comprehensive environmental monitoring and struggle to simultaneously detect multiple PFAS at environmental concentrations. This creates an urgent need for more advanced chemical sensors capable of precise environmental pollution detection.

[0006] In addition to PFAS, lung cancer remains a leading cause of cancer-related morbidity and mortality worldwide, underscoring the continued need for improved diagnostic technologies. TheTEC2025-0129 I TEC2025-0177 I TEC2026-00016550-000538-WO-POAmost promising approach for lung cancer screening is with low dose computed tomography (LDCT) in high-risk populations, which has shown to provide statistically significant mortality reduction in patient. This approach, however, relies on risk modeling methodology to select high-risk populations and to minimize the potential damages to radiation exposure to a broader population. A more accessible, noninvasive diagnostic tool is needed that lacks harmful exposures.

[0007] Additionally, bacterial biofilms pose a significant challenge across clinical, industrial, and environmental settings due to their inherent resistance to conventional antimicrobial treatments. Current methods for detection and diagnosis of bacteria mostly rely on culturing blood or other biological samples to grow a bacterial culture in vitro. These tests take a few days to grow the sample in a lab for results and require skilled personnel. While faster detection methods, such as qPCR and ELISA, have been developed, these methods do not distinguish between biofilm and planktonic, and they can only test a single targeted disease at a time. A non-invasive and rapid alternative diagnostic method is needed.

[0008] Thus, there is a need in the art for new and improved methods and systems for precisely detecting PFAS, lung cancer, and bacterial biofilms.SUMMARY

[0009] This section provides a general summary of the disclosure and is not a comprehensive disclosure of its full scope or all of its features.

[0010] In certain aspects, the present disclosure provides a method of detecting a presence or absence of a per- or polyfluoroalkyl substance (PFAS). The method includes exposing a biological chemosensory array to 1) at least one test gas-phase PFAS mixture and 2) at least one control gasphase PFAS mixture, at least one control gas-phase non-PFAS mixture, or a combination thereof; obtaining at least one test neuronal response and at least one control neuronal response from the biological chemosensory array; comparing the at least one test neuronal response to the at least one control neuronal response; and outputting a result based on the comparison, where the result indicates a presence or an absence of a PFAS or a probability of a presence or an absence of a PFAS.

[0011] Also provided herein is a system for detecting the presence or absence of a PFAS. The system includes an odor stimulus delivery component for delivering 1) at least one test gas-phase PFAS mixture and 2) at least one control gas-phase PFAS mixture, at least one control gas-phase non-PFAS mixture, or a combination thereof to one or more biological chemosensory array, where theTEC2025-0129 I TEC2025-0177 I TEC2026-00016550-000538-WO-POAone of more biological chemosensory array is stabilized by a stabilizing component; a neuron probe for detecting at least one test neuronal response and at least one control neuronal response from the one or more biological chemosensory array; and at least one processor detecting at least one test neuronal response and at least one control neuronal response from the one or more biological chemosensory array.

[0012] In some embodiments, the test gas -phase PFAS mixture is emitted from one or more environmental sample, such as air, water, soil, sediment, dust, a biosolid, or a biological sample.

[0013] In certain aspects, the present disclosure also provides a method of detecting a presence or absence of lung cancer. The method includes exposing a biological chemosensory array to 1) at least one test volatile organic compound (VOC) mixture and 2) at least one control VOC mixture, at least one control non-VOC mixture, or a combination thereof; obtaining at least one test neuronal response and at least one control neuronal response from the biological chemosensory array; comparing the at least one test neuronal response to the at least one control neuronal response; and outputting a result based on the comparison, where the result indicates a presence or absence of lung cancer or a probability of a presence or absence of lung cancer.

[0014] Also provided herein is a system for detecting the presence or absence of lung cancer. The system includes an odor stimulus delivery component for delivering 1) at least one test volatile organic compound (VOC) mixture and 2) at least one control VOC mixture, at least one control non-VOC mixture, or a combination thereof to one or more biological chemosensory array, where the one or more biological chemosensory array is stabilized by a stabilizing component; a neuron probe for detecting at least one test neuronal response and at least one control neuronal response from the one or more biological chemosensory array; and at least one processor which stores the at least one test neuronal response and the at least one control neuronal response in memory, where the at least one test neuronal response and the at least one control neuronal response are one or more neuronal voltage signals.

[0015] In some embodiments, the test VOC mixture is a gas VOC mixture emitted from a biological sample, such as breath, urine, sweat, or blood.

[0016] Also provided herein is a method of diagnosing, preventing, and / or treating lung cancer based on the presence of lung cancer or the probability of the presence of lung cancer obtained from the method of detecting lung cancer.TEC2025-0129 I TEC2025-0177 I TEC2026-00016550-000538-WO-POA

[0017] Further areas of applicability will become apparent from the description provided herein. The description and specific examples in this summary are intended for purposes of illustration only and are not intended to limit the scope of the present disclosure.BRIEF DESCRIPTION OF THE DRAWINGS

[0018] The drawings described herein are for illustrative purposes only of selected embodiments and not all possible implementations and are not intended to limit the scope of the present disclosure.

[0019] FIG. 1A-B illustrate an example overview of electrophysiology design and signaling pathways. FIG. 1A is a schematic showing how clean air is mixed with odor laden air using the olfactometer before delivering to a locust antennae via an odor line. Odor presentation timing is controlled through the olfactometer, recording controller, and Intan board. Information is relayed through the DAQ terminal to the computer. The Faraday cage houses the insect with electrode implanted in the antennal lobe (AL) and the Intan head stage to isolate the recordings, limiting electrical interference. Neural activity is visualized and stored on the computer. FIG. IB is a diagram depicting device signaling pathways with directional arrows indicating communication from one device to the next. Bi-directional arrows indicate communication both ways. The Intan Head Stage and Electrode With DIP Socket are devices found within the Faraday cage.

[0020] FIG. 2A-D show the classification of pure PFAS compounds using root mean squared (R.M.S.) analysis. FIG. 2A depicts odor-evoked R.M.S. trajectories for all seven pure PFAS compounds visualized using principal component analysis (PCA) for dimensionality reduction. These trajectories represent neural activity from 0.25 - 1.0 seconds following odor exposure. The colored arrows depict the progression of trajectories, aligned at 0.25 seconds after odor onset (black dot). R.M.S. captures the real-time evolution of odor responses through the sensor. A total of 51 tetrodes were analyzed. FIG. 2B shows how linear discriminant analysis (LDA), a 3-class supervised dimensionality reduction technique, maps spatiotemporal R.M.S. odor responses between 0.25 - 1.0 seconds post-exposure. Clustering reveals distinct temporal patterns for each odor stimulus (n = 51).FIG. 2C shows how a high-dimensional leave-one-trial-out (LOTO) confusion matrix classifies R.M.S. responses to the seven pure PFAS compounds and the control. Classification of true versus predicted labels was based on the smallest Euclidean distance within the 0.5 - 6.0 second window after stimulus onset. FIG. 2D is a sensitivity and specificity table depicting real-time compound detection. Values were generated from the LOTO matrix.TEC2025-0129 I TEC2025-0177 I TEC2026-00016550-000538-WO-POA

[0021] FIG. 3A-E illustrate that environmentally relevant PFAS concentrations show high classification accuracy using R.M.S. FIG.3A is a table including parts per billion (ppb) calculations for three PFAS compounds at two concentrations (6 odors total) were made using Raoult’ s law and published chemical characteristics. The lowest concentration (0.0001%) represents parts per trillion (ppt) ranges. FIG. 3B shows how a 3-class LDA, a supervised dimensionality reduction technique based on R.M.S. , effectively separates odors and the control into distinct clusters. The stimulus window analyzed was 0.25 - 1.0 seconds post-odor onset (n = 34 tetrodes). FIG.3C shows how PC A dimensionality reduction highlights the separation of R.M.S neural trajectory evolution from 0.25 -1.0 seconds after stimulus onset. Colored arrows illustrate the trajectory evolution, aligned at 0.25 seconds after odor presentation (black dot). Trajectories for different concentrations of a single odor follow similar responses (n = 34 tetrodes). FIG. 3D shows how a LOTO confusion matrix classifies R.M.S. responses of the PFAS concentrations, along with a control, achieving 86% accuracy. Classification assignments were based on the smallest Euclidean distance within the 0.25 - 5.75 second window after stimulus onset. FIG. 3E is a sensitivity and specificity table that demonstrates real-time PFAS concentration detection. Values were generated from the EOTO matrix.

[0022] FIG. 4A-C show unique odor-evoked neural responses to pure PFAS compounds. FIG.4A is a schematic illustrating the odor delivery system used for in-vivo neural recordings. Pure PFAS compounds were placed into airtight odor vials. A fixed volume of clean air (E) from a compressed air cylinder flowed into the olfactometer, passing through the odor vial to collect volatile compounds from a single PFAS chemical. The resulting odor-laden air mixture (L-x) was then delivered to the locust antenna. The experimental setup-controlled odor volume and duration. Neural activity was recorded from the AL using a multichannel multielectrode array. An exhaust system removed any residual odor. In FIG. 4B, extracellular voltage traces were spike-sorted to extract neural responses. Two representative neurons displayed distinct odor-evoked activity patterns for all tested PFAS compounds. Raster plots and peri-stimulus time histograms (PSTHs) were generated for each neuron, showing responses to seven PFAS compounds and an empty control vial. Each vertical black line in the raster plots represents a spiking event across five trials per odor condition. PSTHs generated from the raster plots illustrate changes in firing rate over the stimulus duration, with shaded regions indicating the standard error of the mean (S.E.M.). The box marks the odor stimulus window (4 seconds). FIG.4C shows PSTHs of the neuronal population (n = 86) reveal distinct responses to each of the seven PFAS compounds and the control condition. Colored lines represent different odor responses, with shadedregions denoting the S.E.M. The transient response period (0.25 - 1.0 seconds)TEC2025-0129 I TEC2025-0177 I TEC2026-00016550-000538-WO-POAis highlighted, where neuronal activity most effectively differentiates between odors. The box marks the odor stimulus window (4 seconds).

[0023] FIG. 5A-D illustrate the classification of pure PFAS compounds using population neural responses. FIG. 5A shows how PCA was used to reduce dimensionality and visualize population projection neuron (PN) responses to seven pure PFAS compounds and a control. Trajectories represent odor-evoked neural activity between 0.25 - 1.0 seconds after odor exposure. Colored arrows indicate the evolution of these trajectories, with all trajectories aligned at 0.25 seconds after odor onset (black dot), which corresponds to the delay in odor delivery after opening the final valve (0.0s). A greater angular trajectory from the origin and the other odors indicates a more distinct neural response. A total of 86 neurons were analyzed. FIG. 5B shows how linear discriminant analysis (LDA), a supervised dimensionality reduction technique, was applied to classify odor responses between 0.25 - 1.0 seconds post-exposure. LDA separates responses into eight distinct clusters, minimizing within-cluster variance while maximizing separation between different odors. This clustering demonstrates how PFAS compounds influence neural temporal dynamics over the analyzed timeframe. FIG.5C is a high-dimensional LOTO confusion matrix used to classify responses to seven PFAS compounds and the control vial. The matrix compares true versus predicted labels, with the darkened diagonal indicating correct classifications based on the smallest Euclidean distance. Using the entire response window (0.25 - 5.75 seconds), which includes both odor onset and offset responses, the classification achieved an accuracy of 87.08%. This result confirms that the sensor can reliably detect and distinguish between different PFAS compounds. FIG. 5D is a sensitivity and specificity table highlighting precise compound detection using the locust sensor. Values were calculated from the LOTO matrix.

[0024] FIG. 6A-F display real-time classification of environmentally relevant PFAS concentrations using R.M.S. In FIG. 6A, Raoult’s law was used to calculate the ppb concentration ranges for PFAS. The lowest concentration (0.0001%) represents ppt ranges, which mimic those typically found in environmental samples. For FIG. 6B, VOC-evoked extracellular neural voltage traces for representative neurons exposed to PFAS concentrations were spike-sorted. Spike timing is shown in raster plots across five trials, with each black line indicating a single spiking event. PSTHs generated from the raster plots reveal changes in spiking patterns, with shaded regions representing the SEM. The boxes mark the odor stimulus windows (4 seconds). For FIG. 6C, a 3-class LDA was used to separate R.M.S. odor responses between 0.25 - 1.0 seconds post-stimulus onset. Five distinctTEC2025-0129 I TEC2025-0177 I TEC2026-00016550-000538-WO-POAclusters were visualized, demonstrating clear spatiotemporal separation between different odors and concentrations. A total of 34 tetrodes were used in the analysis. In FIG.6D, R.M.S. neural trajectories of two PFAS compounds at two concentrations, along with a control odor (5 odors total), were visualized using PCA dimensionality reduction. The VOC-evoked responses are plotted between 0.25 - 1.0 seconds after stimulus onset. Colored arrows indicate the trajectory evolution of responses. Trajectories are aligned at 0.25 seconds after odor onset (indicated by a black dot), corresponding to the delay after the final valve is opened. An increased angular trajectory from the origin and other odors signifies a distinct neural response. A total of 34 tetrodes were analyzed. FIG. 6E shows a classification matrix based on the LOTO analysis with 84% classification accuracy within the 0.25 -5.75 seconds time window after stimulus onset, demonstrating that the sensor can effectively detect PFAS concentrations. FIG.6F is a sensitivity and specificity table highlighting the precise detection of PFAS compounds and their concentrations. Values were calculated from the LOTO matrix, demonstrating the sensor’s ability to accurately classify PFAS odors and their concentrations.

[0025] FIG. 7A-F represent precision classification of PFOS at environmentally relevant concentrations using R.M.S. For FIG.7A, PFOS concentrations in ppb and ppt were calculated using Raoult’s law, based on chemical characteristics found in the literature. The lowest concentration represents ppt values. For FIG.7B, extracellular voltage traces were spike-sorted to reveal individual neurons. Raster plots show spike timing across five trials, where each black line represents a spiking event. The average firing frequency across the five trials is displayed in the PSTHs, with shaded regions representing the SEM. The boxes indicate the odor stimulus windows (4 seconds). PSTHs illustrate variations in odor-evoked responses across different concentrations and the control. In FIG.7C, a 2-class LDA dimensionality reduction technique was applied to visualize R.M.S. neural response clustering between 0.25 - 1.0 seconds after odor presentation. A total of 34 tetrodes were used in the analysis. In FIG. 7D, PCA dimensionality reduction analysis highlights distinct odor trajectory evolution for two PFOS concentrations, along with a control, between 0.25 - 1.0 seconds after stimulus onset. The separation between PFOS and the control, as well as differentiation in neural response trajectories between concentrations, shows that the locust can distinguish between low PFOS concentrations. Trajectories are aligned at 0.25 seconds after odor onset (black dot), with colored arrows indicating trajectory evolution. Distinct neural responses are shown by increased angular deviation from the origin and separation from other odors. A total of 34 tetrodes were analyzed. FIG. 7E shows a LOTO classification matrix that demonstrates 100% classification accuracy within the 0.25 - 5.75 seconds time window after stimulus onset. FIG. 7F is a sensitivityTEC2025-0129 I TEC2025-0177 I TEC2026-00016550-000538-WO-POAand specificity table showing 100% precision in detecting environmentally relevant PFOS concentrations. Values were calculated from the LOTO matrix.

[0026] FIG. 8A-G illustrate an example concept schematic, microfabricating process flow, and fabrication results. FIG. 8A is a schematic of the flexible dual-sided microelectrode array (MEA), including the MEA tip, cross section of the tip, and stacking layers. FIG. 8B is a schematic of experimental setup where VOCs are delivered to the chemosensory array (antenna) and neural responses are recorded from the antennal lobe region of the locust brain using a flexible dual-sided MEA. A representative voltage trace is shown displaying a 4 second stimulus and VOC-evoked response. FIG.8C is a schematic of the microfabrication method, including (1) 5 pm PA deposition on a silicon wafer, (2) thermal evaporation and patterning of a metal stack of 200 nm Au and 20 nm Ti on PA substrate, (3) deposition of a 5 pm PA encapsulation layer, (4) dry etching of MEAs and the openings of electrodes and contact pads. (5) releasing of single sided MEAs, (6) folding, (7) annealing and bonding to form dual sided structure, (8) assembling of dual-sided MEAs onto PCB.FIG. 8D illustrates microscope images of the MEAs in planar, folded, and bonded states. Scale bar: 100 pm. FIG. 8E shows a mechanical flexible MEA and the tip. FIG. 8F is a graph depicting electrical impedance spectroscopy and FIG. 8G is a graph depicting 1 kHz impedance of the microelectrodes measured throughout the fabrication processes (n=6).

[0027] FIG. 9A-L depict electrode surface modification and performance of the flexible dualsided MEAs. FIG. 9A includes images of microelectrodes including the microscopic images of the PEDOT: PSS coated electrodes (top left) and the SEM images of the electrodes before (Cycle 0) and after 2-cycle or 4-cycle PEDOT:PSS coatings. FIG. 9B-F show characterizations of the microelectrodes without and with PEDOT:PSS deposited through various CV cycles. FIG.9B shows electrochemical impedance change vs frequency, FIG.9C shows electrochemical phase angle change vs frequency, FIG. 9D shows impedance at 1 kHz, FIG. 9E shows surface roughness, and FIG. 9F shows effective areas. FIG. 9G is an insertion test showing MEAs before and after insertion into 0.6% agarose. FIG.9H shows the 1 kHz impedance of electrodes before and after insertion 15 times.FIG. 91 shows the 1 kHz impedance of electrodes before and after bending to 45 and 90 degrees.FIG. 9J shows the 1 kHz impedance of electrodes during 7-day soaking in PBS at room temperature.FIG. 9K shows noise level of PEDOT:PSS coated electrodes vs. bare gold electrodes. FIG.9L shows signal-to-noise ratio of PEDOT:PSS coated electrodes vs. bare gold electrodes.TEC2025-0129 I TEC2025-0177 I TEC2026-00016550-000538-WO-POA

[0028] FIG. 10A-F illustrate how lung cancer biomarkers are distinguished and classified by individual and population neural responses. FIG. 10A is an example schematic of the experimental setup. Beginning at the left, clean air from a compressed air cylinder enters an olfactometer at a flow rate of L and is delivered to the locust antennae. At the start of the odor stimulus, the olfactometer diverts a portion of the clean air into the headspace of an odor vial containing the VOC mixed with mineral oil (1% vol / vol). The odor laden air (x) and the clean air (L-x) are combined at the final valve and delivered to the locust antennae. A flexible dual-sided microelectrode array is inserted into the antennal lobe of the locust brain and extracellular voltage traces are recorded. FIG. 10B is an image of the locust brain with the dual-sided MEA inserted into the AL. The antennal lobe is -500 pm in diameter as shown by the black dashed circles. The tip region of the 8-channel MEA was 230 pm wide and 20 pm thick, with an electrode diameter of 20 pm and a trace spacing of 20 pm. FIG. 10C shows representative extracellular voltage traces from a single location in the locust antennal lobe for each lung cancer biomarker mixed in mineral oil (1% vol / vol) and pure mineral oil is shown. The light grey box indicates the stimulus presentation window (4 seconds). FIG. 10D shows representative raster and peri-stimulus time histograms (PSTHs) for a single neuron spike sorted from the representative voltage traces shown in FIG. 10C. Raster plots contain 5 trials, in which the VOC was delivered to the locust antennae. Each line in the raster plot indicates a spiking event, or action potential, from the neuron. PSTHs show the average changes in spiking rate for the 5 trials shown in the corresponding raster plot. An increase in average firing rate corresponds to an increase in density of the raster plot. The grey box indicates the stimulus presentation window (4 seconds). Trialaveraged PSTHs are plotted with the shaded region indicating the standard error of the mean (SEM).FIG. 10E depicts population neural trajectories of 69 neurons after dimensionality reduction via PCA. The trajectories are plotted between 0.25 - 1.75 seconds after stimulus onset which encompasses the most discriminatory segment of the population neural response. Distinct and separate trajectories indicate differentiation of neural responses to lung cancer biomarkers at the population neuron level. All trajectories are aligned at 0.25 seconds after odor onset (indicated by a black dot), which roughly corresponds to the time of the odor plume hitting the antennae after the opening of the final valve (at 0.0 seconds). FIG. 10F is a confusion matrix displaying the performance of a LOTO analysis for trial-wise classification. 1.5 s trials are classified for each lung cancer biomarker with an accuracy of 100%. Time window duration used for plots shown in FIG. 10E-F is from 0.25-1.75 s after stimulus onset.TEC2025-0129 I TEC2025-0177 I TEC2026-00016550-000538-WO-POA

[0029] FIG. 11A-C depict efficient classification of lung cancer biomarkers from a single recording of locust antennal lobe neurons. FIG. 11A illustrates raster and PSTHs of 8 of the 12 neurons recorded from a single experiment. Raster plots contain 5 trials, in which the biomarker was delivered to the locust antennae. Each line in the raster plot indicates a spiking event, or action potential, from the neuron. PSTHs show the average changes in spiking rate for the 5 trials shown in the corresponding raster plot. The grey box indicates the 4 second stimulus presentation window. Trial- averaged PSTHs are plotted with the shaded region indicating the SEM. FIG.11B shows neural trajectories of the four biomarkers and mineral oil after dimensionality reduction via PCA. Distinct and separate trajectories indicate the differentiation of neural responses to lung cancer biomarkers. The trajectories are plotted combining the population responses between 0.5 - 1.25 and 4.5 - 5 seconds after stimulus onset which includes the two most discriminatory segments of the population neural response. FIG. 11C is a confusion matrix displaying the performance of a LOTO analysis for trial- wise classification. 1.25 s duration trials are classified for each lung cancer biomarker with an accuracy of 88%. Analysis time window duration used in FIG. 11B and FIG. 11C includes population responses between 0.5-1.25 and 4.5-5 s after stimulus onset.

[0030] FIG. 12A-E depict the detection and differentiation of multiple concentrations of lung cancer biomarkers. FIG. 12A is a table of concentrations for the two biomarkers diluted in mineral oil (vol / vol) and the associated gas phase concentrations values in parts per million (ppm) and ppb (see Methods). FIG. 12B illustrates representative raster and PSTHs for a single neuron spike sorted from the representative voltage traces. Raster plots contain 5 trials, in which the VOC was delivered to the locust antennae. Each line in the raster plot indicates a spiking event, or action potential, from the neuron. PSTHs show the average changes in spiking rate for the 5 trials shown in the corresponding raster plot. An increase in average firing rate corresponds to an increase in density of the raster plot. The grey box indicates the stimulus presentation window (4 seconds). Trial- averaged PSTHs are plotted with the shaded region indicating SEM. FIG. 12C shows population neural responses to nonanal and propylbenzene at multiple concentrations using LDA with dimensionality reduction. The time window plotted is from 0.5 to 1.25 seconds after stimulus onset. Separation of clusters indicates differentiation of neural responses to biomarker concentration at the population neuron level. FIG. 12D is a confusion matrix summarizing the performance of a LOTO analysis for trial-wise classification. 0.75 s trials are classified for each lung cancer biomarker at multiple concentrations with an accuracy of 86%. These results showcase the ability of the insect-basedTEC2025-0129 I TEC2025-0177 I TEC2026-00016550-000538-WO-POAbiosensor to detect multiple concentrations of two lung cancer biomarkers. FIG. 12E is a sensitivity and specificity table summarizing the results from the confusion matrix in FIG. 12D.

[0031] FIG. 13A-E depict the detection and differentiation of human lung cancer cell lines using an insect-based bioelectronic sensor. FIG. 13A illustrates representative images of the human lung cancer cell lines. NSC-NCI-H1437-S1; stage 1; adenocarcinoma; non-small cell lung cancer. NSC-NCI-H1573-S4; stage 4; adenocarcinoma; non-small cell lung cancer. SCSHP-77; carcinoma; small cell lung cancer. SC-H69PR; small cell lung cancer. HLF; Primary Lung Fibroblast, Normal, Human. Media was used as a control (image not shown). The black scale bar indicates 100 pm. The photos were captured on the Olympus CKX53 microscope using the Olympus cellSens Entry 2.1 (Build 17342). FIG. 13B includes representative raster and PSTHs for a single neuron spike sorted from the representative voltage traces. Raster plots contain 5 trials, in which the VOC was delivered to the locust antennae. Each line in the raster plot indicates a spiking event, or action potential, from the neuron. PSTHs show the average changes in spiking rate for the 5 trials shown in the corresponding raster plot. An increase in average firing rate corresponds to an increase in density of the raster plot. The grey box indicates the stimulus presentation window (4 seconds). Trial-averaged PSTHs are plotted with the shaded region indicating the SEM. FIG. 13C shows population neural responses to each cell line using LDA with dimensionality reduction. The time window plotted is from 0.25 to 2.5 seconds after stimulus onset. Separation of clusters indicates differentiation of neural responses to each cell line at the population neuron level. FIG. 13D is a confusion matrix summarizing the performance of a LOTO analysis for trial-wise classification of multiple cell lines, which is shown for 2.25 s trials. The analysis classifies the cell lines with an accuracy of 85%. FIG. 13E is a sensitivity and specificity table summarizing the results from the confusion matrix in FIG. 13D.

[0032] FIG. 14A-E illustrate a schematic of the experiment delivering bacterial odors to a locust olfactory sensor and subsequent data analysis for classification. For FIG. 14A, Pseudomonas aeruginosa (PA) and Staphylococcus aureus (SA) were cultured as biofilms (left) and planktonic (right) to prepare the four odor samples. All cultures were prepared in lysogeny broth (LB), which was used as the fifth odor for control. For FIG. 14B, locust, Schistocerca americanus, were used as a biological sensor to detect odors. Locust antennae (analogous to vertebrate noses) detect volatiles in the air. They were prepared via surgery, which removed the exoskeleton and exposed the AL while maintaining the antennal nerve connections to the antennae. A wire platform was placed under the brain to support it and prevent motion during data recording. A reference wire was placed within theTEC2025-0129 I TEC2025-0177 I TEC2026-00016550-000538-WO-POAheadspace but not touching the brain. The electrode was carefully placed within the AL using a micromanipulator. For FIG. 14C, headspace from the bacteria culture flasks were delivered to the locust antennae using an olfactometer for precise odor delivery timing. A cone pulling a slight vacuum was placed behind the locust to quickly clear the odors after delivery. Raw data was collected as voltage recordings using the electrode and time matched to the odor delivery timing. For FIG. 14D, all of the recorded neurons were combined to give the entire population of neurons’ responses to each odor. Each neuron’s response was analyzed as an orthogonal dimension, which together yields a highdimensional response “fingerprint” for each odor. The data was split between training and testing templates for each odor, and then the testing templates were assigned to whichever training template minimized the high-dimensional Euclidean distance. FIG. 14E shows the results of the highdimensional classification summarized in a confusion matrix indicating the percentage of testing templates assigned to each of the training templates.

[0033] FIG. 15A-D show neural responses can differentiate between biofilm and planktonic cultures of Pseudomonas aeruginosa. FIG. 15A shows representative voltage traces of extracellular neural responses to the bacterial odors from the AL. There are three odors, biofilm (PAbio. top), planktonic (PAplank, middle), and lysogeny broth (LB, bottom). The box corresponds with the four-second odor presentation window. Spikes within the voltage trace are due to PN activity. For FIG.15B, the recording location shown in FIG.15A was spike sorted to identify individual PN spike times from the voltage trace. Here, a single PN’s odor-evoked responses are shown with raster and PSTHs. Each vertical black line in the raster plot indicates a single spiking event from the PN, and each row of black lines indicates a single trial (five trials total). The PSTH is the trial- averaged response with the S.E.M. indicated by the shaded region. The box corresponds with the four-second odor presentation window. FIG. 15C includes population PN trajectory plots created using PC A dimensionality reduction to reduce from 32 dimensions (one dimension for each neuron) to three for visualization. The trajectories trace the evolution of the neural response across the first two seconds of the stimulus starting at the origin. The small numbers within the odor space denote seconds after the onset of odor delivery. There is visual differentiation between the population neural responses of each odor within these three dimensions. FIG. 15D illustrates population PN responses classified using a LOTO cross-validation with no dimensionality reduction and summarized in confusion matrices. The spiking responses across the four second stimulus duration are separated into 50ms time bins (80 total). For each odor, one trial is used for testing while the remaining four are averaged together as the training data. This creates a testing template and a training template for each of theTEC2025-0129 I TEC2025-0177 I TEC2026-00016550-000538-WO-POAfive odors at each of the 80 time bins. Within each time bin, the testing templates are compared to the training templates and assigned to the template that minimizes Euclidean distance. In a winner-take-all (WTA) approach, each testing trial is assigned a single classification using the mode of the time bin classifications. The training templates are on the y-axis and the testing templates are on the x-axis. Here, we can see that all trials were correctly classified, yielding 100% accuracy for the detection of biofilm vs. planktonic P. aeruginosa.

[0034] FIG. 16A-D show how the locust olfactory system also responds to odors from Staphylococcus aureus. FIG. 16A illustrates representative voltage traces of extracellular neural responses to the bacterial odors from the AL. There are three odors, biofilm (SAbio, top), planktonic (SAplank, middle), and lysogeny broth (LB, bottom). For FIG. 16B, the recording location shown in FIG. 16A was spike sorted, and a single PN’s odor-evoked responses were visualized using raster and peri-stimulus time histograms (PSTHs). FIG. 16C illustrates population PN trajectory plots created using PCA dimensionality reduction across the first two seconds of the stimuli. The small numbers within the odor space denote seconds after the onset of odor delivery. There is visual differentiation between the population neural responses of each odor within these three dimensions.FIG. 16D shows population PN responses classified with LOTO cross-validation using a highdimensional Euclidean distance metric. Here, it is shown that all the trials were correctly classified, yielding 100% accuracy for the detection of biofilm vs. planktonic P. aeruginosa.

[0035] FIG. 17A-D illustrate how population neural responses uniquely encode each of the bacterial odors. FIG. 17A shows the population neural responses in PCA space all project in unique trajectories across the first two seconds of stimuli. The large differences in the trajectory angles indicate different neural subpopulations encode for the different bacterial species (red: Pseudomonas aeruginosa and blue: Staphylococcus aureus'). The biofilm (dark) and planktonic (light) odors of each species also elicit distinct trajectories from each other. The small numbers within the odor space denote seconds after the onset of odor delivery. For FIG. 17B, using the Euclidean distance, LOTO analysis across the four second odor presentation window, the P. aeruginosa (PAbio) and S. aureus (SAbio) biofilms were both classified with 100% accuracy. P. aeruginosa planktonic (PAplank) was also classified with 100% accuracy. One out of five trials (80%) of S. aureus planktonic (SAplank) was misclassified to S. aureus biofilm. The control, LB broth (LB), was confused with the S. aureus odors in two out of five trials. In FIG. 17C, a high-dimensional hierarchical clustering using Ward’s method shows that the largest difference in odors is between bacterial species with the next majorTEC2025-0129 I TEC2025-0177 I TEC2026-00016550-000538-WO-POAsplit between biofilm and planktonic odors of the same bacterial species. This highlights that differentiating between bacterial species is easier, but locusts can still do the much more difficult task of differentiating biofilm from planktonic bacterial cultures, and they can do this for both bacterial species. For FIG. 17D, a one-vs-rest analysis for specificity was implemented in which each odor is its own correct classification, and all other odors are incorrect, thereby creating multiple binary scenarios for calculating sensitivity and specificity. Very high sensitivity and specificity are seen for each of the odors.

[0036] FIG. 18A-B show that accurate classification occurs rapidly after odor onset and with less than one second of data. In FIG. 18A, LOTO accuracies for different time windows are plotted. The S.E.M. of the accuracies across the five trials is the shaded region. Using a time range of 0.25s (blue), or only five 50ms time bins, we have roughly 20% accuracy during the pre-stimulus windows (-2s to 0s), which we expect for a completely random 5-way classification test. However, very rapidly after odor presentation the accuracy climbs to over 80% after about a second. The same trends hold for the 0.5s (red) and 1.0s (green) time ranges. The rectangles indicate the time windows with the highest accuracy for each time range (from top to bottom are 0.25s, 0.5s, and 1.0s). FIG. 18B illustrates the highest accuracy confusion matrices for each of the time ranges (from left to right are 0.25, 0.5, and Is). The highest accuracy time windows for the three time ranges are centered around Is after stimulus onset, which aligns with the transient neural response. As more data is included, the accuracy increases from 82% to over 90%; slightly better than including neural response data from the entire four-second stimulus (FIG. 18B).

[0037] FIG. 19A-F model the locust neural olfactory sensor’s capabilities. FIG.19A is a receiver operating characteristic (ROC) curve of the locust sensor’s detection of P. aeruginosa biofilm produced by varying the classification threshold. Out of 32 total recorded PNs, random subsets of neurons (10, 20, and 30) were sampled and a clear trend of better sensor performance correlates with more neurons. For each number of neurons, 100 random samples were chosen and averaged together to produce the curves. The dashed line indicates the performance of a completely random binary classifier. In FIG. 19B, for S. aureus biofilm, there is a similar trend of better sensor performance as the number of neurons increases. For FIG. 19C-D, to further explore the trends seen in FIG. 19A-B, the area under the curve (AUC) was compared with the number of neurons sampled. Each dot is the average AUC for 100 random subsamples of the 32 recorded neurons, and the grey shaded region is one standard deviation. An exponential curve was fitted to the data, assuming that the x-intercept isTEC2025-0129 I TEC2025-0177 I TEC2026-00016550-000538-WO-POA0.5, the classification is completely random when there is no neural data, and that the curve should approach an AUC of 1, or perfect classification, as you increase the number of neurons. From both these models, an AUC of over 0.9 is achieved with just 15 neurons. The coefficients of determination (R2) were both greater than 0.95, indicating a strong fit to the model. In FIG. 19E-F, Youden’s Index, a linear combination of sensitivity and specificity, is calculated at different values of the threshold weighting. A threshold weighting of 1 is the same as minimum Euclidean distance. A threshold weighting greater than 1 makes unknown odors less likely to be classified as the odor of interest, PAbio or SAbio. This decreases sensitivity. However, a threshold weighting less than 1 makes unknown odors more likely to be classified as the odor of interest, which decreases specificity instead. For both PAbio and SAbio, the highest Youden’s Index occurs very close to a threshold of 1, or minimum Euclidean distance, and is over 0.9.

[0038] FIG. 20A-C show that air flow and odor delivery to the insect antennae are carefully controlled by the olfactometer. As shown in the example schematic of FIG. 20A, prior to any odor delivery, the insect receives 200 standard cubic centimeters (seem) of clean air via Mass Flow Controller 3 (MFC3). The final valve is a four-way valve that can switch back and forth between delivering the flow from MFC3 or the combined flows from MFC1 and MFC2. While clean air is being delivered to the insect antennae, the valves around the bacterial culture flask are closed to preserve the headspace containing VOCs, and the bypass valve connecting MFC1 and MFC2 is open. As shown in FIG. 20B, during odor delivery to the insect, the final valve switches to deliver the combined airflows from MFC1 and MFC2. The valves around the bacterial culture flask are opened, and the bypass valve is closed to divert the 80 seem of airflow from MFC2 through the bacterial culture flask. This 80 seem of VOC laden- air mixes with the 120 seem from MFC1, and this combined 200 seem of air with bacterial odor in it is delivered to the insect antennae. This valve configuration is maintained for four seconds of the odor delivery before switching back to the clean air delivery setup. This setup maintains a constant airflow of 200 seem to the insect, preventing possible mechanosensory neural responses from biasing the measured odor response. In FIG. 20C, a photoionization detector (PID) was used to confirm the odor delivery profile using the LB control. The olfactometer odor delivery follows a square wave profile.

[0039] FIG. 21A-C illustrate neural trajectories visualized using PCA and also include the temporal changes across time. In FIG. 21A, the neural responses of two neurons to two different odors are visualized using raster plots. Neuron 1 had a much stronger response (more spikes) to OdorTEC2025-0129 I TEC2025-0177 I TEC2026-00016550-000538-WO-POA1 than to Odor 2. Also, Neuron 1’s response to Odor 2 is delayed. Neuron 2 also had different responses between the two odors, with a slightly more delayed response and more consistent spiking to Odor 1. In FIG. 21B, the neural responses to each odor were plotted with each neuron along different axes and the points were connected in temporal order to reveal the neural trajectories to each odor. Clear differences in the curve of each neural response were seen. In FIG. 21C, when using the neural responses from more neurons (n = 32), there are now 32 orthogonal axes, which we cannot visualize. We used PCA to reduce the dimensionality to three for visualization and then connected the points in temporal order to reveal the trajectories.

[0040] FIG. 22A-B show that using independent training and testing templates yields similar classification accuracies as LOTO analysis. Instead of using LOTO (FIG. 15D, FIG. 16D, and FIG.17B), here, training templates were created by averaging two random trials together, and then tested using the remaining three trials. For FIG. 22A, using the WTA approach yielded 80% accuracy. For FIG. 22B, to check the robustness of these results, each of the possible combinations of two training trials was tested and the accuracies were summarized as a histogram. Nine out of the ten possible combinations (five trials, choose two trials) had accuracy between 70-90% with one combination resulting in a lower accuracy of 60%.

[0041] FIG. 23A-B depict varying the classification model threshold. In FIG. 23A, the training templates for a planktonic and a biofilm odor can be visualized as opposite ends of a spectrum. In reality, these templates are in a high-dimensional neural encoding space. An unknown odor, or testing template, is also within this space. The distances between the unknown odor and each of the training templates can be calculated, and then the unknown odor assigned to whichever training template it is closest to. This would use the minimum Euclidean distance metric. However, not all classification tests want to equally weigh a positive (biofilm) and a negative (planktonic) classification. Therefore, the threshold, which was previously in the middle, can be shifted towards the biofilm, which would increase the chance of the unknown odor being classified as planktonic, or it could be shifted in the opposite direction. In FIG.23B, as the threshold is weighed differently, the sensitivity and specificity change. At one extreme, for low thresholds, all unknown odors are assigned to the biofilm training template, which yields a sensitivity of 1, but a specificity of 0. For the opposite extreme, the sensitivity is 0, and the specificity is 1. However, as the threshold shifts, a region in the middle yields high sensitivity and specificity.TEC2025-0129 I TEC2025-0177 I TEC2026-00016550-000538-WO-POA

[0042] FIG. 24 shows a block diagram of an embodiment of the system for detecting per- and polyfluoroalkyl substances, lung cancer, and bacterial biofilms using a bioelectronic sensor.

[0043] In the drawings, reference numbers may be reused to identify similar and / or identical elements.DETAILED DESCRIPTIONA. Introduction

[0044] VOCs are gaseous chemicals that represent the metabolic processes in the body. VOC gas mixtures have distinct smell profiles, and these profiles can be detected using chemical analysis of biological samples. In addition to VOCs, the inventors have discovered that gas-phase PFAS, whether VOCs or not, can also be detected using chemical analysis of environmental samples. The biological olfactory system in insects can be used for detection and classification of complex mixtures of gasphase PFAS and VOCs. For example, studies using gas chromatography-mass spectrometry (GC-MS) and other chemical detection systems have identified changes in VOCs in the breath of patients with diseases.

[0045] The inventors developed an insect brain-based gas sensing system and computational analytical technique for non-invasive detection and diagnosis of PFAS, lung cancer, and / or bacterial biofilms. Biological olfactory systems have evolved to detect minute differences in complex gasphase PFAS and VOC mixtures in natural environments. Here, the challenge of detecting PFAS, lung cancer, and / or bacterial biofilms non-invasively through a novel olfactory neuron-based sensor is addressed, where the capacity of the entire biological olfactory sensory system of an insect brain is leveraged and neural responses are analyzed to discriminate gas mixtures emitted from PFAS. lung cancer, and / or bacterial biofilms vs. non-PFAS, non-lung cancer, and / or non-bacterial biofilms samples.

[0046] Current PFAS detection technologies struggle in detecting trace concentrations found in the environment and require complex data processing, limiting their on-site applicability. For example, conventional gas sensors, such as liquid and gas chromatography-mass spectrometry (LC-MS and GC-MS), struggle with detecting chemicals at ppt levels, require extensive data processing, and lack portability; making them impractical for fieldwork. By leveraging biological chemical sensing systems (locust olfaction), the inventors discovered that broad ranges of chemicals at trace concentrations can be detected. Locusts’ advanced combinatorial coding mechanism enables highlyTEC2025-0129 I TEC2025-0177 I TEC2026-00016550-000538-WO-POAsensitive and specific odor detection. Here, the locust olfactory system is harnessed to differentiate several PFAS. Extracellular neural recordings displayed unique odor-evoked responses for multiple PFAS at environmental concentrations.

[0047] Additionally, early detection of lung cancer significantly enhances treatment outcomes, yet current screening methods are limited by accessibility, sensitivity, and cost. Thus, described herein is a bioelectronic sensing platform that integrates the highly sensitive locust olfactory system with a flexible dual- sided microelectrode array (MEA) for robust, noninvasive, and label-free detection of volatile lung cancer biomarkers. Advanced dimensionality reduction techniques applied to the electrophysiological recordings identified distinct neural response patterns to each VOC biomarker and the complex “scent” emitted from various cell lines. By integrating biological sensory systems with advanced bioelectronics, a novel and efficient approach to lung cancer biomarker detection is provided. It provides a non-invasive. brain-based cancer screening method, offering an accessible and innovative solution for early lung cancer diagnosis.

[0048] Additionally, the inventors discovered that neural responses from the olfactory brain centers of locusts can be used to differentiate between odors produced by biofilm and planktonic cultures. VOCs generated through normal cellular metabolism are excreted into the lysogeny broth (LB) and then enter the gaseous headspace. While there are shared metabolic mechanisms between species of bacteria, there are clear differences in the metabolomes. These differences result in changes to the VOC profile or odor, which are detectable via biological olfactory systems. Changes in gene expression, the proteome, and the metabolome are also evident between biofilms and planktonic bacteria. The example described herein compares the similarity of volatilomes between bacterial species to the similarity between biofilm and planktonic states.

[0049] The results described herein demonstrate that this novel biosensing technology can discriminate between multiple types of PFASs, lung cancers, and bacterial biofilms and non-PFASs, non-lung cancers, and non-bacterial biofilms. Gas-phase PFAS from environmental samples or VOC mixtures emitted from cell cultures or surface samples were delivered to the biological chemosensory array (antennae) and neuronal responses (extracellular neuronal voltage signals) were obtained from the insect (locust) antennal lobe. Individual neurons can distinguish each component by its ‘smell’ (e.g., emitted gas-phase PFAS or VOC mixture). By combining the neural responses across experiments, high-dimensional population neural response templates were obtained that were used to classify unknown gas mixtures to achieve high classification accuracy. This innovative approachTEC2025-0129 I TEC2025-0177 I TEC2026-00016550-000538-WO-POAharnesses the full power of a biological olfactory system for the detection of PFAS, lung cancer, and / or bacterial biofilms via gas-phase PFAS or VOC gas mixture analysis, and presents a novel detection tool to aid in the diagnosis and / or treatment of PFAS, lung cancer, and bacterial biofilms.

[0050] Example embodiments are provided so that this disclosure will be thorough and will fully convey the scope to those who are skilled in the art. Numerous specific details are set forth, such as examples of specific components, methods, and systems, to provide a thorough understanding of embodiments of the present disclosure. It will be apparent to those skilled in the art that specific details need not be employed, that example embodiments may be embodied in many different forms, and that neither should be construed to limit the scope of the disclosure. In some example embodiments, well-known processes, well-known device structures, and well-known technologies are not described in detail.B. Definitions

[0051] The terminology used herein is for the purpose of describing particular example embodiments only and is not intended to be limiting. As used herein, the singular forms “a.” “an,” and “the” may be intended to include the plural forms as well, unless the context clearly indicates otherwise. The terms “comprises,” “comprising,” “including,” and “having,” are inclusive and therefore specify the presence of stated features, elements, compositions, steps, integers, operations, and / or components, but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups thereof. Although the open-ended term “comprising,” is to be understood as a non-restrictive term used to describe and claim various embodiments set forth herein, in certain aspects, the term may alternatively be understood to instead be a more limiting and restrictive term, such as “consisting of’ or “consisting essentially of.” Thus, for any given embodiment reciting compositions, materials, components, elements, features, integers, operations, and / or process steps, the present disclosure also specifically includes embodiments consisting of, or consisting essentially of, such recited compositions, materials, components, elements, features, integers, operations, and / or process steps. In the case of “consisting of,” the alternative embodiment excludes any additional compositions, materials, components, elements, features, integers, operations, and / or process steps, while in the case of “consisting essentially of,” any additional compositions, materials, components, elements, features, integers, operations, and / or process steps that materially affect the basic and novel characteristics are excluded from such an embodiment, but any compositions, materials, components, elements, features, integers, operations,TEC2025-0129 I TEC2025-0177 I TEC2026-00016550-000538-WO-POAand / or process steps that do not materially affect the basic and novel characteristics can be included in the embodiment.

[0052] Any method steps, processes, and operations described herein are not to be construed as necessarily requiring their performance in the particular order discussed or illustrated, unless specifically identified as an order of performance. It is also to be understood that additional or alternative steps may be employed, unless otherwise indicated.

[0053] The use of the term "a" or "an" when used in conjunction with the term "comprising" in the claims and / or the specification may mean "one," but it is also consistent with the meaning of "one or more," "at least one," and "one or more than one." As such, the terms "a," "an," and "the" include plural referents unless the context clearly indicates otherwise. Thus, for example, reference to "a compound" may refer to one or more compounds, two or more compounds, three or more compounds, four or more compounds, or greater numbers of compounds.

[0054] The use of the term "at least one" will be understood to include one as well as any quantity more than one. including but not limited to. 2, 3. 4, 5, 10. 15. 20. 30, 40, 50, 100, etc. The term "at least one" may extend up to 100 or 1000 or more, depending on the term to which it is attached; in addition, the quantities of 100 / 1000 are not to be considered limiting, as higher limits may also produce satisfactory results. In addition, the use of the term "at least one of X, Y, and Z" will be understood to include X alone, Y alone, and Z alone, as well as any combination of X, Y, and Z. The use of ordinal number terminology (i.e., "first," "second." "third," "fourth," etc.) is solely for the purpose of differentiating between two or more items and is not meant to imply any sequence or order or importance to one item over another or any order of addition, for example.

[0055] The use of the term "or" in the claims is used to mean an inclusive "and / or" unless explicitly indicated to refer to alternatives only or unless the alternatives are mutually exclusive. For example, a condition "A or B" is satisfied by any of the following: A is true (or present) and B is false (or not present), A is false (or not present) and B is true (or present), and both A and B are true (or present).

[0056] As used herein, any reference to "one embodiment," "an embodiment," "some embodiments," "one example," "for example," or "an example" means that a particular element, feature, structure, or characteristic described in connection with the embodiment is included in at least one embodiment. The appearance of the phrase "in some embodiments" or "one example" in variousTEC2025-0129 I TEC2025-0177 I TEC2026-00016550-000538-WO-POAplaces in the specification is not necessarily all referring to the same embodiment, for example. Further, all references to one or more embodiments or examples are to be construed as non-limiting to the claims.

[0057] Throughout this disclosure, the term "about" is used to indicate that a value includes the inherent variation of error for a composition / apparatus / device, the method being employed to determine the value, or the variation that exists among the study subjects. For example, but not by way of limitation, when the term "about" is utilized, the designated value may vary by plus or minus twenty percent, or fifteen percent, or twelve percent, or eleven percent, or ten percent, or nine percent, or eight percent, or seven percent, or six percent, or five percent, or four percent, or three percent, or two percent, or one percent from the specified value, as such variations are appropriate to perform the disclosed methods and as understood by persons having ordinary skill in the art. Particularly in reference to a given quantity, number or percentage, “about” is meant to encompass deviations of plus or minus ten percent (± 10). For example, about 5% encompasses any value between 4.5% to 5.5%, such as 4.5, 4.6, 4.7, 4.8, 4.9, 5, 5.1, 5.2, 5.3, 5.4, or 5.5. Accordingly, unless otherwise indicated, the numerical parameters set forth in this specification and attached claims are approximations that can vary depending upon the desired properties sought to be obtained by the presently disclosed subject matter.

[0058] The term "or combinations thereof" as used herein refers to all permutations and combinations of the listed items preceding the term. For example, "A, B, C, or combinations thereof" is intended to include at least one of: A, B, C, AB, AC, BC, or ABC, and if order is important in a particular context, also BA, CA, CB, CBA, BCA, ACB, BAC, or CAB. Continuing with this example, expressly included are combinations that contain repeats of one or more item or term, such as BB, AAA, AAB, BBC, AAABCCCC, CBBAAA, CABABB, and so forth. The skilled artisan will understand that typically there is no limit on the number of items or terms in any combination, unless otherwise apparent from the context.

[0059] As will be understood by one skilled in the art, for any and all purposes, all ranges disclosed herein also encompass any and all possible subranges and combinations of subranges thereof. Furthermore, as will be understood by one skilled in the art, a range includes each individual member.

[0060] The term “olfactory receptor” or “odorant receptor” as used herein refers to chemoreceptors expressed in the cell membranes of olfactory receptor neurons and are responsibleTEC2025-0129 I TEC2025-0177 I TEC2026-00016550-000538-WO-POAfor the detection of odorants (for example, compounds that have an odor, such as VOCs) which give rise to the sense of smell. Activated olfactory receptors trigger nerve impulses that transmit information about odor to the brain. In insects, olfactory receptors are members of an unrelated group of ligand-gated ion channels.

[0061] The term “ vivo" as used herein refers to experiments on living organisms. Thus, in some embodiments, a neural response may be taken from a live insect, such as a locust. The term “zn vitro" as used herein may refer to experiments on non-living organisms or parts of non-living organisms. Thus, in some embodiments, a neural response may be taken from a non-living insect, such as a locust or a locust antenna.

[0062] The terms “per- and polyfluoroalkyl substances” or “PFAS” as used herein refer to a class of synthetic organic compounds characterized by carbon-fluorine bonds, including molecules in which all or part of the carbon chain is fully fluorinated. “PFAS” can refer to any per- and polyfluoroalkyl substance, including but not limited to a perfluoroalkyl acid (PFAA), perfluoroalkyl sulfonamide (FOSA), perfluoroalkyl sulfonamidoethanol (FOSE), fluorotelomer alcohol (FTOH), fluorotelomer based PFAS, or polyfluoroalkyl precursor.

[0063] The term “gas-phase PFAS” as used herein refers to per- and polyfluoroalkyl substances measurable in the vapor phase, including compounds capable of partitioning into the gas phase despite low volatility. Gas-phase PFAS can include both volatile precursors and low-volatility PFAS detectable at trace concentrations.

[0064] The term “low volatility” as used herein refers to compounds having vapor pressures sufficiently low that they do not readily evaporate under ambient conditions, typically exhibiting vapor pressures below about 1x103atm at 20-25 °C, or otherwise volatilizing only at trace levels. Such compounds can still be present in the gas phase at low concentrations.

[0065] The term “lung cancer” as used herein refers to a malignant neoplastic disease arising from the epithelial cells of the lung, characterized by uncontrolled cellular proliferation within pulmonary tissues. “Fung cancer” can include any type of lung cancer, such as non-small cell lung cancer (NSCEC) and small cell lung cancer (SCEC).

[0066] The terms “bacterial biofilms” or “biofilms” as used herein refer to structured communities of bacteria embedded within a self-produced extracellular polymeric matrix and attached to a surface or interface. Examples include, but are not limited to. dental plaque.TEC2025-0129 I TEC2025-0177 I TEC2026-00016550-000538-WO-POAdevice-associated biofilms, environmental surface biofilms, and biofilms present in industrial or water- treatment systems.

[0067] Unless defined otherwise, technical and scientific terms used herein have the same meaning as commonly understood by a person of ordinary skill in the art. In particular, this disclosure utilizes routine techniques in the field of gas-phase substance and VOC detection via bioelectronic sensors.C. Method of Detecting PFAS, Lung Cancer, or Bacterial BiofilmsPFAS

[0068] Provided herein is a method for detecting a presence or absence of a per- or polyfluoroalkyl substance (PFAS). For detecting PFAS, the method includes exposing a biological chemosensory array to 1) at least one test gas-phase PFAS mixture and 2) at least one control gasphase PFAS mixture, at least one control gas-phase non-PFAS mixture, or a combination thereof; obtaining at least one test neuronal response and at least one control neuronal response from the biological chemosensory array; comparing the at least one test neuronal response to the at least one control neuronal response; and outputting a result based on the comparison. The result can indicate a presence or absence of a PFAS or a probability of a presence or absence of a PFAS.

[0069] PFAS may be referred to as “forever chemicals” due to their durability and resistance to degradation. PFAS may be present in a variety of household items, contributing to increased human exposure. PFAS can pose significant health risks due to their bioaccumulation in, for example, environments, animals, agricultural crops, and drinking water. In some embodiments, accumulation of PFAS may be linked to a range of diseases and immunotoxicity, including potential carcinogenic effects.

[0070] The PFAS can be any per- or polyfluoroalkyl substance, and may be any synthetic organic compound containing one or more carbon atoms on which all attached hydrogens are replaced by fluorine. The PFAS can encompass fully fluorinated perfluoroalkyl structures as well as partially fluorinated polyfluoroalkyl structures and their precursor species. In some embodiments, the PFAS may be a perfluoroalkyl acid (PFAA), perfluoroalkyl sulfonamide (FOSA), perfluoroalkyl sulfonamidoethanol (FOSE), fluorotelomer alcohol (FTOH), fluorotelomer based PFAS, polyfluoroalkyl precursor, or a combination thereof. In some embodiments, the PFAA may be perfluorooctanesulfonic acid (PFOS). In some embodiments, the PFAA may be PFOA. In someTEC2025-0129 I TEC2025-0177 I TEC2026-00016550-000538-WO-POAembodiments, the PFAS may be fluorotelomer alcohol (FTOH), fluorotelomer acrylate (FTAcr), methyl perfluorooctane sulfonamide (MeFOSA), perfluorooctane sulfonamide (PFOSA), ethyl perfluorooctane sulfonamidoethanol (EtFOSE), methyl perfluorooctane sulfonamidoethanol (MeFOSE), perfluorooctane sulfonic acid (PFOS), perfluorooctanoic acid (PFOA), or a combination thereof. In some embodiments, the PFAS may be a PFAS regulated by the United States Environmental Protection Agency (EPA) or equivalent agency outside the United States, such as the European Environment Agency (EEA). In some embodiments, the PFAS may be PFOA, PFOS, PFHxS, PFNA, HFPO-DA / GenX, PFBS, or a combination thereof.

[0071] In some embodiments, the probability of a presence of a PFAS may be at least about 1%, 2%, 3%, 4%, 5%, 6%, 7%, 8%, 9%, 10%, 11%, 12%, 13%, 14%, 15%, 16%, 17%, 18%, 19%, 20%, 21%, 22%, 23%, 24%, 25%, 26%, 27%, 28%, 29%, 30%, 31%, 32%, 33%, 34%, 35%, 36%, 37%, 38%, 39%, 40%, 41%, 42%, 43%. 44%, 45%, 46%, 47%. 48%, 49%, 50%, 51%, 52%, 53%, 54%, 55%, 56%, 57%, 58%, 59%, 60%, 61%, 62%, 63%, 64%, 65%, 66%, 67%, 68%, 69%, 70%, 71%, 72%, 73%, 74%. 75%, 76%, 77%, 78%, 79%, 80%, 81%, 82%, 83%, 84%, 85%, 86%, 87%, 88%, 89%, 90%. 91%. 92%, 93%, 94%, 95%. 96%. 97%, 98%, 99%, or 100%. In some embodiments, the probability of a presence of a PFAS may be at least about 40%. In some embodiments, the probability of a presence of a PFAS may be at least about 50%. In some embodiments, the probability of a presence of a PFAS may be at least about 60%. In some embodiments, the probability of a presence of a PFAS may be at least about 70%. In some embodiments, the probability of a presence of a PFAS may be at least about 80%. In some embodiments, the probability of a presence of a PFAS may be at least about 90%. In some embodiments, the probability of a presence of a PFAS may be at least about 99%. In some embodiments, the probability of a presence of a PFAS may be about 100%.

[0072] In some embodiments, the probability of an absence of a PFAS may be at least about 1%, 2%. 3%, 4%. 5%, 6%. 7%, 8%. 9%, 10%. 11%. 12%, 13%, 14%, 15%. 16%. 17%, 18%, 19%, 20%.21%, 22%, 23%, 24%, 25%, 26%, 27%, 28%, 29%, 30%, 31%, 32%, 33%, 34%, 35%, 36%, 37%, 38%, 39%, 40%. 41%, 42%, 43%, 44%, 45%, 46%, 47%, 48%, 49%, 50%, 51%, 52%, 53%, 54%, 55%, 56%, 57%, 58%, 59%, 60%, 61%, 62%, 63%, 64%, 65%, 66%, 67%, 68%, 69%, 70%, 71%, 72%, 73%, 74%, 75%, 76%, 77%, 78%, 79%, 80%, 81%, 82%, 83%, 84%, 85%, 86%, 87%, 88%, 89%, 90%, 91%. 92%, 93%, 94%, 95%, 96%. 97%, 98%, 99%, or 100%. In some embodiments, the probability of an absence of a PFAS may be at least about 40%. In some embodiments, the probability of an absence of a PFAS may be at least about 50%. In some embodiments, the probability of anTEC2025-0129 I TEC2025-0177 I TEC2026-00016550-000538-WO-POAabsence of a PFAS may be at least about 60%. In some embodiments, the probability of an absence of a PFAS may be at least about 70%. In some embodiments, the probability of an absence of a PFAS may be at least about 80%. In some embodiments, the probability of an absence of a PFAS may be at least about 90%. In some embodiments, the probability of an absence of a PFAS may be at least about 99%. In some embodiments, the probability of an absence of a PFAS may be about 100%.

[0073] In some embodiments, the method may include determining the concentration or concentration range of the PFAS. For example, in some embodiments, the PFAS may be detected at an environmental concentration, such as about 0.01% vol / vol to about 0.0001% vol / vol, about 0.01% to about 0.005% vol / vol, about 0.005% to about 0.001% vol / vol, about 0.001% to about 0.0005% vol / vol, about 0.0005% to about 0.0001% vol / vol, about 0.01% to about 0.001% vol / vol, about 0.005% to about 0.0005% vol / vol, or about 0.001% to about 0.0001% vol / vol. In certain embodiments, the PFAS may be detected at an environmental concentration of about 0.01% vol / vol to about 0.0001% vol / vol. In some embodiments, PFAS can be detected at concentrations as low as about 34ppt within the tested range corresponding to approximately 0.01% to 0.0001% vol / vol. In some embodiments, concentrations above 0.01% vol / vol were also evaluated, and the sensor demonstrated a lower detection capability reaching down to approximately 34ppt and reliably detecting any concentration above this threshold. In some embodiments, PFAS can be detected at about 34ppt to about lOOppt, about lOOppt to about 500 ppt, about 500 ppt to about 1 ppb, about 1 ppb to about 10 ppb, about 10 ppb to about 50 ppb, about 50 ppb to about 100 ppb, about 34 ppt to about 1 ppb, about 1 ppb to any concentration above 1 ppb. In some embodiments, PFAS can be detected at about 85 ppb. In some embodiments, PFAS can be detected at about 34 ppt.Lung Cancer

[0074] Provided herein is a method for detecting a presence or absence of lung cancer. For detecting lung cancer, the method includes exposing a biological chemosensory array to 1) at least one test volatile organic compound (VOC) mixture and 2) at least one control VOC mixture, at least one control non- VOC mixture, or a combination thereof; obtaining at least one test neuronal response and at least one control neuronal response from the biological chemosensory array; comparing the at least one test neuronal response to the at least one control neuronal response; and outputting a result based on the comparison. The result can indicate a presence or absence of lung cancer or a probability of a presence or absence of lung cancer.TEC2025-0129 I TEC2025-0177 I TEC2026-00016550-000538-WO-POA

[0075] Lung cancer is a malignant neoplastic condition of the respiratory epithelium, and can be characterized by, for example, genomic instability, aberrant signaling pathways, progressive infiltration of surrounding tissues, or a combination thereof. In some embodiments, the lung cancer may be non small cell lung cancer (NSCLC), small cell lung cancer (SCLC), or a combination thereof. In some embodiments, NSCLC and SCLC may have differences in cellular morphology, growth kinetics, therapeutic susceptibility, or a combination thereof.

[0076] In some embodiments, the probability of a presence of lung cancer may be at least about 1%, 2%, 3%, 4%, 5%, 6%, 7%, 8%, 9%, 10%, 11%, 12%, 13%, 14%, 15%, 16%, 17%, 18%, 19%, 20%, 21%, 22%, 23%, 24%, 25%, 26%, 27%, 28%, 29%. 30%, 31%, 32%, 33%, 34%, 35%, 36%, 37%, 38%, 39%, 40%, 41%, 42%, 43%, 44%, 45%, 46%, 47%, 48%, 49%, 50%, 51%, 52%, 53%, 54%, 55%, 56%, 57%, 58%, 59%, 60%, 61%, 62%, 63%, 64%, 65%, 66%, 67%, 68%, 69%, 70%, 71%, 72%, 73%, 74%, 75%, 76%. 77%, 78%, 79%, 80%. 81%, 82%, 83%, 84%, 85%, 86%, 87%, 88%, 89%, 90%, 91%, 92%, 93%, 94%, 95%, 96%, 97%, 98%, 99%, or 100%. In some embodiments, the probability of a presence of lung cancer may be at least about 40%. In some embodiments, the probability of a presence of lung cancer may be at least about 50%. In some embodiments, the probability of a presence of lung cancer may be at least about 60%. In some embodiments, the probability of a presence of lung cancer may be at least about 70%. In some embodiments, the probability of a presence of lung cancer may be at least about 80%. In some embodiments, the probability of a presence of lung cancer may be at least about 90%. In some embodiments, the probability of a presence of lung cancer may be at least about 99%. In some embodiments, the probability of a presence of lung cancer may be about 100%.

[0077] In some embodiments, the probability of an absence of lung cancer may be at least about 1%, 2%, 3%, 4%, 5%. 6%, 7%, 8%, 9%, 10%, 11%, 12%. 13%, 14%, 15%, 16%. 17%, 18%, 19%, 20%, 21%. 22%, 23%, 24%, 25%. 26%, 27%, 28%, 29%. 30%, 31%, 32%. 33%, 34%, 35%, 36%.37%, 38%, 39%, 40%, 41%, 42%, 43%, 44%, 45%, 46%, 47%, 48%, 49%, 50%, 51%, 52%, 53%, 54%, 55%, 56%. 57%, 58%, 59%, 60%, 61%, 62%, 63%, 64%, 65%, 66%, 67%, 68%, 69%, 70%, 71%, 72%, 73%, 74%, 75%, 76%, 77%, 78%, 79%, 80%, 81%, 82%, 83%, 84%, 85%, 86%, 87%, 88%, 89%, 90%, 91%, 92%, 93%, 94%, 95%, 96%, 97%, 98%, 99%, or 100%. In some embodiments, the probability of an absence of lung cancer may be at least about 40%. In some embodiments, the probability of an absence of lung cancer may be at least about 50%. In some embodiments, the probability of an absence of lung cancer may be at least about 60%. In some embodiments, theTEC2025-0129 I TEC2025-0177 I TEC2026-00016550-000538-WO-POAprobability of an absence of lung cancer may be at least about 70%. In some embodiments, the probability of an absence of lung cancer may be at least about 80%. In some embodiments, the probability of an absence of lung cancer may be at least about 90%. In some embodiments, the probability of an absence of lung cancer may be at least about 99%. In some embodiments, the probability of an absence of lung cancer may be about 100%.Bacterial Biofilms

[0078] Provided herein is a method for detecting a presence or absence of a bacterial biofilm. For detecting a bacterial biofilm, the method includes exposing a biological chemosensory array to 1) at least one test volatile organic compound (VOC) mixture and 2) at least one control VOC mixture, at least one control non-VOC mixture, or a combination thereof; obtaining at least one test neuronal response and at least one control neuronal response from the biological chemosensory array; comparing the at least one test neuronal response to the at least one control neuronal response; and outputting a result based on the comparison. The result can indicate a presence or absence of a bacterial biofilm or a probability of a presence or absence of a bacterial biofilm.

[0079] Bacterial biofilms are multicellular, heterogeneous colonies that can form on solid surfaces and create a complex extracellular environment, constituting, but not limited to, a polysaccharide matrix filled with extracellular DNA (eDNA). The combination of the extracellular matrix and heterogeneous nature can make it difficult for antibiotics to completely eliminate a biofilm. Some almost dormant bacteria buried within the matrix, known as persisters. have slower metabolisms that enable them to survive and recreate the biofilm post-treatment. Biofilm formation can enable bacteria to adhere to biotic or abiotic surfaces, communicate through coordinated signaling, and / or exhibit increased tolerance to antimicrobial agents and environmental stresses. Biofilms are widely observed in clinical, industrial, and environmental settings and represent a persistent challenge for detection and control. Bacteria emit volatile organic compounds (VOCs) that can be targeted for disease detection.

[0080] In some embodiments, the VOC mixture is emitted from one or more environmental sample that may have a bacterial biofilm formed on the surface of the sample. In some embodiments, the bacterial biofilm may be present on any solid, semi-solid, or hydrated surface capable of supporting microbial attachment and extracellular polymeric substance (EPS) production. Exemplary surfaces include, but are not limited to, metallic surfaces, polymeric surfaces, glass surfaces, ceramic surfaces, mineral surfaces, or composite materials. Additional examples include surfaces of medicalTEC2025-0129 I TEC2025-0177 I TEC2026-00016550-000538-WO-POAdevices, industrial equipment, water distribution infrastructure, food-processing equipment, HVAC components, or environmental substrates such as soil aggregates, sediment particles, plant tissue, or biological tissue. In some embodiments, the VOC mixture may be emitted from a biofilm formed on the surface of a biological sample, including but not limited to lung tissue, airway epithelial tissue, sputum, mucus, or ex vivo cultured tissue models. In some embodiments, the VOC mixture may be emitted from a biofilm formed on the surface of an environmental or industrial sample, such as a pipe interior, filtration membrane, storage tank, or wetted surface exposed to microbial colonization.

[0081] In some embodiments, the probability of a presence of a bacterial biofilm may be at least about 1%, 2%, 3%, 4%. 5%, 6%. 7%, 8%, 9%, 10%. 11%, 12%, 13%, 14%, 15%, 16%, 17%, 18%, 19%, 20%, 21%, 22%, 23%, 24%, 25%, 26%, 27%, 28%, 29%, 30%, 31%, 32%, 33%, 34%, 35%, 36%, 37%, 38%, 39%, 40%, 41%, 42%, 43%, 44%, 45%, 46%, 47%, 48%, 49%, 50%, 51%, 52%, 53%, 54%, 55%, 56%, 57%, 58%. 59%, 60%, 61%, 62%. 63%, 64%, 65%, 66%. 67%, 68%, 69%, 70%, 71%, 72%, 73%, 74%, 75%, 76%, 77%, 78%, 79%, 80%, 81%, 82%, 83%, 84%, 85%, 86%, 87%, 88%, 89%, 90%, 91%, 92%, 93%, 94%, 95%, 96%, 97%, 98%, 99%, or 100%. In some embodiments, the probability of a presence of a bacterial biofilm may be at least about 40%. In some embodiments, the probability of a presence of a bacterial biofilm may be at least about 50%. In some embodiments, the probability of a presence of a bacterial biofilm may be at least about 60%. In some embodiments, the probability of a presence of a bacterial biofilm may be at least about 70%. In some embodiments, the probability of a presence of a bacterial biofilm may be at least about 80%. In some embodiments, the probability of a presence of a bacterial biofilm may be at least about 90%. In some embodiments, the probability of a presence of a bacterial biofilm may be at least about 99%. In some embodiments, the probability of a presence of a bacterial biofilm may be about 100%.

[0082] In some embodiments, the probability of an absence of a bacterial biofilm may be at least about 1%, 2%. 3%, 4%. 5%, 6%, 7%, 8%, 9%, 10%. 11%, 12%, 13%, 14%. 15%, 16%, 17%, 18%.19%, 20%, 21%, 22%, 23%, 24%, 25%, 26%, 27%, 28%, 29%, 30%, 31%, 32%, 33%, 34%, 35%, 36%, 37%, 38%, 39%, 40%, 41%, 42%, 43%, 44%, 45%, 46%, 47%, 48%, 49%, 50%, 51%, 52%, 53%, 54%, 55%, 56%, 57%, 58%, 59%, 60%, 61%, 62%, 63%, 64%, 65%, 66%, 67%, 68%, 69%, 70%, 71%, 72%, 73%, 74%, 75%, 76%, 77%, 78%, 79%, 80%, 81%, 82%, 83%, 84%, 85%, 86%, 87%, 88%. 89%, 90%, 91%, 92%, 93%, 94%, 95%. 96%, 97%. 98%, 99%, or 100%. In some embodiments, the probability of an absence of a bacterial biofilm may be at least about 40%. In some embodiments, the probability of an absence of a bacterial biofilm may be at least about 50%. In someTEC2025-0129 I TEC2025-0177 I TEC2026-00016550-000538-WO-POAembodiments, the probability of an absence of a bacterial biofilm may be at least about 60%. In some embodiments, the probability of an absence of a bacterial biofilm may be at least about 70%. In some embodiments, the probability of an absence of a bacterial biofilm may be at least about 80%. In some embodiments, the probability of an absence of a bacterial biofilm may be at least about 90%. In some embodiments, the probability of an absence of a bacterial biofilm may be at least about 99%. In some embodiments, the probability of an absence of a bacterial biofilm may be about 100%.Gas-phase mixture or VOC mixture and olfactometer

[0083] In the context of this disclosure, a gas-phase mixture refers broadly to any mixture in which one or more chemical species are present in the vapor phase, regardless of their intrinsic volatility. This category includes compounds that appear in the gas phase at trace or low concentrations due to volatilization, desorption, aerosolization, or partitioning mechanisms. Gas-phase PFAS fall within this broader category, as certain per- and polyfluoroalkyl substances can be detected in the vapor phase even when their inherent vapor pressures are extremely low. Low volatility compounds have vapor pressures sufficiently low that they do not readily evaporate under ambient conditions, typically exhibiting vapor pressures below about 1x103atm at 20-25 °C, or otherwise volatilizing only at trace levels.

[0084] A volatile organic compound (VOC) mixture is a specific subset of gas-phase mixtures composed of volatile organic compounds (i.e., organic chemicals that readily evaporate under ambient conditions and exhibit measurable vapor pressures). VOC mixtures are commonly associated with applications involving lung cancer research, environmental exposure studies, and bacterial biofilm formation, where volatile organic species are relevant to inhalation pathways or microbial interactions.

[0085] Thus, while all VOC mixtures are gas-phase mixtures, not all gas-phase mixtures are VOC mixtures. Gas-phase PFAS mixtures are distinct in that they may contain low-volatility or semi-volatile PFAS species that do not meet regulatory or chemical definitions of VOCs but are nonetheless detectable in the vapor phase.

[0086] As used herein, the term “gas-phase PFAS mixture” and “VOC mixture” may be referring to the test mixture, one or more control mixture, or both, unless specified.

[0087] VOCs are organic compounds that have a high vapor pressure at room temperature. VOCs may be produced by one or more biological processes in an organism and can be released to theTEC2025-0129 I TEC2025-0177 I TEC2026-00016550-000538-WO-POAexterior in several ways. For example, VOCs may be emitted from an organism via exhaled breath, perspiration, urine, feces, or lacrimal fluid.

[0088] In some embodiments, a VOC mixture may be a VOC gas or liquid mixture. In certain embodiments, the VOC mixture may be a VOC gas mixture.

[0089] In some embodiments, the test gas-phase PFAS mixture may be emitted from one or more environmental sample. In some embodiments, the environmental sample may be air, water, soil, sediment, dust, a biosolid, or a biological sample. In some embodiments, the test VOC mixture may be emitted from a biological sample. In some embodiments, the biological sample may be breath, urine, sweat, or blood. In some embodiments, a gas-phase PFAS mixture may contain at least 2, 5, 10, 15, 20, 25, 30, 35, 40, 45, 50, 55, 60, 65, 70, 75, 80, 85, 90, 95, 100, 200, 300. 400, 500, 600, 700, 800, 900, or 1,000 PFAS. In some embodiments, a VOC mixture may contain at least 2, 5, 10, 15, 20, 25, 30, 35, 40, 45, 50, 55, 60, 65, 70, 75, 80, 85, 90, 95, 100, 200, 300, 400, 500, 600, 700, 800, 900, or 1,000 VOCs.

[0090] In some embodiments, the gas-phase PFAS mixture or VOC mixture may be in a container or vessel, such as a cell culture flask. In some embodiments, the gas-phase PFAS mixture or VOC mixture may be delivered to the biological chemosensory array via one or more tubes. In some embodiments, the gas-phase PFAS mixture or VOC mixture may be delivered to the biological chemosensory array via an olfactometer system.

[0091] In some embodiments, the olfactometer system may be a flow-olfactometer system. In some embodiments, air may flow continuously through or over a biological sample emitting a gasphase PFAS mixture or VOC mixture, and is then transported (e.g., via tubing) to the biological chemosensory array. Thus, the olfactometer system may allow precision cell culture headspace stimulus delivery. The headspace refers to air / space immediately above, for example within about 0-24 inches above, the biological sample containing a gas-phase PFAS mixture or VOC mixture. In some embodiments, the headspace is contained in a container or vessel with the biological sample. In some embodiments, the olfactometer may use vacuum pressure to switch between test gas-phase PFAS mixture or test VOC mixture flow and zero contaminant air (i.e., does not contain gas-phase PFAS mixture or VOCs). In some embodiments, air flow to the biological chemosensory array may be continuous.TEC2025-0129 I TEC2025-0177 I TEC2026-00016550-000538-WO-POA

[0092] In some embodiments, the olfactometer system may include a stimulus air flow line and a dilution air flow line. In some embodiments, the stimulus air flow line may include a gas-phase PFAS mixture or a VOC mixture. In some embodiments, the dilution air flow line may be added to the stimulus air flow line to dilute the gas-phase PFAS mixture or VOC mixture. In some embodiments, the dilution air flow line may include zero contaminant air. In some embodiments, the dilution air flow line may be added upstream of the cell culture vessel. In some embodiments, the dilution air flow line may be added downstream of the cell culture vessel.

[0093] In some embodiments, air may pass through the flow line(s) at about 150-400 standard cubic centimeters per minute (seem). In certain embodiments, air may pass through the flow line(s) at about 200 seem. In some embodiments, the air flow line is about 1 / 8 in. to about 1 / 16 in. in diameter. In certain embodiments, the air flow line is about 1 / 16 in. in diameter. In some embodiments, the end of the stimulus air flow line is placed about 0.1, 0.2, 0.3. 0.4, 0.5, 0.6, 0.7, 0.8, 0.9, 1.0, 1.1, 1.2, 1.3, 1.4, 1.5, 1.6, 1.7, 1.8, 1.9, 2.0, 2.1, 2.2, 2.3, 2.4, 2.5, 2.6, 2.7, 2.8, 2.9, 3.0, 3.1, 3.2, 3.3, 3.4, 3.5, 3.6, 3.7, 3.8, 3.9, 4.0, 4.1, 4.2, 4.3, 4.4, 4.5, 4.6, 4.7, 4.8, 4.9, 5.0, 5.5, 6.0, 6.5, 7.0, 7.5, 8.0, 8.5, 9.0. 9.5, or 10 cm from the biological chemosensory array. In certain embodiments, the end of the stimulus air flow line may be placed about 2-3 cm from the biological chemosensory array. In certain embodiments, the stimulus flow line may be placed about 2-3 cm from the most distal part of the biological chemosensory array. In some embodiments, the stimulus air flow line may be delivered to the biological chemosensory array for about 1-5 seconds. In certain embodiments, the stimulus air flow line may be delivered to the biological chemosensory array for about 4 seconds. In some embodiments, stimulus delivery may be repeated at least 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, or 25 times. In certain embodiments, stimulus delivery may be repeated at least 5 times. In some embodiments, the interstimulus interval (i.e., time between two stimulus (i.e., test gas-phase PFAS mixture or test VOC mixture) delivery intervals) may be at least about 1, 5, 10, 15, 20, 25, 30, 35, 40, 45. 50, 55, 60, 65, 70, 75, 80, 85, or 90 seconds. In certain embodiments, the interstimulus delivery may be about 60 seconds. In some embodiments, before and / or after stimulus air flow (i.e., test gas-phase PFAS mixture or test VOC mixture air flow) is delivered to the biological chemosensory array, constant air flow may be maintained that does not contain the test gas-phase PFAS mixture or test VOC mixture (e.g., zero contaminant air). Thus, in some embodiments, the stimulus air flow delivery may be “turned on” and / or “turned off” while the zero contaminant air maintains a constant flow. In some embodiments, keeping a constant air flow may reduce confounding neuronal responses due to changes in air pressure via mechanosensoryTEC2025-0129 I TEC2025-0177 I TEC2026-00016550-000538-WO-POAdetection. In some embodiments, a funnel positioned behind the biological chemosensory array may be used to remove gas -phase PFASs or VOCs. In some embodiments, the funnel may include a vacuum. In some embodiments, the funnel may be about 2-10 inches in diameter. In certain embodiments, the funnel may be about 6 inches in diameter.Biological sample

[0094] In some embodiments, the method of detection may be non-invasive. In other words, in some embodiments, the method of detection may not introduce instruments into a subject. Instead, a biological sample may be taken from a subject to use in the disclosed method.

[0095] Thus, in some embodiments, the VOC mixture may be emitted from one or more biological sample. In some embodiments, the VOC mixture may be emitted from one or more lung cancer biological sample. In some embodiments, the VOC mixture may be emitted from one or more non-lung cancer (e.g., healthy) biological sample. In some embodiments, the test VOC mixture may be obtained from one or more lung cancer biological sample. In some embodiments, the control VOC mixture may be obtained from one or more non-lung cancer biological sample.

[0096] In some embodiments, the biological sample may be cheek tissue, blood (e.g., whole blood, plasma, dried blood spot, etc.), organ tissue, feces, skin, hair, breath, urine, sweat, or a combination thereof. In some embodiments, the biological sample may be breath, urine, sweat, blood, or a combination thereof. In certain embodiments, the biological sample may be breath.

[0097] “Subject,” ‘ ‘individual,” and “patient” interchangeably refer to a mammal, for example, a human or a non-human primate, but also domesticated mammals (e.g., canine or feline), laboratory mammals (e.g., mouse, rat, rabbit, hamster, guinea pig), and agricultural mammals (e.g., equine, bovine, porcine, ovine). In some embodiments, the subject may be human (e.g., adult female, adult male, adolescent female, adolescent male, female child, male child). Alternatively, in some embodiments, the subject may be a non-human animal. For example, in some embodiments, the non-human animal may be a mouse or primate. In certain embodiments, the subject can be under the care of a physician or other health worker. In certain embodiments, the subject may not be under the care of a physician or other health worker.

[0098] In some embodiments, a subject may be suffering from lung cancer. In some embodiments, the subject may have a history of lung cancer. In some embodiments, the subject may present symptoms such as persistent cough, chest pain, shortness of breath, unexplained weight loss.TEC2025-0129 I TEC2025-0177 I TEC2026-00016550-000538-WO-POAhemoptysis, recurrent respiratory infections, or a combination thereof. In some embodiments, the subject may exhibit radiographic abnormalities or other clinical indicators consistent with pulmonary malignancy.Environmental Sample

[0099] In some embodiments, the PFAS mixture may be taken from an environmental sample, such as air, water, soil, sediment, dust, a biosolid, or a biological sample. In some embodiments, the environmental sample may include any gas-phase, liquid-phase, or solid-phase matrix encountered in environmental, industrial, or biological settings. In some embodiments, air samples may include ambient air, indoor air, stack emissions, headspace samples, or a combination thereof. In some embodiments, water samples may include surface water, groundwater, drinking water, wastewater, stormwater, process water, or a combination thereof. In some embodiments, soil and sediment samples may encompass natural soils, contaminated soils, subsurface materials, aquatic sediments, or a combination thereof. In some embodiments, dust samples may include airborne particulate matter, settled dust, particulate residues collected from surfaces, or a combination thereof. In some embodiments, biosolid samples may include sludge, treated biosolids, other solid residues generated during wastewater treatment, or a combination thereof. In some embodiments, biological samples may include plant tissue, animal tissue, microbial cultures, other biologically derived matrices capable of containing an analyte of interest, or a combination thereof.Biological chemosensory array

[0100] The biological chemosensory array includes one or more insect antenna. Thus, in some embodiments, the insect may be any insect with one or more antenna.

[0101] Thus, the insect may belong to the order Protura, Collembola, Diplura, Microcoryphia, Thysanura, Ephemeroptera, Odonata, Orthoptera, Phasmatodea, Grylloblattodea, Mantophasmatodea, Dermaptera, Plecoptera, Embiidina, Zoraptera, Isoptera, Mantodea, Blattodea, Hemiptera, Thysanoptera, Psocoptera, Phthiraptera, Coleoptera, Neuroptera, Hymenoptera, Trichoptera, Lepidoptera, Siphonaptera, Mecoptera, Strepsiptera, or Diptera.

[0102] In certain embodiments, the insect may belong to the order Orthoptera, Hymenoptera, or Diptera. In some embodiments, the insect belonging to the order Orthoptera may be a cricket, grasshopper, or locust. In some embodiments, the insect may belong to the order Acrididae and / or theTEC2025-0129 I TEC2025-0177 I TEC2026-00016550-000538-WO-POAsuborder Caelifera. In certain embodiments, the locust may be a post-fifth instar locust (Schistocerca americand).

[0103] In insects, olfactory receptors are the functional receptor complexes expressed on olfactory sensory neurons that detect and respond to odorant molecules. These receptors typically consist of a ligand-specific odorant receptor subunit paired with the conserved odorant receptor co-receptor (Oreo). As used herein, “olfactory receptor types” refers to the distinct gene-encoded classes of odorant receptors, each representing a unique receptor sequence that confers differential odorant sensitivity. Thus, while an organism may express numerous individual olfactory receptors across its sensory neurons, these receptors arise from a finite set of olfactory receptor types encoded in the genome. In some embodiments, the locust olfactory system may comprise about 142 olfactory receptors derived from a repertoire of about 60 to 80 olfactory receptor types, together with a single conserved Oreo type.

[0104] Thus, in some embodiments, locusts may have about 142 olfactory receptors. In some embodiments, the insect belonging to the order Hymenoptera may be a honeybee or an ant. In some embodiments, honeybees may have about 170 olfactory receptors. In some embodiments, ants may have about 400 olfactory receptors. In some embodiments, the insect belonging to the order Diptera may be a fruit fly. In some embodiments, the insect may belong to the Family Drosophilidae. In some embodiments, fruit flies may have about 60 olfactory receptors.

[0105] In some embodiments, the insect, insect head, and / or insect antennae may be stabilized via a stabilizing component. In some embodiments, the stabilizing component may be a surgical platform. In some embodiments, the insect may be stabilized and / or immobilized with wax. In some embodiments, the exoskeleton between the antennae may be removed. In some embodiments, glandular tissue may be removed until the insect brain is fully visible. In some embodiments, treatment with protease may be used to remove the neural sheath on the antennal lobe.Olfactory receptors

[0106] In some embodiments, the biological chemosensory array (antenna) may have olfactory receptors, also known as odorant receptors, which are chemoreceptors expressed in the cell membranes of olfactory receptor neurons and are responsible for the detection of PFASs or VOCs. Olfactory receptor neurons contain olfactory receptor neurons. A single olfactory receptor neuron may contain multiple olfactory receptors. In some embodiments, the terms “olfactory receptorTEC2025-0129 I TEC2025-0177 I TEC2026-00016550-000538-WO-POAneuron” and “olfactory receptor” may be used interchangeably. Thus, olfactory receptors may be used to detect PFASs or VOCs in control and / or test PFAS or VOC mixtures.

[0107] In some embodiments, the insect may have at least about 50, 51, 52, 53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63, 64, 65. 66. 67. 68. 69. 70. 71, 72, 73, 74, 75, 76, 77, 78, 79, 80, 81, 82. 83. 84.85, 86, 87, 88, 89, 90, 91, 92, 93, 94, 95, 96, 97, 98, 99, 100, 101, 102, 103, 104, 105, 106, 107, 108, 109. 110, 111, 112, 113, 114, 115, 116, 117, 118. 119, 120, 121, 122, 123, 124, 125, 126, 127, 128, 129, 130, 131, 132, 133, 134, 135, 136, 137, 138, 139, 140, 141, 142, 143, 144, 145, 146, 147, 148, 149, 150, 151, 152, 153, 154, 155, 156, 157, 158, 159, 160, 161, 162, 163, 164, 165, 166, 167, 168, 169, 170, 171, 172, 173, 174. 175, 176, 177, 178, 179, 180, 181, 182, 183. 184, 185, 186, 187, 188, 189, 190, 191, 192, 193, 194, 195, 196, 197, 198, 199, 200, 210, 220, 230, 240, 250, 260, 270, 280, 290, 300, 310, 320, 330, 340, 350, 360, 370, 380, 390, 400, 410, 420, 430, 440, 450, 460, 470, 480, 490, or 500 olfactory receptors. In some embodiments, the insect may have at least about 142 olfactory receptors. In some embodiments, the insect may have at least about 170 olfactory receptors. In some embodiments, the insect may have at least about 400 olfactory receptors. In some embodiments, the insect may have at least about 60 olfactory receptors. In certain embodiments, the insect may have at least about 100 olfactory receptors.Neuronal response

[0108] In some embodiments, the at least one test neuronal response and the at least one control neuronal response may be obtained from one or more insect antennal lobes. In some embodiments, the test or control neuronal response may be obtained from one insect antennal lobe. Thus, in some embodiments, a neuronal response may be triggered by contact of one or more olfactory receptors with one or more PFASs or VOCs in a test or control mixture.

[0109] In some embodiments, the at least one test neuronal response and the at least one control neuronal response may be extracellular neuronal voltage signals. In some embodiments, extracellular recording may use an electrode probe inserted into living or non-living tissue to measure electrical activity coming from adjacent cells, such as neurons. In some embodiments, the probe may be inserted into the antennal lobe of an insect, such as a locust. In certain embodiments, a 16-channel silicon probe with impedances between 100 and 400 kQ may be inserted into the antennal lobe for all neural recordings. In certain embodiments, a 16-channel silicon probe with impedances between 200 and 300 k may be inserted into the antennal lobe for all neural recordings.TEC2025-0129 I TEC2025-0177 I TEC2026-00016550-000538-WO-POA

[0110] In some embodiments, the at least one test neuronal response and the at least one control neuronal response may be obtained from an insect in vivo. In other words, in some embodiments, the neuronal response may be taken from a live insect (e.g.. a locust). Thus, in some embodiments, the biological chemosensory array (antenna) may not be detached from the insect. In some embodiments, the at least one test neuronal response and the at least one control neuronal response may be obtained from an insect in vitro. In other words, in some embodiments, the neuronal response may be taken from a non-live insect (e.g., a locust). Thus, in some embodiments, the biological chemosensory array (antenna) may be detached from the insect.

[0111] In some embodiments, the olfactory receptors may transform chemical stimuli (e.g., via contact with one or more gas-phase PFASs or VOCs) into electrical signals that are transmitted to the antennal lobe, which contains a dense interconnected network of excitatory projection neurons and inhibitory local neurons. In some embodiments, voltage signals may be sampled at about 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, or 50 kHz and then digitized. In some embodiments, voltage signals may be sampled at about 20 kHz. In some embodiments, the digitized signals may be transmitted to a recording controller before being visualized and / or stored.

[0112] In some embodiments, in the antennal lobe, neural codes may take the form of spatiotemporal patterns of activity distributed across the projection neurons and inhibitory local neurons. Thus, the collective gas-phase PFAS- or VOC-evoked population response may contain information about the gas-phase PFAS or VOC identity, intensity, timing, or a combination thereof. In some embodiments, Thus, the collective gas-phase PFAS- or VOC-evoked population response may be reliable over repeated trials.

[0113] In some embodiments, one or more neuronal response may be recorded. In some embodiments, a neuronal response recording may be obtained before, during, or after stimulus delivery. In some embodiments, a neuronal response recording may be obtained during stimulus delivery. In other words, a neuronal response recording may be taken during the timeframe in which a test gas-phase PFAS mixture or test VOC mixture is delivered to the biological chemosensory array. In some embodiments, neuronal responses recorded during test gas-phase PFAS mixture or test VOC mixture delivery may be referred to as a test neuronal response.

[0114] In some embodiments, a neuronal response may be recorded before, during, or after the timeframe in which a control gas-phase or control VOC mixture is delivered to the biologicalTEC2025-0129 I TEC2025-0177 I TEC2026-00016550-000538-WO-POAchemosensory array. Tn some embodiments, a neuronal response may be recorded during the timeframe in which a control gas-phase or control VOC mixture is delivered to the biological chemosensory array. In some embodiments, neuronal responses recorded during control gas-phase or control VOC mixture delivery may be referred to as a control neuronal response. As mentioned herein, the control gas-phase or control VOC mixture may be a control gas-phase PFAS mixture, a control gas-phase non-PFAS mixture, or a combination thereof. Additionally, or alternatively, the control gas-phase or control VOC mixture may be a control VOC mixture, a control non-VOC mixture, or a combination thereof. Thus, a control neuronal response may be obtained from a PFAS, lung cancer, or bacterial biofilm control, a non-PFAS, non-lung cancer, or non-bacterial biofilm control, or a combination thereof.Comparison results

[0115] In some embodiments, one or more test neuronal responses may be compared to one or more control neuronal responses. In some embodiments, at least 1. 2, 3, 4, 5, 6, 7, 8, 9, 10. 11. 12. 13.14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, or 30 neuronal responses may be recorded for each of the test and the control gas-phase PFAS or control VOC mixtures. In some embodiments, at least five neuronal responses may be recorded for each of the test and the control gas-phase PFAS or VOC mixtures. In some embodiments, the neuronal response recordings are averaged together prior to comparison. In some embodiments, more than one insect is used to generate the neuronal responses. In some embodiments, at least 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, or 30 insects may be used for each of the test and the control gas-phase PFAS or VOC mixtures. In certain embodiments. 5 locusts may be used to obtain neural recordings for each of the test and control gas-phase PFAS or VOC mixtures.

[0116] In some embodiments, dimensionality reduction analysis may be performed, for example via principal component analysis (PCA) and / or linear discriminant analysis (LDA), to visualize a population of neuronal responses. In some embodiments, one or more test neuronal responses may be compared to one or more non-PFAS, non-lung cancer, or non-bacterial biofilm control neuronal responses and / or one or more PFAS, lung cancer, or bacterial biofilm control neuronal responses. Thus, in some embodiments, using dimensionality reduction analysis, the test neuronal response may be classified as containing PFAS or not containing PFAS, containing lung cancer or not containing lung cancer, or containing bacterial biofilm or not containing bacterial biofilm. Thus, in some embodiments, using dimensionality reduction analysis, the test neuronal response may be classifiedTEC2025-0129 I TEC2025-0177 I TEC2026-00016550-000538-WO-POAas containing or not containing PFAS, lung cancer, or bacterial biofilm based on a probability of containing or not containing PFAS, lung cancer, or bacterial biofilm. Additionally, statistical tests, such as a leave-one-trial-out (LOTO) analysis, may be used to determine the accuracy of each prediction and / or the probability that a test gas-phase PFAS or VOC mixture either contains or does not contain PFAS, lung cancer, or bacterial biofilm.Clinical Workflow

[0117] In some embodiments, an example clinical workflow may include patients visiting a healthcare provider for symptoms related to lung cancer. At this initial point of contact, if the provider suspects a lung cancer diagnosis, a test using the method described herein can be ordered with sample collection occurring at the location of the provider. For sample collection, in some embodiments, exhaled breath or urine samples can be collected at any location (e.g., at-home, hospital, clinic, community healthcare setting, retail pharmacy, or other). In some embodiments, the urine or breath samples can be collected and sent to a central location for analysis. In some embodiments, within a few days, the patients may get their results back that will deliver the probability of lung cancer. In some embodiments, the method for detecting lung cancer may be done in a laboratory. For example, in some embodiments, the method may use specific infrastructure in order to obtain neural recordings from the insect.Optional treatment / prevention step

[0118] Optionally, in some embodiments, when the method of detecting lung cancer results in a presence of lung cancer or a probability of a presence of lung cancer in a sample obtained by a subject, the method may further include a treatment / prevention step. For example, where the test VOC mixture is obtained from a subject, the method may also include providing clinical care and / or treatment to the subject if the result indicates a presence or probability of a presence of lung cancer. Treating and / or preventing lung cancer is further described in Sec. E.D. Method of Diagnosing Lung Cancer

[0119] Also provided herein is a method for diagnosing lung cancer.

[0120] The method for diagnosing lung cancer includes exposing a biological chemosensory array to a test volatile organic compound (VOC) mixture, obtaining a test neuronal response from the biological chemosensory array, comparing the test neuronal response to one or more control neuronal responses obtained from one or more lung cancer control VOC mixture and one or more non-lungTEC2025-0129 I TEC2025-0177 I TEC2026-00016550-000538-WO-POAcancer control VOC mixture, and outputting a result based on the comparison. The result can indicate a presence or absence of lung cancer or a probability of a presence or absence of lung cancer.

[0121] The method for diagnosing lung cancer includes exposing a biological chemosensory array to a test volatile organic compound (VOC) mixture, obtaining a test neuronal response from the biological chemosensory array, comparing the test neuronal response to one or more control neuronal responses obtained from one or more lung cancer control VOC mixture and / or one or more non- lung cancer control VOC mixture, and outputting a result based on the comparison. The result can indicate a presence or absence of lung cancer or a probability of a presence or absence of lung cancer.

[0122] Thus, the present disclosure provides the method described in Sec. C, where the VOC mixture is obtained from a subject, further comprising diagnosing the subject with lung cancer if the result indicates a presence or probability of a presence of lung cancer.

[0123] Optionally, in some embodiments, the method may further include a treatment / prevention step. For example, where the test VOC mixture is obtained from a subject, the method may also include providing clinical care and / or treatment to the subject if the subject is diagnosed with lung cancer. Treating and / or preventing is further described in Sec. E.E. Method of Treating and / or Preventing Lung Cancer

[0124] Additionally, the present disclosure provides the method described in Sec. C, where the VOC mixture is obtained from a subject, further comprising providing clinical care and / or treatment to the subject if the result indicates a presence or probability of a presence of lung cancer.

[0125] In some embodiments, “treat” and “treatment” refer to an approach for obtaining beneficial or desired results, including clinical results. In some embodiments, the subject in need of treatment may include a subject diagnosed as having, or suspected to have, lung cancer. In some embodiments, prevention includes treatment of lung cancer that causes the clinical symptoms of the lung not to develop or progress, or treatment of lung cancer that reduces the occurrence of lung cancer. In some embodiments, “treatment” may also include treatment of environmental samples that contain PFAS.

[0126] In some embodiments, the clinical care and / or treatment may include performing one or more surgical procedures on a subject, such as laparoscopy to remove lung cancer, etc. In some embodiments, the clinical care and / or treatment may include one or more dietary and / or nutritional changes for the subject. In some embodiments, the clinical care and / or treatment may includeTEC2025-0129 I TEC2025-0177 I TEC2026-00016550-000538-WO-POAadministering to a subject one or more hormone therapies. In some embodiments, the clinical care and / or treatment may include administering to a subject an effective amount of a lung cancer medication. In some embodiments, the clinical care and / or treatment may include chemotherapy and / or radiation therapy.

[0127] In some embodiments, administration may be in an “effective amount” or a “therapeutically effective amount” (as the case may be), this being sufficient to show benefit to the individual / subject. The actual amount administered, and rate and time-course of administration, may depend on the nature and severity of lung cancer, or the type of lung cancer.

[0128] A pharmaceutical composition including a lung cancer medication may be administered as an individual therapeutic agent or in combination with other therapeutic agents, and may be administered sequentially or concurrently with an existing therapeutic agent and once or multiple times. In some embodiments, the pharmaceutical composition may be a chemotherapy treatment.

[0129] Administration of the pharmaceutical composition can be systemic, mucosal, and / or proximal to the location of a target site (e.g., near lung cancer tissue). Suitable routes of administration will be apparent to those of skill in the art. Various acceptable methods of administration include, but are not limited to, intravenous administration, intraperitoneal administration, intramuscular administration, intranodal administration, intracoronary administration, intraarterial administration (e.g., into a carotid artery), subcutaneous administration, retroorbital administration, transdermal delivery, intratracheal administration, subcutaneous administration, intraarticular administration, intraventricular administration, inhalation (e.g., aerosol), intracranial, intraspinal, intraocular, aural, intranasal, oral, pulmonary administration, impregnation of a catheter, and direct injection into a tissue. In one aspect, routes of administration include: intravenous, intraperitoneal, subcutaneous, intradermal, intranodal, intramuscular, transdermal, inhaled, intranasal, oral, intraocular, intraarticular, intracranial, and intraspinal. Parenteral delivery can include intradermal, intramuscular, intraperitoneal, intrapleural, intrapulmonary, intravenous, subcutaneous, atrial catheter, and venal catheter routes. Aural delivery can include ear drops, intranasal delivery can include nose drops or intranasal injection, and intraocular delivery can include eye drops. Aerosol (inhalation) delivery can also be performed using methods standard in the art.

[0130] A suitable amount of the pharmaceutical composition to be administered can be determined by routine experiments with animal models. Such models include, without implying any limitation, rabbit, sheep, mouse, rat, dog, and non-human primate models. Example unit dose formsTEC2025-0129 I TEC2025-0177 I TEC2026-00016550-000538-WO-POAfor injection include sterile solutions of water, physiological saline, or mixtures thereof. The pH of such solutions should be adjusted to about 7.4. Suitable carriers for injection include hydrogels, devices for controlled or delayed release, polylactic acid, and collagen matrices. For injection, the pharmaceutical composition may be provided, for example, in a pre-filled syringe. Suitable pharmaceutically acceptable carriers for topical application include those which are suitable for use in lotions, creams, gels, and the like. If the composition is to be administered orally, tablets, capsules, and the like may be used as a unit dose form. The pharmaceutically acceptable carriers for the preparation of unit dose forms which can be used for oral administration are well known in the prior art. The choice thereof will depend on secondary considerations such as taste, costs, and storability, which are not critical for the purposes of the present disclosure, and can be made without difficulty by a person skilled in the art.F. System for Detecting PFAS, Lung Cancer, or Bacterial Biofilm

[0131] FIG. 24 illustrates a block diagram of an example system 202, which may be used for detecting the presence or absence of a probability of a presence or absence of PFAS, lung cancer, or bacterial biofilm as described in Sec. C. For example, the method described herein for detecting a presence or absence of PFAS, lung cancer, or bacterial biofilms. The system includes one or more biological chemosensory array, an odor stimulus delivery component for delivering 1) at least one test gas-phase PFAS or VOC mixture and 2) at least one control gas-phase PFAS or VOC mixture, at least one control gas-phase non-PFAS or non-VOC mixture, or a combination thereof to the one or more biological chemosensory array, where the one or more biological chemosensory array is stabilized by a stabilizing component; a neuron probe for detecting at least one test neuronal response and at least one control neuronal response from the one or more biological chemosensory array; and at least one processor which stores the at least one test neuronal response and the at least one control neuronal response in memory, wherein the at least one test neuronal response and the at least one control neuronal response are one or more neuronal voltage signals.

[0132] In some embodiments, the clinical care and / or treatment may include administering to a subject, which may be carried out via a system 202 for detecting PFAS, lung cancer, or bacterial biofilms. In some embodiments, the clinical care and / or treatment may include administering to a subject or environment. The system 202 includes an odor stimulus delivery component 206 for delivering a test VOC mixture 207 to one or more biological chemosensory array 204, a neuron probe 203 for detecting one or more test neuronal responses from the one or more biological chemosensoryTEC2025-0129 I TEC2025-0177 I TEC2026-00016550-000538-WO-POAarray 204, and at least one processor 216 which stores the one or more test neuronal responses in memory. The biological chemosensory array 204 can be stabilized by a stabilizing component 205, and the test neuronal response can be one or more neuronal voltage signals.

[0133] As mentioned in Sec. C, the VOC mixture may be a VOC gas mixture. In some embodiments, the gas-phase PFAS or VOC mixture may be emitted from one or more biological sample, such as from breath, urine, sweat, or blood. In some embodiments, the biological sample may be obtained from a mouse, primate, or human.

[0134] Also as mentioned in Sec. C, the biological chemosensory array 204 may be one or more insect antenna. In some embodiments, the neuronal response may be obtained from an insect antennal lobe. In some embodiments, the insect may belong to the order Orthoptera, Hymenoptera, or Diptera. In some embodiments, an insect belonging to the order Orthoptera may be a locust. In some embodiments, an insect belonging to the order Hymenoptera may be a honeybee or an ant. In some embodiments, an insect belonging to the order Diptera may be a fruit fly. In some embodiments, the insect may have at least about 100 olfactory receptors. In some embodiments, a neuronal response may be obtained from an insect in vivo.

[0135] In some embodiments, the system may be used to diagnose, treat, and / or prevent lung cancer. The method of diagnosing lung cancer is described in Sec. D, while the method of treating and / or preventing lung cancer is described in Sec. E.

[0136] The system 202 may be configured to control operation of one or more aspects of the system, such as a neuron probe 203. As shown in FIG. 24, the system 202 includes one or more processors 216 (which may be referred to as a central processor unit (CPU) that is in communication with memory 214 including optional read only memory (ROM) 218 and optional random access memory (RAM) 220, and optional secondary storage 222 (such as disk drives). The processor 216 may be implemented as one or more CPU chips. The system 202 further includes optional input / output (I / O) devices 224, and network connectivity devices (e.g., a communication interface) 212.

[0137] In various implementations, the memory 214 may be configured to store computerexecutable instructions, and one or more processors 216 may be configured to execute the instructions to control operation of the neuron probe 203. The neuron probe 203 may include a sensor associatedTEC2025-0129 I TEC2025-0177 I TEC2026-00016550-000538-WO-POAwith the biological chemosensory array 204 and configured to sense at least one parameter of the biological chemosensory array 204.

[0138] The instructions may include receiving the at least one parameter of the biological chemosensory array 204 as detected by the sensor and implementing one or more control operations according to the at least one parameter. For example, the one or more processors 216 may be configured to determine one or more neuronal response (voltage signal) from a biological chemosensory array 204, to determine a presence or absence of PFAS, lung cancer, or bacterial biofilm or a probability of a presence or absence of PFAS, lung cancer, or bacterial biofilm.

[0139] The neuron probe 203 may include any suitable sensors, such as a commercial Neuronexus 16-channel silicon probe (A2x2-tet-3mm-150-150-121) with impedances between 200 and 300 kQ. This sensor may be used to detect or determine one or more voltage signals from the biological chemosensory array 204.

[0140] In various implementations, the memory 214 may be configured to store an artificial intelligence algorithm. The one or more processors 216 may be configured to use the artificial intelligence algorithm to calculate at least one value based on a sensed or determined parameter, and control operation of the neuron probe 203, etc. The artificial intelligence algorithm may include any suitable algorithm, such as a machine learning model. For example, a machine learning model may be trained to predict a presence or absence of PFAS, lung cancer, or bacterial biofilm or a probability of a presence or absence of PFAS, lung cancer, or bacterial biofilm according to historical values. The historical values may indicate levels of voltage signals obtained from the biological chemosensory array 204 when it is exposed to test gas-phase PFAS or VOC mixtures and / or control gas-phase PFAS or control VOC mixtures (e.g.. PFAS and non-PFAS controls).

[0141] Any suitable machine learning models may be used, and the models may be trained in any suitable fashion. For example, historical data may be separated into training data and test data, where the training data is used to train the model, and the test data is used to test the model performance and prediction accuracy. Typically, the set of training data is selected to be larger than the set of test data, depending on the desired model development parameters. Separating a portion of the acquired data as test data allows for testing of the trained model against actual historical output data, to facilitate more accurate training and development of the model. This arrangement may allow the system to simulate an outcome of the machine learning prediction when it processes a new voltage signal, etc., in the future. The model may be trained using any suitable machine learning model techniques.TEC2025-0129 I TEC2025-0177 I TEC2026-00016550-000538-WO-POAincluding those described herein, such as random forest logistic regression, decision tree (for example, a light gradient boosted tree), and neural networks.

[0142] The trained model may be tested using the test data, and the results of the output data from the tested model may be compared to actual historical outputs of the test data, to determine a level of accuracy. The model results may be evaluated using any suitable machine learning model analysis, such as cumulative gain and lift charts. Lift is a measure of the effectiveness of a predictive model calculated as the ratio between the results obtained with and without the predictive model (for example, by comparing the tested model outputs to the actual outputs of the test data). Cumulative gains lift charts provide visual aids for measuring model performance. Both charts include a lift curve and a baseline, where a greater area between the lift curve and the baseline indicates a stronger model.

[0143] After evaluating the model test results, the model may be deployed if the model test results are satisfactory. Deploying the model may include using the model to make predictions for a large-scale input dataset with unknown outputs, such as determining if a test gas-phase PFAS or VOC mixture is obtained from a biological sample from a subject with or without PFAS, lung cancer, or bacterial biofilm. If the evaluation of the model test results is unsatisfactory, the model may be developed further using different parameters, using different modeling techniques, or using other model types.

[0144] One example machine learning model is a recurrent neural-network-based model, which may be used to directly predict dependent variables without casting relationships between variables into mathematical formulas. The neural network model includes a large number of virtual neurons operating in parallel and arranged in layers. The first layer is the input layer and receives raw input data. Each successive layer modifies outputs from a preceding layer and sends them to a next layer. The last layer is the output layer and produces output of the system.

[0145] In some embodiments, a convolutional neural network may be implemented. Similar to LSTM neural networks, convolutional neural networks include an input layer, a hidden layer, and an output layer. However, in a convolutional neural network, the output layer includes one fewer output than the number of neurons in the hidden layer and each neuron is connected to each output. Additionally, each input in the input layer is connected to each neuron in the hidden layer.

[0146] In various implementations, each input node in the input layer may be associated with a numerical value, which can be any real number. In each layer, each connection that departs from anTEC2025-0129 I TEC2025-0177 I TEC2026-00016550-000538-WO-POAinput node has a weight associated with it, which can also be any real number. In the input layer, the number of neurons equals the number of features (columns) in a dataset. The output layer may have multiple continuous outputs.

[0147] As mentioned above, the layers between the input and output layers are hidden layers. The number of hidden layers can be one or more (one hidden layer may be sufficient for many applications). A neural network with no hidden layers can represent linear separable functions or decisions. A neural network with one hidden layer can perform continuous mapping from one finite space to another. A neural network with two hidden layers can approximate any smooth mapping to any accuracy.

[0148] The number of neurons can be optimized. At the beginning of training, a network configuration is more likely to have excess nodes. Some of the nodes may be removed from the network during training that would not noticeably affect network performance. For example, nodes with weights approaching zero after training can be removed (this process is called pruning). The number of neurons can cause under- fitting (inability to adequately capture signals in dataset) or overfitting (insufficient information to train all neurons; network performs well on training dataset but not on test dataset).

[0149] Various methods and criteria can be used to measure performance of a neural network model. For example, root mean squared error (RMSE) measures the average distance between observed values and model predictions. Coefficient of Determination (R2) measures correlation (not accuracy) between observed and predicted outcomes. This method may not be reliable if the data has a large variance. Other performance measures include irreducible noise, model bias, and model variance. A high model bias for a model indicates that the model is not able to capture true relationship between predictors and the outcome. Model variance may indicate whether a model is not stable (a slight perturbation in the data will significantly change the model fit).

[0150] Referring again to FIG.24, the secondary storage 222 may include one or more disk drives or tape drives. The secondary storage 222 may be used for non-volatile storage of data and as an overflow data storage device if RAM 220 is not large enough to hold all working data. The secondary storage 222 may be used to store programs which are loaded into RAM 220 when such programs are selected for execution.TEC2025-0129 I TEC2025-0177 I TEC2026-00016550-000538-WO-POA

[0151] In this embodiment, the secondary storage 222 has a processing component 222a comprising non-transitory instructions operative by the processor 216 to perform various operations of the methods of the present disclosure. The ROM 218 is used to store instructions and perhaps data which are read during program execution. The secondary storage 222, the memory 214, the RAM 220, and / or the ROM 218 may be referred to in some contexts as computer readable storage media and / or non-transitory computer readable media.

[0152] The optional VO devices 224 may include printers, video monitors, liquid crystal displays (LCDs), plasma displays, touch screen displays, keyboards, keypads, switches, dials, mice, track balls, voice recognizers, card readers, paper tape readers, or other suitable input devices.

[0153] The network connectivity devices 212 may take the form of modems, modem banks, Ethernet cards, universal serial bus (USB) interface cards, serial interfaces, token ring cards, fiber distributed data interface (FDDI) cards, wireless local area network (WLAN) cards, radio transceiver cards. The devices 212 may promote radio communications using protocols, such as code division multiple access (CDMA), global system for mobile communications (GSM), long-term evolution (LTE), worldwide interoperability for microwave access (WiMAX), near field communications (NFC), radio frequency identity (RFID), and / or other air interface protocol radio transceiver cards, and other suitable network devices. These network connectivity devices 212 may enable the processor 216 to communicate with the Internet and / or one or more intranets. With such a network connection, it is contemplated that the processor 216 might receive information from the network, might output information to the network in the course of performing the above-described method operations, etc. Such information, which is often represented as a sequence of instructions to be executed using processor 216, may be received from and outputted to the network, for example, in the form of a computer data signal embodied in a carrier wave.

[0154] The processor 216 executes instructions, codes, computer programs, scripts, which it accesses from hard disk, floppy disk, optical disk (these various disk based systems may all be considered secondary storage 222), flash drive, memory 214, ROM 218, RAM 220, the network connectivity devices 212, etc. While only one processor 216 is shown, multiple processors may be present. Thus, while instructions may be discussed as executed by a processor, the instructions may be executed simultaneously, serially, or otherwise executed by one or multiple processors.

[0155] Although the system 202 is described with reference to a computing device, it should be appreciated that the system may be formed by two or more computing devices in communication withTEC2025-0129 I TEC2025-0177 I TEC2026-00016550-000538-WO-POAeach other that collaborate to perform a task. For example, but not by way of limitation, an application may be partitioned in such a way as to permit concurrent and / or parallel processing of the instructions of the application. Alternatively, the data processed by the application may be partitioned in such a way as to permit concurrent and / or parallel processing of different portions of a dataset by the two or more computers.

[0156] In an embodiment, virtualization software may be employed by the system 202 to provide the functionality of a number of servers that is not directly bound to the number of computers in the system 202. The functionality disclosed above may be provided by executing an application and / or applications in a cloud computing environment. Cloud computing may include providing computing services via a network connection using dynamically scalable computing resources. A cloud computing environment may be established by an enterprise and / or may be hired on an as-needed basis from a third party provider.

[0157] It is understood that by programming and / or loading executable instructions onto the system 202, at least one of the CPU 216, the memory 214, the ROM 218, and the RAM 220 are changed, transforming the system 202 in part into a specific purpose machine and / or apparatus having the novel functionality taught by the present disclosure. It is fundamental to the electrical engineering and software engineering arts that functionality that can be implemented by loading executable software into a computer can be converted to a hardware implementation by well-known design rules.

[0158] The foregoing description is merely illustrative in nature and is in no way intended to limit the disclosure, its application, or uses. The broad teachings of the disclosure can be implemented in a variety of forms. Therefore, while this disclosure includes particular examples, the true scope of the disclosure should not be so limited since other modifications will become apparent upon a study of the drawings, the specification, and the following claims. It should be understood that one or more steps within a method may be executed in different order (or concurrently) without altering the principles of the present disclosure. Further, although each of the embodiments is described above as having certain features, any one or more of those features described with respect to any embodiment of the disclosure can be implemented in and / or combined with features of any of the other embodiments, even if that combination is not explicitly described. In other words, the described embodiments are not mutually exclusive, and permutations of one or more embodiments with one another remain within the scope of this disclosure.TEC2025-0129 I TEC2025-0177 I TEC2026-00016550-000538-WO-POA

[0159] Spatial and functional relationships between elements (for example, between modules) are described using various terms, including “connected,” “engaged,” “interfaced,” and “coupled.” Unless explicitly described as being “direct.” when a relationship between first and second elements is described in the above disclosure, that relationship encompasses a direct relationship where no other intervening elements are present between the first and second elements, and also an indirect relationship where one or more intervening elements are present (either spatially or functionally) between the first and second elements. The phrase “at least one of A, B, and C” should be construed to mean a logical (A OR B OR C), using a non-exclusive logical OR, and should not be consumed to mean “at least one of A, at least one of B, and at least one of C.”

[0160] In the figures, the direction of an arrow, as indicated by the arrowhead, generally demonstrates the flow of information (such as data or instructions) that is of interest to the illustration. For example, when element A and element B exchange a variety of information but information transmitted from element A to element B is relevant to the illustration, the arrow may point from element A to element B. This unidirectional arrow does not imply that no other information is transmitted from element B to element A. Further, for information sent from element A to element B. element B may send requests for, or receipt acknowledgments of, the information to element A. The term “subset” does not necessarily require a proper subset. In other words, a first subset of a first set may be coextensive with (equal to) the first set.

[0161] In this application, including the definitions below, the term “module” or the term “controller” may be replaced with the term “circuit.” The term “module” may refer to, be part of, or include processor hardware (shared, dedicated, or group) that executes code and memory hardware (shared, dedicated, or group) that stores code executed by the processor hardware.

[0162] The module may include one or more interface circuits. In some examples, the interface circuit(s) may implement wired or wireless interfaces that connect to a local area network (LAN) or a wireless personal area network (WPAN). Examples of a LAN are Institute of Electrical and Electronics Engineers (IEEE) Standard 802.11-2016 (also known as the WIFI wireless networking standard) and IEEE Standard 802.3-2015 (also known as the ETHERNET wired networking standard). Examples of a WPAN are IEEE Standard 802.15.4 (including the ZIGBEE standard from the ZigBee Alliance) and, from the Bluetooth Special Interest Group (SIG), the BLUETOOTH wireless networking standard (including Core Specification versions 3.0, 4.0, 4.1, 4.2, 5.0, and 5.1 from the Bluetooth SIG).TEC2025-0129 I TEC2025-0177 I TEC2026-00016550-000538-WO-POA

[0163] The module may communicate with other modules using the interface circuit(s). Although the module may be depicted in the present disclosure as logically communicating directly with other modules, in various implementations the module may actually communicate via a communications system. The communications system includes physical and / or virtual networking equipment such as hubs, switches, routers, and gateways. In some implementations, the communications system connects to or traverses a wide area network (WAN) such as the Internet. For example, the communications system may include multiple LANs connected to each other over the Internet or point-to-point leased lines using technologies including Multiprotocol Label Switching (MPLS) and virtual private networks (VPNs).

[0164] In various implementations, the functionality of the module may be distributed among multiple modules that are connected via the communications system. For example, multiple modules may implement the same functionality distributed by a load balancing system. In a further example, the functionality of the module may be split between a server (also known as remote, or cloud) module and a client (or user) module.

[0165] The term “code”, as used above, may include software, firmware, and / or microcode, and may refer to programs, routines, functions, classes, data structures, and / or objects. Shared processor hardware encompasses a single microprocessor that executes some or all code from multiple modules. Group processor hardware encompasses a microprocessor that, in combination with additional microprocessors, executes some or all code from one or more modules. References to multiple microprocessors encompass multiple microprocessors on discrete dies, multiple microprocessors on a single die, multiple cores of a single microprocessor, multiple threads of a single microprocessor, or a combination of the above.

[0166] Shared memory hardware encompasses a single memory device that stores some or all code from multiple modules. Group memory hardware encompasses a memory device that, in combination with other memory devices, stores some or all code from one or more modules.

[0167] The term “memory hardware” is a subset of the term “computer-readable medium”. The term computer-readable medium, as used herein, does not encompass transitory electrical or electromagnetic signals propagating through a medium (such as on a carrier wave); the term computer-readable medium is therefore considered tangible and non-transitory. Non-limiting examples of a non-transitory computer-readable medium are nonvolatile memory devices (such as a flash memory device, an erasable programmable read-only memory device, or a mask read-onlyTEC2025-0129 I TEC2025-0177 I TEC2026-00016550-000538-WO-POAmemory device), volatile memory devices (such as a static random access memory device or a dynamic random access memory device), magnetic storage media (such as an analog or digital magnetic tape or a hard disk drive), and optical storage media (such as a CD, a DVD, or a Blu-ray Disc).

[0168] The apparatuses and methods described in this application may be partially or fully implemented by a special purpose computer created by configuring a general purpose computer to execute one or more particular functions embodied in computer programs. The functional blocks and flowchart elements described above serve as software specifications, which can be translated into the computer programs by the routine work of a skilled technician or programmer.

[0169] The computer programs include processor-executable instructions that are stored on at least one non-transitory computer-readable medium. The computer programs may also include or rely on stored data. The computer programs may encompass a basic input / output system (BIOS) that interacts with hardware of the special purpose computer, device drivers that interact with particular devices of the special purpose computer, one or more operating systems, user applications, background services, background applications, etc.

[0170] The computer programs may include: (i) descriptive text to be parsed, such as HTML (hypertext markup language), XML (extensible markup language), or JSON (JavaScript Object Notation), (ii) assembly code, (iii) object code generated from source code by a compiler, (iv) source code for execution by an interpreter, (v) source code for compilation and execution by a just-in-time compiler, etc. As examples only, source code may be written using syntax from languages including C, C++, C#, Objective C, Swift, Haskell, Go, SQL, R, Lisp, Java®, Fortran, Perl, Pascal, Curl, OCaml, JavaScript®, HTML5 (Hypertext Markup Language 5th revision), Ada, ASP (Active Server Pages), PHP (PHP: Hypertext Preprocessor), Scala, Eiffel, Smalltalk, Erlang, Ruby, Flash®, Visual Basic®, Lua, MATLAB®, SIMULINK, and Python®.TEC2025-0129 I TEC2025-0177 I TEC2026-00016550-000538-WO-POAEXAMPLES

[0171] The following examples are merely illustrative and do not limit this disclosure in any way.EXAMPLE 1: General Methods for Detecting PEAS Using In Vivo Bioelectric Sensor

[0172] Locust husbandry

[0173] Locusts (Schistocerca americand) were raised in a crowded colony and kept in an incubator with a 12-hour day and night cycle. The incubator simulated natural temperature fluctuations, maintaining 36.5°C during the day and 25°C at night. The locusts were fed a daily mixture of fresh grass and wheat germ. Locust eggs were kept in a separate incubator, maintained at a temperature of 37.7 °C.

[0174] Locust surgery

[0175] Both sexes of post-fifth instar locusts were used for electrophysiological recordings (D. Saha et al., A spatiotemporal coding mechanism for background-invariant odor recognition. Nature neuroscience 16. 1830-1839 (2013); D. Saha, K. Leong, N. Katta, B. Raman, Multi-unit recording methods to characterize neural activity in the locust (Schistocerca americana) olfactory circuits. Journal of visualized experiments; JoVE, 50139 (2013)). The limbs and wings were removed, and the openings were sealed with VETBOND®. The locust was then immobilized on a custom-made surgical platform, secured with electrical tape. Batik wax was used to immobilize the antennae by first constructing a wax pillar on either side of the head, and then securing the antennae into the pillars using tubes. A bowl of batik wax was formed around the head to hold a room-temperature saline solution, ensuring the brain remained moist during excision. A two-part epoxy was used to stabilize the antennae. The head’s exoskeleton was carefully cut to remove the glandular tissue, exposing the brain and gut. The gut was then removed, and a platform was inserted to stabilize and lift the brain. Finally, the antennal lobes were desheathed using a protease treatment.

[0176] Odor vial preparation

[0177] For the pure PFAS chemicals, the following seven odorants were used: 6:2 FTOH, 8:2 FTOH, 6:2 FTAcr, EtFOSE, MeFOSE, MeFOSA, and PFOSA. These odorants were distributed into individual vials at a concentration of 100 pg (or the equivalent volume in pL, depending on their state of matter), ensuring an equal volume of each odor. An empty vial was used as a control.TEC2025-0129 I TEC2025-0177 I TEC2026-00016550-000538-WO-POA

[0178] For varying concentrations of the three selected PFAS chemicals - 6:2 FTOH, 6:2 FTAcr, and PFOS - each odorant was diluted in 10 mF of ddPEO at a concentration of 0.01% vol / vol. Serial dilutions were then performed in 10 mL of ddPEO to achieve a concentration of 0.0001% vol / vol. A vial containing only 10 mF of ddPEO was used as a control. Raoult’s Faw was used to calculate the ppb and ppt of concentrations, using the density of ddlFO. Chemical characteristics (i.e., vapor pressure, etc.) needed for Raoult’s Faw were taken from averaging values found in literature (C. M. Eichler, J. C. Eittle, A framework to model exposure to per-and polyfluoroalkyl substances in indoor environments. Environmental Science: Processes & Impacts 22, 500-511 (2020); H. P. H. Arp, C. Niederer, K..-U. Goss, Predicting the partitioning behavior of various highly fluorinated compounds. Environmental science & technology 40, 7298-7304 (2006); B. Bhhatarai, P. Gramatica, Prediction of aqueous solubility, vapor pressure and critical micelle concentration for aquatic partitioning of perfluorinated chemicals. Environmental science & technology 45, 8120-8128 (2011); Z. Wang, M. MacLeod, I. T. Cousins, M. Scheringer, K. Hungerbiihler, Using COSMOtherm to predict physicochemical properties of poly-and perfluorinated alkyl substances (PFASs). Environmental Chemistry 8, 389-398 (2011); M. Kim, L. Y. Li, J. R. Grace, C. Yue, Selecting reliable physicochemical properties of perfluoroalkyl and polyfluoroalkyl substances (PFASs) based on molecular descriptors. Environmental pollution 196, 462-472 (2015); USEPA, CompTox Chemical Dashboard. U. S. Environmental Protection Agency, (2023); Y. D. Lei, F. Wania, D. Mathers, S. A. Mabury, Determination of vapor pressures, octanol- air, and water- air partition coefficients for polyfluorinated sulfonamide, sulfonamidoethanols, and telomer alcohols. Journal of Chemical & Engineering Data 49, 1013-1022 (2004); N. L. Stock, D. A. Ellis, L. Deleebeeck, D. C. Muir, S. A. Mabury, Vapor pressures of the fluorinated telomer alcohols limitations of estimation methods. Environmental science & technology 38, 1693-1699 (2004); P. J. Krusic et al., Vapor pressure and intramolecular hydrogen bonding in fluorotelomer alcohols. The Journal of Physical Chemistry A 109, 6232-6241 (2005): M. Zhang (Brown University, 2021)).

[0179] Electrophysiology

[0180] In-vivo extracellular electrophysiology recordings were performed on either sex of postfifth instar locusts. A commercial 16-channel NeuroNexus® silicon electrode (A2x2-tet-3mm-150-150-121) was used to extract neural activity from the antennal lobes during odor stimulus presentation. The electrodes were electroplated prior to recordings to achieve impedances between 200 and 300 kQ. A silver-chloride reference wire was placed in the saline solution within the waxTEC2025-0129 I TEC2025-0177 I TEC2026-00016550-000538-WO-POAbowl surrounding the locust’s head to ground the experiments. The electrode was inserted approximately 100 pm into the antennal lobes to collect voltage signals. Extracellular recordings were sampled at 20 kHz from eight channels, comprising two tetrodes. Neural activity was digitized using an Intan® pre-amplifier board (C3334 RHD 16-channel head stage), which transmitted data to an Intan® recording controller (C3100 RHD USB interface board) (FIG. 1A and FIG. IB). The neural activity throughout odor presentation was visualized and stored via the Intan® graphical user interface and a Lab VIEW® data acquisition system, which synchronized both odor delivery and recording windows (FIG. IB). For pure PFAS electrophysiology recordings, the same stimulus panel was used, consisting of seven pure PFAS odors and a control vial. For concentration experiments, the stimulus panel consisted of three PFAS odors, each diluted to two separate concentrations (0.01% and 0.0001% vol / vol), for a total of six odors, plus a control vial. Each odor panel was pseudorandomized before recordings, with five trials per odor and a 1 -minute interstimulus interval. A total of 13 locusts were used for pure PFAS experiments, resulting in 29 antennal lobe recordings. From those recordings, 86 neurons were spike-sorted and used for classification analysis. For the concentration experiments, eight locusts were used, yielding 18 antennal lobe recordings. From these recordings, 69 neurons were spike-sorted. Root- mean- square (RMS) voltage traces were used for classification analysis.

[0181] Odor stimulation

[0182] Odor stimulus presentation followed pre-established methodology (M. Parnas et al., Precision detection of select human lung cancer biomarkers and cell lines using honeybee olfactory neural circuitry as a novel gas sensor. Biosensors and Bioelectronics, 116466 (2024); D. Saha et al., A spatiotemporal coding mechanism for background-invariant odor recognition. Nature neuroscience 16, 1830-1839 (2013); D. Saha, K. Leong, N. Katta, B. Raman, Multi-unit recording methods to characterize neural activity in the locust (Schistocerca americana) olfactory circuits. Journal of visualized experiments: JoVE, 50139 (2013): A. Famum et al., Harnessing insect olfactory neural circuits for detecting and discriminating human cancers. Biosensors and Bioelectronics 219, 114814 (2023)). Briefly, a commercial olfactometer (Aurora Scientific, 220A) controlled the odor presentation and airflow (FIG. 1A and FIG. IB). A fresh air line delivered 200 seem of contaminant-free air through a 1 / 16” diameter polytetrafluoroethylene (PTFE) stimulus flow line to the locust antennae, while 200 seem of air was directed through the dilution flow line to exhaust. The end of the stimulus flow line was positioned 2 - 3 cm from the antennae’s distal segment. Approximately five seconds before odor stimulus, 40% (80 seem) of the dilution flow was redirected through the odorTEC2025-0129 I TEC2025-0177 I TEC2026-00016550-000538-WO-POAflow line, upstream of the odorant vials. Both lines merged downstream of the vials, mixing with 120 seem of clean air from the dilution flow, creating a total of 200 seem. This allowed a volatile mixture from the stimulus odor to prime the lines up to the final valve, which was previously directed to the exhaust. Once activated, the final valve switched the fresh air flow to exhaust and directed the volatile mixture to the stimulus flow line, delivering the odor to the locust antennae. The odor stimulus lasted four seconds before the final valve diverted fresh air to the antennae and exhausted the volatile mixture. One second after stimulus removal, the 80 seem flowing through the odor flow line was rerouted to prevent potential depletion of the odorant in the vial. A 6’ diameter funnel, generating a slight vacuum, was placed directly behind the locust antennae to remove lingering odorants. The odor stimulation protocol was designed to maintain consistent air flow, minimizing potential responses due to changes in air pressure.

[0183] Spike sorting

[0184] A high-pass Butterworth filter, set to 300 Hz, was applied to remove low-frequency components. The raw neural data was imported and analyzed using custom-written codes in MATLAB® R2024a scripts. The data was then converted into a format readable by IGOR PRO® 4 for spike-sorting analysis, following previously described methods to isolate active neurons during stimulus windows (M. Parnas et al., Precision detection of select human lung cancer biomarkers and cell lines using honeybee olfactory neural circuitry as a novel gas sensor. Biosensors and Bioelectronics, 116466 (2024); D. Saha et al., A spatiotemporal coding mechanism for backgroundinvariant odor recognition. Nature neuroscience 16, 1830-1839 (2013); D. Saha, K. Leong, N. Katta, B. Raman, Multi-unit recording methods to characterize neural activity in the locust (Schistocerca americana) olfactory circuits. Journal of visualized experiments: JoVE, 50139 (2013); A. Farnum et al., Harnessing insect olfactory neural circuits for detecting and discriminating human cancers. Biosensors and Bioelectronics 219, 114814 (2023)). Spike-sorting was performed with detection thresholds set between 2.5 to 3.5 standard deviations (SDs) from baseline fluctuations. Neurons were considered acceptable if they met the following criteria: < 10% inter-spike intervals (ISIs), spike amplitudes > 5 SDs apart from other sorted neurons, and spike waveform variance < 10%. Additionally, neurons that passed these initial criteria had to exhibit consistent baseline firing for each odor stimulus across five trials, as visualized on raster plots. A total of 86 neurons were isolated for pure PFAS compound recordings from 13 locusts, and 69 neurons were isolated for PFAS concentration recordings using spike sorting.TEC2025-0129 I TEC2025-0177 I TEC2026-00016550-000538-WO-POA

[0185] RMS transformation

[0186] The raw data was imported into MATLAB® and processed using a 300 Hz filter. Artifacts caused by electrical interference were identified and removed, resulting in 180 samples centered around each artifact, which were then normalized to the mean voltage value. This process effectively eliminated any electrical interference artifacts from the raw data. For pure PFAS compounds, of the total recorded root-mean-square (RMS) responses (2,040 samples: 51 tetrodes x 8 odors x 5 trials), only 29 voltage traces contained artifacts that required removal. No artifacts were found in the PFAS concentration data; therefore, no voltage samples were discarded. The filtered data was then trimmed to the time window of interest before being processed through an RMS filter using previously described method (M. Parnas et al., Precision detection of select human lung cancer biomarkers and cell lines using honeybee olfactory neural circuitry as a novel gas sensor. Biosensors and Bioelectronics. 116466 (2024); D. Saha et al.. A spatiotemporal coding mechanism for backgroundinvariant odor recognition. Nature neuroscience 16, 1830-1839 (2013); D. Saha, K. Leong, N. Katta, B. Raman, Multi-unit recording methods to characterize neural activity in the locust (Schistocerca americana) olfactory circuits. Journal of visualized experiments: JoVE, 50139 (2013); A. Famum et al., Harnessing insect olfactory neural circuits for detecting and discriminating human cancers. Biosensors and Bioelectronics 219, 114814 (2023)). Briefly, a 500-point continuously moving RMS filter, accompanied by a smoothing, 500-point continuously moving averaging filter were applied to the voltage data. Baseline values were calculated by averaging all time bins and trials from the two seconds prior to odor presentation. The data were then normalized by subtracting these baseline values to secure ARMS values. These values were binned into 50 ms non-overlapping time bins, and the average of each bin was calculated. For pure PFAS experiments, RMS transformed voltage data from four channels on each tetrode were averaged over 29 positions (51 tetrodes) from 13 locusts (FIG.2A-D). A similar procedure was applied to the recorded 18 positions (34 tetrodes) from eight locusts in the PFAS concentration experiments. RMS data were used for analysis in FIG. 6A-F and FIG.7A-F.

[0187] Dimensionality reduction analysis

[0188] Two dimensionality reduction techniques - principal component analysis (PC A) and linear discriminant analysis (LDA) - were used to qualitatively visualize odor-evoked responses in neuronal activity. In PCA, each neuron’s baseline response was calculated by averaging the firing rate during the two seconds before odor presentation. The binned baseline was then subtracted fromTEC2025-0129 I TEC2025-0177 I TEC2026-00016550-000538-WO-POAthe spike-sorted neuronal activity. Neuron signals were binned in 50 ms non-overlapping time intervals and averaged across five trials, each with a one-minute interstimulus interval. All spike-sorted neurons were pooled across multiple experiments. For example, in FIG. 5A, the binned responses of all recorded neurons (n = 86) and the number of 50 ms bins between 250 ms and 1000 ms were combined into a matrix with neuron number (86) x time bins (t = 15). This matrix represented the spike count of each neuron in each 50 ms time bin. This approach was applied to the entire neuronal time-series population, capturing changes in neural trajectory evolution over time for each odor stimulus. PCA was performed on the time-series data from all seven pure PFAS compounds and the control, maximizing the variance within the data set. High-dimensional time bin vectors were projected onto the principal component axes to visualize the directions of maximum variance. The three-dimensional vectors with the highest eigenvalues were used to plot and connect the data points in adjacent time bins, illustrating neural trajectory evolution over time (FIG. 5A). The trajectories converge at the origin to explore the dynamics of odor-specific responses and trajectory separation. PCA analysis using baseline-subtracted. RMS transformed population data was also applied to the pure PFAS data (FIG.2A). A similar PCA analysis was applied to the PFAS concentration data using baseline- subtracted, RMS transformed population data (FIG. 6D, FIG. 7D, FIG. 3C). For LDA, analogous neural population data matrices were used for the evaluation of pure PFAS compounds (FIG. 5B), and RMS transformed population data was further used for both pure PFAS compounds and PFAS concentration data (FIG. 6C, FIG. 7C, FIG. 2B, and FIG. 3B). LDA minimizes within-stimulus variation while maximizing the separation between stimuli, highlighting the differentiation in spatio-temporal space for each odor. The time bins were plotted as points in the LDA space, revealing distinct neural clustering for stimulus response. All dimensionality reduction analyses were performed using custom-written MATLAB® R2024a scripts.

[0189] Leave-One-Trial-Out (LOTO) confusion matrix

[0190] To quantitatively classify individual PFAS concentrations, a leave-one-trial-out (LOTO) confusion matrix was employed. This method evaluates the accuracy of PFAS classification within our model, demonstrating how effectively locust neural responses differentiate between PFAS compounds, various concentrations of the same chemical, and an odor control. After spike sorting, neuronal population activity was divided into 50 ms time bins, creating a three-dimensional matrix (e.g., 86 neurons x 5 trials x 110-time bins). The number of bins depends on the time window of interest; for this data, the time window was 0.25 - 5.75 s (5,500 ms) after odor presentation. This timeTEC2025-0129 I TEC2025-0177 I TEC2026-00016550-000538-WO-POAwindow encompasses both the ON and OFF responses, capturing the entire neuronal odor-evoked response. The spiking events in each time bin are counted for all neurons in the population. This process is repeated for each neuron across five trials. Four of the five trials were averaged to create a training template (predicted label), while the fifth trial (the left-out trial) served as a testing template (true label). The number of predicted and true labels corresponds to the number of odors (e.g., for eight odors, there will be eight predicted and eight true labels). These templates were plotted in highdimensional space, and each testing template was assigned to a training template based on the smallest Euclidean distance. A confusion matrix was then generated to depict classification accuracy of the assignments. Using a “winner-takes-all” approach, or the mode of time bin assignments, the testing trials were matched to their respective training templates. This resulted in a trial-wise LOTO confusion matrix that indicated the accuracy of odor classification. A fully diagonal matrix indicates 100% classification accuracy. A similar analysis was performed for RMS transformed pure and concentration PFAS data (FIG. 6E, FIG. 7E, FIG. 2C, and FIG. 3D). In this case, the three-dimensional matrix was based on the number of tetrodes instead of neurons (e.g., 34 tetrodes x five trials x 110-time bins). All quantitative classification analyses were carried out using custom written MATLAB® R2024a scripts.

[0191] Sensitivity and Specificity Tables

[0192] Values used to calculate sensitivity and specificity were obtained from the LOTO confusion matrices. The columns of the matrix were divided between true positives (TP) and false negatives (FN), while the rows were split between false positives (FP) and TP. Any data not in the odor column or row of interest was considered a true negative (TN). A TP was identified as the diagonal classification accuracy, where the true label was correctly assigned to the predicted label.

[0193] Sensitivity was calculated as the ratio of TP to the sum of TP and FN, representing all positive conditions.TPSensitivity = - TP + FN

[0194] Specificity was calculated as the ratio of TN to the sum of TN and FP, highlighting all negative conditions.TNSpecificity =TN + FPTEC2025-0129 I TEC2025-0177 I TEC2026-00016550-000538-WO-POA

[0195] For calculations of groups (i.e., PFAS vs Air, PFAS families, PFAS vs ddFEO), all odors of interest values were combined. For the FTOH family, TP values were defined as any classification that was accurately assigned to either 6:2 FTOH or 8:2 FTOH, including the misclassification of one as the other (FIG. 5C and FIG. 5D). False positives (FP) were any odors outside the FTOH family assigned to either FTOH compound. True negatives (TN) were considered as any odor assignments not in the FTOH family. A similar process was applied to other PFAS families analyzed (FIG. 5D and FIG. 2D). For the PFAS vs Controls groupings, any assignment of a PFAS odor to a PFAS odor was considered a TP. Assigning a PFAS odor to a control was a FN, and assigning a control to a PFAS odor was a FP. TN were defined as an accurate assignment of the control odor to itself. Analogous sensitivity and specificity tables were created for PFAS concentrations (Figs. 6F, FIG.7F, and FIG.3E).EXAMPLE 2: PFAS Are Detected By Neurons In The Locust Al

[0196] Odor-evoked neural responses were measured from the locust olfactory circuitry to seven pure PFAS and an odor control. The primary objective of these experiments was to investigate the broad detection range of PFAS. In brief, a multichannel electrode was inserted into one of the antennal lobes; odor delivery and airflow were controlled using a commercial-grade olfactometer (FIG. 4A and FIG. 1A). The stimuli were presented in a pseudorandomized order across five trials to avoid temporal biases. In-vivo electrophysiological data was recorded and voltage traces were spike- sorted to identify individual neurons that responded to the odor stimuli.

[0197] It was found that odor-evoked neural activity from a single extracted neuron varies in response to all seven PFAS and the control (FIG. 4B). Raster plots depicted individual action potentials across five trials. Averaging the trials allowed visualization of the firing frequency of one neuron in peri-stimulus time histograms (PSTHs), revealing changes in firing frequency throughout odor stimulation (ON-response). as shown by the boxes (FIG. 4B). Two representative neurons demonstrated distinct differences in responses to multiple PFAS compounds and the control (FIG.4B). Each neuron in the population exhibited a unique response to the odors, prompting examination of the neuronal population’s response. PSTHs from 86 neurons (n = 86) were combined to investigate the population response (FIG. 4C). Analysis displayed the collective neural response to each stimulus, with distinct ON- and OFF-responses (post- stimulus) observed for certain PFAS (FIG.4C).These findings suggest that the neuronal population responds to PFAS compounds through aTEC2025-0129 I TEC2025-0177 I TEC2026-00016550-000538-WO-POAcombination of ON and OFF responses, creating a unique neural fingerprint for each PFAS. The most variance in the data was observed between 0.25 - 1.0 seconds post-stimulus onset, likely due to a delay in odor response following the final valve opening (FIG. 4C).EXAMPLE 3: Neural Responses Can Differentiate Between PFAS

[0198] High-dimensional reduction techniques were applied to examine the neuronal population's response between various PFAS. Neurons from electrophysiology recordings (n = 86) were combined to generate an odor-evoked population response matrix (neurons x time). Responses were aligned at stimulus onset and averaged across five trials. Firing rates for each neuron were binned into nonoverlapping, 50 ms time intervals, chosen based on 20 Hz oscillations observed in the insect antennal lobes (M. Stopfer, G. Laurent, Short-term memory in olfactory network dynamics. Nature 402, 664-668 (1999); J. Perez-Orive et al., Oscillations and sparsening of odor representations in the mushroom body. Science 297, 359-365 (2002)). The number of spikes within each time bin was counted to create odor-specific, high-dimensional population response vectors. Principal component analysis (PCA) was used to visualize the temporal odor-evoked responses of the neuronal population. PCA was performed using a 750 ms time window from 0.25 to 1.0 seconds post-stimulus onset to capture the period of highest odor-evoked response (FIG. 4C). The three principal components with the highest eigenvalues, which explained the most variance in the dataset, were projected onto three dimensions for visualization (FIG.5A). Odor-specific response vectors were connected across adjacent time bins, forming stimulus neural trajectories. Trajectories were aligned at the origin, represented by a black dot. Notably, each odor stimulus produced distinct neural trajectories, indicating unique neuronal population responses to all PFAS, as well as the control (FIG. 5A). Further analysis using root mean square (RMS) for real-time data analysis revealed similar results (n = 51, FIG. 2A). Additionally, a second dimensionality reduction technique - supervised linear discriminant analysis (LDA) - was applied to maximize variance between stimulus clusters while minimizing variance within each stimulus. Clear neuronal clustering was observed for each PFAS in neuronal populations and RMS analyses (FIG.5B and FIG.2B). These findings suggest that the locust neural circuitry exhibits both temporal and spatial discrimination across a range of PFAS.

[0199] Quantitative classification was performed using a leave-one-trial-out (LOTO) matrix. In brief, the neuron response was aligned and divided into 50 ms non-overlapping time bins. The number of spikes within each time bin was counted to create a 3D matrix (trials x neurons x time bins). FourTEC2025-0129 I TEC2025-0177 I TEC2026-00016550-000538-WO-POAout of five trials were combined to form a training template, while the remaining trial formed a testing template. The templates were plotted in high-dimensional space; each testing template was assigned to the closest training template based on Euclidean distance. This procedure was repeated for each trial serving as the testing template once. The classification results were determined by the mode of time bin assignments, with higher classification accuracy observed along the diagonal of the matrix, where testing templates were correctly matched to the corresponding training templates. A time window from 0.25 to 5.75 seconds post-stimulus onset (5,500 ms) was used to capture both ON and OFF responses. The classification success rate for the neuronal population (n = 86) was 87.08% (FIG.5C). Notably, five out of eight PFAS achieved 100% classification accuracy, while success rates for other stimuli were lower (FIG. 5C). RMS data (n = 51) yielded an 85% classification success rate, with 100% accuracy for four of the PFAS (FIG. 2C).

[0200] To further assess the precision of PFAS classification, sensitivity and specificity values were calculated. These values were derived from the LOTO confusion matrix, where columns represented true positives (TP) and false negatives (FN), and rows represented false positives (FP) and TP. Any data outside the relevant odor column or row was considered a true negative (TN). True positives were identified as the diagonal classification accuracy, where the true label was correctly assigned to the predicted label. It was observed that most PFAS achieved sensitivity and specificity values of 90% or higher, with three odors exhibiting lower sensitivity values (FIG.5D). Additionally, when discriminating between PFAS families (e.g., FTOH, FOSE, FOSA), all families achieved sensitivity and specificity values of approximately 90% or higher, except for one, which showed reduced sensitivity (FIG. 5D). When distinguishing between PFAS and air (non-PFAS), the sensor demonstrated 100% precision (FIG. 5D). For real-time RMS analysis, a 90% sensitivity and specificity were obtained for all but three PFAS and the FOSA family, showing decreased sensitivity (FIG. 2D). This cyborg gas sensor generated by the inventors can accurately differentiate between multiple PFAS, PFAS families, and controls.EXAMPLE 4: Population Responses Can Differentiate Environmental PFAS Concentrations

[0201] Neural responses were generated to a subset of PFAS at environmentally relevant concentrations (i.e., ppb and ppt) to investigate the sensor's detection limit. For the PFAS tested - 6:2 FTOH and 6:2 FTAcr, both receiving 100% classification in previous experiments - two concentrations were selected that mimic environmental conditions (0.01% vol / vol and 0.0001%TEC2025-0129 I TEC2025-0177 I TEC2026-00016550-000538-WO-POAvol / vol diluted in ddH2O). The previously high classification accuracy, along with strong odor-evoked responses, made these compounds ideal for examining neural responses at environmental levels. Parts per billion (ppb) concentrations were calculated using Raoult’s Law. with ppt levels observed at the lowest concentration (FIG. 6A). Odor stimuli were delivered to the locust antennae using previously described methods (FIG. 1A-B).

[0202] Interestingly, recordings from multiple antennal lobes (ALs) revealed that neurons responded differently to PFAS concentration and the odor control. A representative neuron demonstrated stimulus-specific changes in spiking activity, visualized via raster plots and corresponding firing frequency shifts shown in PSTHs (FIG. 6B). This data suggests that the sensor can detect PFAS odors at trace concentrations. To further characterize neural dynamics, RMS analysis was applied using discrete 50 ms time bins to visualize and classify responses in real-time (n = 34). The most significant neural activity changes occurred between 0.25 - 1.0 seconds following stimulus onset. LDA revealed clear separability of neural response clusters for each PFAS, its concentrations, and the control, indicating distinct spatial encoding for each condition (FIG. 6C). PCA further demonstrated unique neural trajectory evolutions over the course of odor presentation (n = 34, FIG.6D). Odor-specific response vectors, linked across consecutive 50 ms bins, formed distinct trajectories, all aligned to a common origin. The divergence of these trajectories from the origin and other stimuli underscores the specificity of the voltage trace responses. Overall, these results indicate that the cyborg sensor can detect and encode differences in concentration of the same PFAS, enabling discrimination through spatiotemporal neural activity patterns.

[0203] Classification success rates remained high for both PFAS across varying concentrations during quantitative analysis. A high dimensional LOTO analysis was employed to assess odor classification accuracy. A post-stimulus time window of 0.25 - 5.75 seconds was used for trial-wise classification, encompassing both ON and OFF phases of neural population responses (n = 34). Within this window, classification accuracy reached 84% (FIG. 6E). The sensor consistently classified all PFAS and concentrations with an accuracy of 80% or higher (FIG. 6E). Additionally, sensitivity and specificity analyses displayed values above 80% for all stimuli, including the comparison of PFAS versus control. A slight drop in sensitivity was observed for the lowest concentration (0.0001%) (FIG.6F). These findings quantitatively demonstrate that the cyborg sensor can reliably distinguish between different PFAS and their concentrations at environmental levels (i.e., ppb and ppt).TEC2025-0129 I TEC2025-0177 I TEC2026-00016550-000538-WO-POAEXAMPLE 5: Population Responses Can Differentiate PFOS At Environmental Concentrations

[0204] Stimuli tested thus far have primarily included volatile PFAS with measurable vapor pressures. While these compounds are detectable by the locust olfactory system at environmentally relevant concentrations, they are not the most commonly found PFAS (L. Liang et al., Immunotoxicity mechanisms of perfluorinated compounds PFOA and PFOS. Chemosphere 291, 132892 (2022); S. Kurwadkar et al., Per-and polyfluoroalkyl substances in water and wastewater: A critical review of their global occurrence and distribution. Science of The Total Environment 809, 151003 (2022); Z. W. Tang, F. S. Hamid, I. Yusoff, V. Chan, A review of PFAS research in Asia and occurrence of PFOA and PFOS in groundwater, surface water, and coastal water in Asia. Groundwater for Sustainable Development 22, 100947 (2023)).

[0205] Perfluorooctanesulfonic acid (PFOS), one of the most environmentally prevalent PFAS, is known for its extremely low vapor pressure, making it effectively non-volatile and challenging to detect via gas (H. Ryu, B. Li, S. De Guise, J. McCutcheon, Y. Lei, Recent progress in the detection of emerging contaminants PFASs. Journal of Hazardous Materials 408, 124437 (2021)). To investigate whether the sensor can detect common but non-volatile PFAS, PFOS was tested at environmental concentrations (0.01% and 0.0001%). Ppb and ppt values were calculated using Raoult’s law and known physicochemical properties of PFOS. The lowest concentration tested (0.0001%) represented approximately 34 ppt, simulating real- world environmental conditions (FIG.6A). Compared to previously tested PFAS, PFOS concentrations represent lower vapor-phase levels (FIG. 3A). Odor stimuli were delivered pseudo randomly to the locust antennae and odor-evoked responses were recorded from multiple ALs. Distinct differences in neural activity were observed between PFOS concentrations and the odor control. A representative neuron demonstrated concentration- specific spiking patterns in raster plots and corresponding PSTHs (FIG.7B). Neuronal responses were distinguishable from the odor control and between PFOS concentrations, indicating that the locust neural circuitry can discriminate minute PFOS concentrations, independent of the background matrix (FIG.7B).

[0206] To visualize odor-evoked neural responses in real-time, an RMS analysis was applied using non- overlapping 50 ms time bins (n = 34). The most significant changes in neural activity occurred between 0.25 - 1.0 seconds after stimulus onset. Using LDA, clear separation of neuralTEC2025-0129 I TEC2025-0177 I TEC2026-00016550-000538-WO-POAclustering into three distinct groups was observed (FIG. 7C). Notably, the odor control formed a cluster in a spatial region opposite from those representing PFOS responses, indicating that the locust olfactory system uses different neural spatial codes to represent concentrations of PFOS. When combining PFOS data with previously tested PFAS concentrations, high-concentration PFAS (0.01%) and the odor control each formed well- separated clusters, while lower concentrations (0.0001%) showed more spatial overlap (FIG.3B). Despite this overlap, the 0.0001% responses still exhibited partial separation, suggesting that the sensor can differentiate multiple PFAS at ppt levels. PCA revealed distinct neural trajectory evolutions for the control versus PFOS stimuli (FIG. 7D).The control trajectory diverged separately from PFOS concentrations. Both PFOS concentrations showed greater expansion from the origin, reflecting active and differentiated neural responses to low concentrations. Additionally, differences in trajectory evolution were evident between the PFOS concentrations (FIG.7D). When integrating PFOS trajectories with previous PFAS compounds, each PFAS exhibited distinct trajectories (FIG. 3C). Controls consistently followed a trajectory opposite to PFAS. While trajectories of the same compound at different concentrations were generally similar, they retained distinct differences. These findings emphasize that the locust neural circuitry encodes unique signatures for environmental PFOS concentrations, distinguishable from other PFAS and the control.

[0207] For quantitative classification, a high dimension EOTO analysis was used. Trial-wise classification was performed using a 0.25 - 5.75 second post-stimulus window, encompassing both ON and OFF phases of neural activity (n = 34). Analysis achieved a 100% classification success rate, accurately distinguishing environmental PFOS concentrations and the control (FIG.7E). Sensitivity and specificity values showcased PFOS concentrations and PFAS versus control comparison achieving 100% (FIG. 7F). When combining PFOS data with previous PFAS responses, the overall classification accuracy remained high at 85.71% (n = 34, FIG. 3D). All PFAS concentrations achieved an accuracy of 80% or higher, with the exception of 0.0001% 6:2 FTOH. Sensitivity and specificity for the combined dataset were greater than 80% across PFAS concentrations, with minor decreases in 0.0001% 6:2 FTOH sensitivity (FIG. 3E). Findings indicate that the sensor can reliably distinguish between PFOS and other PFAS at environmental concentrations. The locust-based cyborg sensor exhibits high sensitivity and specificity, making it a powerful tool for detecting trace levels of PFOS and PFAS.TEC2025-0129 I TEC2025-0177 I TEC2026-00016550-000538-WO-POA

[0208] These findings showcase a highly sensitive and specific biological cyborg sensor capable of distinguishing multiple PFAS at environmental concentrations. Individual neurons respond to minute changes in PFAS identity and concentrations, compared to control odors (FIG. 4A-C, FIG.5A-D, FIG. 6A-F, and FIG. 7A-F). Differences in neural spiking frequency were observed across varying concentrations of PFAS as well (FIG.6A-F and FIG.7A-F). These variations correspond to unique spatio-temporal responses and distinct encoding patterns for PFAS identity versus concentration, highlighting the sensor’s precise ability in detecting multiple PFAS. Furthermore, PFOS is the most abundant PFAS found in the environment, characterized by its non-volatile properties and presence at ppt concentrations (R. Xu et al., Effects of perfluorooctanoic acid (PFOA) and perfluorooctane sulfonic acid (PFOS) on soil microbial community. Microbial ecology, 1-13 (2022); K. Schulz, M. R. Silva, R. Klaper, Distribution and effects of branched versus linear isomers of PFOA, PFOS, and PFHxS: A review of recent literature. Science of the Total Environment 733, 139186 (2020)).

[0209] In this study, the inventors have discovered a method for detecting PFOS at environmental concentrations in real-time, revealing neural activity changes to PFOS relative to baseline, with clear distinctions across concentrations (FIG. 7B). Spatio-temporal encoding patterns were identified that accurately distinguished environmental PFOS concentrations in real-time (FIG. 7C-F). This is the first study to demonstrate a brain-based cyborg gas sensor capable of detecting a broad range of PFAS compounds, including PFOS, at ppt levels in real-time. These findings address several limitations of current PFAS detection technologies, paving the way for employing a brain-based, cyborg sensor for environmental pollution monitoring.EXAMPLE 6: General Methods for Detecting Lung Cancer

[0210] Design and microfabrication of the flexible Dual-Sided Polymer microelectrode arrays (MEAs)

[0211] FIG. 8C shows a graphical depiction of the device fabrication steps. A 5-pm-thick layer of parylene C was deposited on a 4-inch-diameter silicon wafer using room temperature chemical vapor deposition (SCS Labcoter® 3, Specialty Coating Systems, Indianapolis, IN, USA). Then, a metal stack of Ti (10 nm) and Au (200 nm) was sequentially deposited using a thermal evaporator (Edwards Auto 306, Edwards, UK). The electrode sites, traces, and contact pads of the microelectrodeTEC2025-0129 I TEC2025-0177 I TEC2026-00016550-000538-WO-POAarrays (MEAs) were defined by ultraviolet (UV) photolithography of a positive photoresist (PR, S 1813, Shipley, Marlborough, MA, USA), and patterned by wet etching in gold etchant (MILLIPORE SIGMA®, St. Louis, MO, USA). After that, photoresist was removed with acetone, followed by rinsing with ethanol and deionized (DI) water. Next, a second 5-pm-thick layer of parylene was deposited as the top encapsulation, followed by thermal evaporation of a 200-pm-thick layer of copper. The outline of the planar MEA, the folding slot, and open windows of electrode sites were defined by UV photolithography of photoresist. Subsequently, copper was patterned by wet etching in copper etchant (MILLIPORE SIGMA®, St. Louis, MO, USA) to form as a hard mask. Oxygen plasma etching was performed in a reactive ion etching system (RIE-1701 plasma system, Nordson March, Inc, Westlake, OH, USA) to remove unwanted parylene C, to expose the electrodes sites and contact pads as well as to form the shape of the planar MEAs. After that, the planar MEAs were released from the silicon substrate, aligned and folded along the middle line, and then thermally bonded to form the dual-sided MEAs. The bonded dual-sided MEAs were assembled onto the printed circuit board (PCB) and electrically bonded using conductive silver paste (8331- 14G. MG Chemicals. Canada).

[0212] Electropolymerization of PEDOT: PSS on gold electrodes

[0213] Electropolymerization of poly(3,4-ethylenedi oxythiophene) (PEDOT): polystyrene sulfonate) (PSS) on gold electrodes was achieved using a cyclic voltammetry (CV) technique with a three-electrode setup, in which the MEAs act as the working electrode (WE), a platinum wire was utilized as the counter electrode (CE), and a standard Ag / AgCl electrode as the reference electrode (RE). This process was conducted in a monomer solution of 10 mM 3,4-Ethylenedioxythiophene (EDOT, MILLIPORE SIGMA®, St. Louis, MO, USA) and 0.7 wt.% Poly (sodium 4- styrenesulfonate) (PSS, MILLIPORE SIGMA®, St. Louis, MO, USA) in DI water. The solution was stored in a 4°C refrigerator and deoxygenated by purging with nitrogen gas and then vacuuming for 20 mins before use. CV was performed using a potentiostat (6149E, CH Instruments, Austin, TX, USA), with voltage ranging from -0.7 V to 0.9 V vs. the Ag / AgCl RE and a scan rate of 10 mV / s.

[0214] Approaches to characterizing the MEAs:

[0215] Electrochemical impedance spectroscopy (EIS):

[0216] EIS was conducted using a potentiostat (6149E, CH Instruments. Austin, TX, USA) in a phosphate buffered saline (PBS) solution in a three-electrode setup with the MEA WE, a Pt CE andTEC2025-0129 I TEC2025-0177 I TEC2026-00016550-000538-WO-POAan Ag / AgCl RE. The alternating-current impedance (A.C. impedance) of the electrodes was measured in a wide frequency range of 1 Hz to 100 kHz and acquired using commercial CHI software (CH Instruments).

[0217] Surface morphology and effective surface area:

[0218] Scanning electron microscopy (SEM, Hitachi® S-4700, Hitachi High Technologies America, Inc., USA) was utilized to analyze the surface morphology of the electrodes with and without the PEDOTiPSS coating. Surface roughness was measured with an optical profilometer (WYKO nt2000, Broker®, MA).

[0219] Effective surface area:

[0220] Effective areas measurement:

[0221] In this experiment, the CV responses of the electrode were electrochemically measured in 0.5 mM K3[Fe(CN)6] at varying scan rates of 0.02 V / s, 0.05 V / s, 0.1 V / s, 0.2 V / s, and 0.5 V / s. The effective areas of electrodes were calculated using the Randles-Sevcik equation (Tran, L. T„ Tran, H. V., Cao, H. H., Tran, T. H., & Huynh, C. D. (2022). Electrochemically Effective Surface Area of a Polyaniline Nanowire-Based Platinum Microelectrode and Development of an Electrochemical DNA Sensor. Journal of Nanotechnology, 2022, 8947080; Trouillon, R., Lin, Y., Mellander, L. J., Keighron, J. D., & Ewing, A. G. (2013). Evaluating the diffusion coefficient of dopamine at the cell surface during amperometric detection: disk vs ring microelectrodes. Anal Chem, 85(13), 6421-6428).ip= 2.69xl0sn2AD2Cv2 (1)where ipis the cathodic peak current, n is the number of transported electrons, A (cm2) is the effective electrochemical surface, D (cm2- s-1) is the diffusion coefficient, C (mol- cm-3) is the concentration of redox species, and v (V s1) is the potential scan rate. The dependency of the anodic peak potentials on the natural logarithm of the potential scan rate was investigated and linear regression lines were generated using the cathodic peak current and the square root of the potential scan rate as variables.

[0222] Mechanical properties for bending and insertion:

[0223] To further evaluate the mechanical robustness of the electrodes, additional tests were conducted to assess their performance under bending and insertion conditions. Electrodes were bentTEC2025-0129 I TEC2025-0177 I TEC2026-00016550-000538-WO-POAto angles of 45° and 90° (n = 4), and their EIS properties were measured both before and after bending (FIG. 91).

[0224] A brushed motor controller (KDC101, Thorlabs®, USA) was utilized to perform the insertion of the dual-sided MEAs into a brain phantom (0.6 wt.% agarose gel), with a penetration depth of about 500 pm inside the gel (shown in FIG. 9G). The electrochemical impedance of the electrode was measured before and after 10 repeated insertions. The device’s EIS was tested before and after insertion.

[0225] Locust surgery and implantation of the MEAs

[0226] Insect surgery procedure was performed following pre-established methodology (Farnum, A., Parnas, M., Apu, E.). Briefly, locusts were immobilized onto a surgical platform and antennae were stabilized. A batik wax bowl around the head was constructed to isolate the region and then filled with a room temperature, physiologically balanced locust saline solution. The brain was kept hydrated with locust physiological saline solution described by Laurent et al. (in mM as follows: 140 NaCl. 5 KC1. 5 CaCh, 4 NaHCCh, 1 MgCh, 6.3 HEPES, pH 7.0; all chemicals from Sigma- Aldrich®) (Laurent, G., & Davidowitz, H. (1994). Encoding of olfactory information with oscillating neural assemblies. Science, 265(5180), 1872-1875; Laurent, G., & Naraghi, M. (1994). Odorant-induced oscillations in the mushroom bodies of the locust. The Journal of Neuroscience, 14(5), 2993).

[0227] The removal of the exoskeleton between the antennae was performed and glandular tissue was removed until the brain was fully visible. Treatment with protease was done to desheath the antennal lobes. After undergoing surgery, the insect was secured on a stage within a Faraday cage. The MEA was mounted on a micromanipulator and inserted into the appropriate location within the locust olfactory brain under a microscope, at an approximate speed of 0.05 mm / s. The MEA is implanted in the locust brain, with an insertion area of about 500 pm in diameter within the antennal lobe and a depth of about 100 pm into the brain tissue. The tip region of the 8-channel MEA was 230 pm wide and 20 pm thick, with an electrode diameter of 20 pm and a trace spacing of 20 pm. All electrodes were positioned within 100 pm of the tip.

[0228] Odor vial preparation and odor delivery

[0229] For the individual lung cancer biomarkers, the following odorants were used: nonanal, hexanal, pentanal, and propylbenzene (all from Sigma-Aldrich®). For the initial experiments displayed in FIG. 10A-F, each of these odors was diluted in 10 mL of mineral oil (MILLIPORETEC2025-0129 I TEC2025-0177 I TEC2026-00016550-000538-WO-POASIGMA®, St. Louis, MO, USA) at a 1% vol / vol concentration, and pure mineral oil was used as a control. Odor delivery was performed (FIG. 10A) following our pre-established methodology (Farnum, A., Parnas, M., Hoque Apu. E„ Cox, E., Lefevre. N.. Contag, C. H., & Saha, D. (2022). Harnessing insect olfactory neural circuits for detecting and discriminating human cancers. Biosens Bioelectron, 219, 114814). Briefly, a commercial olfactometer (Aurora Scientific®, 220A) was used for controlled odor stimulus delivery (FIG. 10A). Initially, at the beginning of each trial, 200 standard cubic centimeters per minute (seem) of zero contaminant air was delivered through a 1 / 16-inch diameter polytetrafluoroethylene (PTFE) flow line to the locust antenna denoted the stimulus flow line. The end of the stimulus flow line was placed approximately 2-3 cm from the most distal antennal segment. An additional 200 seem of zero contaminant air was passed through a separate flow line to the exhaust denoted the dilution flow line. Five seconds before odor stimulus delivery, 40% (80 seem) of the dilution flow line was redirected through the odor flow line directly upstream of the odorant vials. The dilution flow line and the odor flow line joined downstream of the odorant vials. This allowed for the 80 seem odor flow to mix with the 120 seem of the dilution flow line’s clean air.

[0230] The air-volatiles mixture primed the line with volatiles up to the final valve, where the combined odor + dilution flow was delivered to the exhaust. Upon stimulus onset, the final valve redirected the clean air flow to the exhaust and the odor + dilution flow to the locust antenna via the stimulus flow line. After 4 s of constant flow and odor stimulus delivery, the final valve redirected the clean air flow back to the locust antenna via the stimulus flow line and the odor + dilution flow back to the exhaust. One second after odor stimulus offset, the 80 seem flowing through the odor line was redirected through the dilution flow line, alleviating potential headspace gas depletion during the odor delivery. This protocol was designed to keep a constant flow rate through the stimulus flow line, thereby eliminating any potentially confounding neuronal responses due to changes in air pressure via mechanosensory detection.

[0231] A 6-inch diameter funnel pulling a slight vacuum was positioned immediately behind the locust during odor stimulus delivery to ensure rapid removal of odorants. Each four second-duration odor stimulus was repeated five times with an interstimulus interval of 60 seconds. The order of the stimuli was pseudorandomized for all experiments including the 1% biomarker vials, the varying concentration biomarker vials, and the cell culture. It should be noted that the VOCs were delivered to the locust antennae, which was physically separate from the recording location (i.e., antennal lobe) where MEAs were inserted. This experimental setup assured environmental stability within the VOCTEC2025-0129 I TEC2025-0177 I TEC2026-00016550-000538-WO-POAdelivery region at the biological sensor (i.e., locust antenna) and the data collection region at the antennal lobe.

[0232] For biomarker concentration vial preparation, two preselected biomarkers were chosen (nonanal and propylbenzene) and diluted in 10 mL of mineral oil at 0.1% vol / vol and 0.01% vol / vol. To create the 0.0001% vol / vol concentration, a serial dilution with the 1% vol / vol vials was used in 10 mL of mineral oil. To calculate the ppm and ppb values for each of these concentrations, the vapor pressure of each VOC (nonanal and propylbenzene) was used with Raoult’s Law to obtain the gas phase concentration of each chemical compound (FIG. 12A).

[0233] Cell culture preparation

[0234] For cell odor experiments, four human lung cancer cell lines were used, including two non-small cell adenocarcinoma cell lines: NCI-H1437 [stage 1] (derived from site: metastatic; pleural effusion) and NCI-H1573 [stage 4] (derived from site: in situ; lung), and two small cell carcinoma cell lines: H69PR (derived from site: metastatic; pleural effusion) and SHP-77, non-adherent cells (derived from site: in situ; lung, left upper lobe). Non-cancer cell line, primary normal healthy lung fibroblast (HLF) (derived from site: in situ, lung), was used as a non-cancer control. All cell lines were purchased from the American Type Culture Collection (ATCC®).

[0235] The cell culturing method followed Parnas, M., McLane-Svoboda, A. K., Cox, E., McLane-Svoboda, S. B., Sanchez, S. W., Farnum, A., . . . Saha, D. (2024). Precision detection of select human lung cancer biomarkers and cell lines using honeybee olfactory neural circuitry as a novel gas sensor. Biosensors and Bioelectronics, 261, 116466. All cancer cell lines were seeded at a density of 1.2 x 106cells per flask in surface-treated sterile tissue culture flasks (T75, Nunc® EasYFlask®, with vented caps, Thermo Fisher Scientific®, MA, USA) using Roswell Park Memorial Institute medium (RPMI) 1640 medium, supplemented with 10% heat-inactivated fetal bovine serum (FBS) and 1% penicillin-streptomycin (10,000 U / mL). The medium, FBS, and penicillinstreptomycin were all purchased from Thermo Fisher Scientific®, MA, USA. The HLF control cell line was cultured in Fibroblast Basal Medium (ATCC®) with the Fibroblast Growth Kit-Low Serum (ATCC®) in the same T75 flasks under identical conditions (37°C in 5% CO?).

[0236] After four days, all adherent cells were detached using 6 mL of Accutase® solution (Sigma- Aldrich®) for five minutes, centrifuged at 160xg for six minutes, and the supernatant was removed. All cell lines were resuspended in 4 mL of RPMI complete medium at a density of 2.25 xTEC2025-0129 I TEC2025-0177 I TEC2026-00016550-000538-WO-POAIO6cells per modified T-25 flask. The cell lines and an RPMI complete medium control were incubated at 37°C in 5% CO2, followed by olfactory testing. Cell culture images were captured on the Olympus® CKX53 microscope using the Olympus cellSens® Entry 2.1 (Build 17342) the day of experimentation.

[0237] To better facilitate the odor delivery from flasks, the T-25 flasks were modified in a sterile cell culture hood before cell culturing. Using a handheld Dremel® tool, 0.7 mm holes were drilled at the top rear of the flask and through the cap. Inlet and outlet 20-gauge needles were inserted into each hole, with the sharp ends cut off. The bases of the needles were secured with a low-volatile, two-part Pennatex® epoxy at least 24 hours before seeding. The ends of the needles were then covered with a cap during cell culture preparation.

[0238] Electrophysiology

[0239] All in vivo extracellular neural recordings were performed on post-fifth instar locust (Schistocerca americana) of either sex raised in a crowded colony. For population responses, in vivo extracellular neural data were collected from two locusts (from 12 different recordings). VOC-evoked neural recordings were made from the antennal lobe of the locust brain, which was submerged into locust saline (pH 7.0) as described in earlier publications (Saha, D., Leong, K., Li, C., Peterson, S., Siegel, G., Raman, B„ 2013c. A spatiotemporal coding mechanism for background-invariant odor recognition. Nature Neuroscience 16(12), 1830-1839; Saha, D., Li, C., Peterson, S., Padovano, W., Katta. N., & Raman, B. (2015). Behavioural correlates of combinatorial versus temporal features of odour codes. Nature Communications, 6(1), 6953; Saha, D., Sun, W., Li, C., Nizampatnam, S., Padovano, W., Chen, Z., . . . Raman, B. (2017). Engaging and disengaging recurrent inhibition coincides with sensing and unsensing of a sensory stimulus. Nature Communications, 8(1), 15413). This saline solution was used to keep the brain hydrated. Spontaneous firing activities were constantly monitored (without any odorant) of the recorded neurons to assure that the neural recording setup is stable over time. Initial lung cancer VOC biomarker experiments were performed using a VOC panel of four VOCs (nonanal, hexanal, pentanal, propylbenzene) and mineral oil. Subsequent concentration experiments included stimulus panels of 0.1% vol / vol, 0.01% vol / vol, and 0.0001% vol / vol for both nonanal and propylbenzene diluted in mineral oil and an additional odor vial with pure mineral oil used as a control. Human lung cancer cell line experiments consisted of a panel with four cultured cell lines and a media control. The cultured cell lines included two non-small cell lung cancerTEC2025-0129 I TEC2025-0177 I TEC2026-00016550-000538-WO-POA(NSCLC) lines, two small cell lung cancer lines (SCLC), and one healthy lung fibroblast cell line used as a control.

[0240] Locust surgery was performed according to previously published methods (Saha, D., Leong, K., Katta, N., Raman, B„ 2013a. Multi-unit recording methods to characterize neural activity in the locust (Schistocerca americana) olfactory circuits. J Vis Exp(71); Saha, D., Leong, K., Li, C., Peterson, S., Siegel, G., Raman, B., 2013b. A spatiotemporal coding mechanism for backgroundinvariantodorrecognition. Nat Neurosci 16(12), 1830-1839). Briefly, locusts were immobilized onto a surgical platform and antennae were stabilized. A batik wax bowl around the head was constructed to isolate the region and then filled with a room temperature, physiologically balanced locust saline solution. The removal of the exoskeleton between the antennae was performed and glandular tissue was removed until the brain was fully visible. Treatment with protease was done to desheath the antennal lobes. Following surgery, the animals were placed in a Faraday cage isolated bench and a silver-chloride ground wire was placed in the saline bath. Voltage signals from antennal lobe projection neurons were recorded by inserting the multichannel MEA with an impedance between 20 and 50 kQ into the antennal lobe (FIG. 10A). Voltage signals were sampled at 20 kHz and then digitized using an Intan® pre-amplifier board (C3334 RHD 32-channel headstage). The digitized signals were transmitted to the Intan® recording controller (C3100 RHD USB interface board) before being visualized and stored using the Intan® graphical user interface and Lab VIEW® data acquisition system.

[0241] In vivo extracellular neural data for the 1% vol / vol cancer biomarkers were collected from two locusts. The electrode was placed into the AL of the locust 12 times and a total of 24 electrophysiological recordings were conducted. In vivo extracellular neural data for the varying concentration experiments were conducted from six locusts. The electrode was placed into the AL of the locust 18 times and a total of 29 electrophysiological recordings were conducted. In vivo extracellular neural data for the cell line experiments were conducted from three locusts. The electrode was placed into the AL of the locust ten times and a total of 20 electrophysiological recordings were conducted. Every antennal lobe projection neuron recording was conducted using the same panel depending on the experiment and each stimulus delivery to the antenna was repeated five times with a 60 s inter-stimulus-interval for repeatability. Neural responses were pooled across animals due to the identifiable neurons across animals and conditions satisfied by previous studies (Burrows, M., 1996. The Neurobiology of an Insect Brain. Oxford University Press; Marin, E.C.,TEC2025-0129 I TEC2025-0177 I TEC2026-00016550-000538-WO-POAJefferis, G.S.X.E., Komiyama, T., Zhu, EL, Luo, L., 2002. Representation of the Glomerular Olfactory Map in e. Drosophila Brain. Cell 109(2), 243-255; Stopfer, M., Jayaraman, V., Laurent, G., 2003. Intensity versus identity coding in an olfactory system. Neuron 39(6), 991-1004; Wong, A.M., Wang, J.W., Axel, R., 2002. Spatial Representation of the Glomerular Map in the Drosophila Protocerebrum. Cell 109(2), 229-241).

[0242] Spike Sorting

[0243] All neural data were imported into MATLAB® after high pass filtering using a 300 Hz Butterworth filter to eliminate frequencies below 300 Hz. The data were analyzed by custom-written codes in MATLAB® R2023b. All data were processed with Igor Pro® for spike sorting analysis using previously described methods (Pouzat, C., Mazor, O., Laurent, G„ 2002. Using noise signature to optimize spike-sorting and to assess neuronal classification quality. J Neurosci Methods 122(1), 43-57) or an offline spike sorting software ROSS (Toosi, R., Akhaee, M.A., Dehaqani, M.-R.A., 2021. An automatic spike sorting algorithm based on adaptive spike detection and a mixture of skew-t distributions. Scientific Reports 11(1), 13925). Spiking events were identified using a detection threshold between 2.5 and 3.5 standard deviations (SD) of baseline fluctuations. Individual projection neurons were identified if they passed the following criteria: cluster separation > 5 SD, inter-spike intervals (ISI) < 10%, and spike waveform variance < 10%. A total of 69 projection neurons were identified from 12 recordings displayed in FIG. 10A-F. Spike sorted data were used for analysis in FIG. 10A-F, FIG. 11A-C, FIG. 12A-E, and FIG. 13A-E.

[0244] Root mean square (RMS) Transformation

[0245] All neural data were imported into MATLAB® after high pass filtering using a 300 Hz Butterworth filter to eliminate frequencies below 300 Hz. The filtered data were trimmed to the time window of interest and all data were passed through a 500-point continuous moving root mean squared (RMS) filter followed by a smoothing step via a 500-point continuous moving average filter to the raw voltage data as described in our previous work (Farnum, A., Parnas, M., Apu. E.). Stimulusspecific baseline values were calculated as the average voltage over all time bins for the two seconds prior to the stimulus onset. Baseline responses were averaged over all trials and subsequently subtracted from the data to obtain the change in root mean squared values (i.e., A RMS). These values were binned into non-overlapping 50 millisecond bins and the average of each bin was computed. For each recording location, root mean squared transformed voltage data of each 4, 6, or 8-channelTEC2025-0129 I TEC2025-0177 I TEC2026-00016550-000538-WO-POAelectrode were averaged together. RMS transformed data were used for analysis in FIG. 12A-E and FIG. 13A-E.

[0246] Scatter plots

[0247] The total number of spikes for each trial during the four second stimulus presentation window was summed and averaged over the five trials. The standard error of the mean (SEM) was calculated across the trials for each neuron. Both the mean spike counts and the SEM were plotted for two stimulus conditions along the X- and Y-axes (e.g., nonanal spike counts vs. mineral oil spike counts). Statistically significant differences in mean spike counts for different stimulus conditions were determined by applying a one-way ANOVA with Bonferroni correction due to multiple comparisons (P < 0.05, d.f. = 4, 20. one-way ANOVA with Bonferroni correction).

[0248] Dimensionality reduction analysis

[0249] Two methods of dimensionality reduction were performed - PCA and LDA. In PCA, neural signals were binned baseline subtracted, spike sorted in 50 ms non- overlapping time bins and averaged across trials (n = 5, each VOC was repeated five times with a 60 s inter-stimulus interval). The baseline response was calculated for each neuron by averaging the firing rate over the two second time window prior to odor stimulus onset across trials. The spike sorted neurons were pooled across electrophysiological experiments. For example, in FIG. 10E, spike sorted and binned responses of all recorded neurons (69 total) were combined to generate a neuron number (n = 69) x time bins (t = 30, number of 50 ms bins between 0.25 - 1.75 s) matrix, where each element in the matrix corresponds to the spike count of one neuron in one 50 ms time bin. Similar neural population time-series data matrices were generated for each odor stimulus. PCA dimensionality reduction analysis was performed on the time-series data involving all four lung cancer biomarkers (nonanal, hexanal, pentanal, and propylbenzene) including mineral oil control and directions of maximum variance were found (FIG. 10E).

[0250] The consequent high-dimensional vector in each time bin was projected along the eigenvectors of the covariance matrix. Only the three dimensions with the highest eigenvalues were used for visualization, and data points in subsequent time bins were connected to produce lowdimensional neural trajectories. The trajectories were smoothed using a third order IIR Butterworth filter (Half Power Frequency = 0.15). Lastly, all trajectories were shifted to begin at the origin to analyze stimulus-specific response dynamics and trajectory divergence. The same PCA analysis wasTEC2025-0129 I TEC2025-0177 I TEC2026-00016550-000538-WO-POAapplied to the concentration panel for nonanal and propylbenzene by using the baseline subtracted RMS transformed population time-series data matrix. For LDA, neural population time-series data matrix was used for either spike sorted or RMS transformed data. Here, we maximized the separation between interclass distances while minimizing the within class distances. For visualization purposes, time bins were plotted as unique points in this transformed LDA space and stimulus- specific VOC clusters became easily noticeable (FIG. 12C and FIG. 13C). All dimensionality reduction analyses were accomplished using custom written MATLAB® (R2023b) codes.

[0251] Quantitative classification analysis: Leave-one-trial-out (LOTO)

[0252] To classify the individual lung cancer biomarkers, a leave-one-trial-out (LOTO) analysis was performed. Spike sorted or RMS transformed data was binned into 50 ms bins to create a three-dimensional matrix (e.g., 69 neurons x 5 trials x n time bins). For example, FIG. 10F uses 30-time bins for the time window between 0.25 - 1.75 s after stimulus onset, which corresponds to the phase of the odor-evoked neural activity that is most discriminatory termed the transient phase. For LOTO analysis, four trials (out of five) were averaged together to form a training template, while the fifth trial, the left-out trial, was used as the testing template.

[0253] This was repeated for all odors to create five training templates and 5 testing templates and each time bin from these templates was considered as points in high-dimensional space. In a binwise matched manner, the Euclidean or Manhattan distance between the testing template time bins and the training template time bins were calculated and the testing template time bins were assigned based on the shortest Euclidean distance. This was repeated for each of the five trials, each time leaving out a different trial for the testing template and using the remaining four trials to create the training template. The results for this type of analysis are summarized with a confusion matrix. This type of analysis is termed a bin-wise classification. Further analysis was completed by calculating the mode for each testing template across the n time bins for the time window of interest. The mode was used to classify the entire trial in a winner-take-all approach and the results are summarized in FIG.10F, FIG. 11C, FIG. 12D, and FIG. 13D. This analysis is termed a trial-wise classification. A fully diagonal matrix indicates 100% classification accuracy with any cells filled outside of this diagonal indicating misclassifications. All quantitative classification analysis was accomplished using custom written MATLAB® (R2023b) codes.TEC2025-0129 I TEC2025-0177 I TEC2026-00016550-000538-WO-POA

[0254] Sensitivity and Specificity

[0255] The sensitivity and specificity were calculated for the two biomarkers (nonanal and propylbenzene) at varying concentrations with respect to each individual concentration and pure mineral oil (FIG. 12E). For example, true positives are instances where a trial with a true label of nonanal 0.1% was predicted to be nonanal 0.1%. True negatives are instances where a trial with a true label of not nonanal 0.1% was predicted to have a label of not nonanal 0.1%. False negatives are instances where a trial with a true label of nonanal 0.1% was predicted to have a label of not nonanal 0.1%. False positives are instances where a trial with a true label of not nonanal 0.1% was predicted to have a label of nonanal 0.1%. This was repeated for each concentration individually. Similarly, the sensitivity and specificity were calculated with respect to each individual cell line and media (FIG.13E).EXAMPLE 7: Microfabrication processes of flexible dual-sided MEAs

[0256] FIG. 8A and FIG. 8B depict the conceptual design of the flexible dual- sided MEAs and their application for human lung cancer VOC detection. The MEAs feature a multilayer structure with parylene C (substrate and encapsulation) and metallization (electrodes, traces, and pads). Configurations include 4, 6, and 8 electrodes, with a 20 pm electrode diameter, a 50 pm minimal center to center spacing, and tip widths of 120, 180, and 230 pm, respectively. These dual-sided MEAs achieved electrode densities of 463, 687, and 766 channels / mm2.

[0257] Fabrication began with single-sided planar MEAs using microfabrication techniques like chemical vapor deposition, metal evaporation, and photolithography (FIG. 8C). Planar MEAs were released, folded along the middle line, and thermally bonded to form dual-sided structures. FIG. 8D shows microscope images of an 8-channel single-sided MEA and its dual-sided counterpart. The thermal bonding effectively enhanced adhesion between parylene layers, eliminating the Newton’s rings observed before bonding (Cartamil-Bueno, S.J., Steeneken, P.G., Centeno, A., Zurutuza, A., van der Zant, H.S.J., Houri, S., 2016. Colorimetry Technique for Scalable Characterization of Suspended Graphene. Nano Letters 16(11), 6792-6796).

[0258] The MEAs were assembled onto a custom printed circuit board (PCB) using conductive silver paste, with uniquely designed contact pads ensuring robust electrical connections. The flexible MEA tip and PCB-enforced backbone facilitated handling during insertion and recording (FIG.8E).TEC2025-0129 I TEC2025-0177 I TEC2026-00016550-000538-WO-POAElectrochemical impedance measurements after each fabrication step confirmed the stability of electrode properties, indicating no compromise from the fabrication processes (FIG. 8F and FIG.8G).EXAMPLE 8: Effect of PEDOT:PSS coating on performance of gold electrodes with various cyclic voltammetry (CV) cycles

[0259] The impact of repeated coating with PEDOT:PSS was investigated by conducting 1 to 4 CV cycles for electro-polymerization of PEDOT:PSS onto the gold electrodes, as shown in FIG.9A.After a single cycle of CV coating with PEDOT:PSS, the impedance at 1 kHz decreased significantly by two orders of magnitude. A second cycle of CV coating further decreased the impedance, with a greater reduction observed for smaller electrodes. However, a further increase in the number of CV cycles to three or four did not lead to a significant additional reduction in impedance (FIG. 9B-D).The impedance at 1 kHz for the 20 pm electrodes converged to a range between 103and 104ohms, suitable for neurophysiological recording. FIG. 9E shows the roughness of the surface of the electrodes with and without the PEDOT:PSS coating. The results indicate that the PEDOT:PSS coating increased the surface roughness of the electrodes; however, two or more cycles of coating did not further increase the surface roughness. The effective surface areas of the electrodes were quantified following the reported method (Tran, L.T., Tran, H.V., Cao, H.H., Tran, T.H., Huynh, C.D., 2022. Electrochemically Effective Surface Area of a Polyaniline Nanowire-Based Platinum Microelectrode and Development of an Electrochemical DNA Sensor. Journal of Nanotechnology 2022, 8947080; Trouillon, R., Lin, Y., Mellander, L.J., Keighron, J.D., Ewing, A.G., 2013. Evaluating the diffusion coefficient of dopamine at the cell surface during amperometric detection: disk vs ring microelectrodes. Anal Chem 85(13), 6421-6428.) As the CV coating cycles increased, the PEDOT:PSS coated electrodes exhibited an increased effective surface area (FIG. 9F), leading to decreased impedance of the electrodes. In the subsequent experiments, the MEAs were treated with two cycles of PEDOT:PSS as this protocol effectively optimizes the electrochemical impedance within the desired range for neurophysiological recording.TEC2025-0129 I TEC2025-0177 I TEC2026-00016550-000538-WO-POAEXAMPLE 9: Characterization of flexible dual-sided MEAs

[0260] To investigate the device’s insertion capability, the MEAs were inserted into a brain tissue phantom, 0.6% (w / v) agarose, and the electrochemical impedance of the electrodes was evaluated before and after undergoing 15 repeated insertions. The results indicated that the flexible dual-sided MEAs could be directly inserted into the brain tissue phantom without any mechanical reinforcement (FIG. 9G). The electrodes maintained their electrochemical performance after multiple insertions (FIG. 9H), demonstrating the device’s robustness and stability under repeated insertion conditions.

[0261] To evaluate the device’s mechanical flexible properties for insertion, we conducted tests where the electrodes were bent to 45° and 90° (n=4) and their electrochemical impedance spectroscopy (EIS) properties were evaluated before and after bending. The EIS properties did not show any significant changes, confirming that the electrodes remain stable under bending conditions (FIG. 91).

[0262] The noise levels of the microelectrodes with or without PEDOT:PSS coating were evaluated and quantified using simulated biosignals generated by a function generator. The basic noise level was 6.01 ± 0.24 pV for bare gold electrodes, which decreased to 2.52 + 0.65 pV for 2-cycle PEDOT:PSS coated electrodes (FIG. 9J). The signal-to-noise ratio (SNR) increased from 8.32 ± 0.24 to 23.22 ± 2.66 when the peak voltage of the signal applied was 100 pV, and from 18.77 ± 2.80 to 53.86 ± 5.44 when the peak voltage of the signal applied was 200 pV (FIG. 9K). Compared to pure gold electrodes, the PEDOT:PSS coated electrodes exhibit significantly lower base noise levels and higher SNR because of their lower impedance (FIG. 9L). Maintaining low noise levels during recording is crucial to detect action potentials from background noise, and to further isolate single neuron activities.

[0263] The capability of the flexible dual-sided MEAs for multichannel in vivo recording was further validated by recording of odor-evoked neural activities from the locust antennal lobe neurons. The raw neural voltage signals were successfully recorded concurrently from all channels using an 8-channel MEA, with action potentials observed following a 300 Hz high-pass filter. An example recording includes a peak voltage exceeding 309 pV, with an amplitude cutoff of 239.35 pV (data not shown). The spike-sorted clusters of the spike waveforms derived from the raw neural data show distinct neural units recorded in each channel. The MEAs' effectiveness in segmenting and characterizing neural signals was further tested. The inter- spike interval (IS I) histograms of three single units of sorted neural responses, which display positively skewed distributions, result in mostTEC2025-0129 I TEC2025-0177 I TEC2026-00016550-000538-WO-POAintervals clustering at shorter durations. Each unit lacks a peak around 1 ms, suggesting the recordings were free from high-frequency noise artifacts, confirming the reliability of the data.EXAMPLE 10: Detection and classification of human lung cancer biomarkers by locust antennal lobe neurons

[0264] Next, the responses of projection neurons in the antennal lobe to four different lung cancer biomarkers were examined. These biomarkers (nonanal, hexanal, pentanal, propylbenzene) have been shown to exist at different concentrations in exhaled breaths of patients with lung cancer (Fuchs. P., Loeseken, C., Schubert, J.K., Miekisch, W., 2010. Breath gas aldehydes as biomarkers of lung cancer. International Journal of Cancer 126(11), 2663-2670; Poli, D., Carbognani, P., Corradi, M., Goldoni, M., Acampa, O., Balbi, B., Bianchi, L„ Rusca, M., Mutti, A., 2005. Exhaled volatile organic compounds in patients with non-small cell lung cancer: cross sectional and nested short-term followup study. Respiratory Research 6(1), 71). Each biomarker was diluted in mineral oil (1% vol / vol) and pure mineral oil was used as a control odor. A schematic of the experimental setup is shown in FIG.10A. Lung cancer biomarker VOCs were delivered to the antennae of the locust via an olfactometer, and in vivo extracellular neural recordings were obtained from projection neurons (PNs) in the antennal lobe (FIG. 10A and FIG. 10B). Each VOC and mineral oil were exposed to the locust antenna five times for a total of five trials.

[0265] FIG. 10C shows a voltage trace for each VOC and mineral oil from a single electrophysiological recording. Distinct VOC-evoked neural responses were observed for each of the biomarkers as displayed in the voltage traces (FIG. 10C). From the voltage traces shown in FIG.10C, individual neurons were obtained via spike sorting, and the firing rate was observed by plotting peri-stimulus time histograms and raster plots (Pouzat, C., Mazor, O., Laurent, G., 2002. Using noise signature to optimize spike-sorting and to assess neuronal classification quality. J Neurosci Methods 122(1), 43-57). An individual neuron that was spike sorted from the voltage trace shown in FIG. 10C displayed a unique and distinct response to each of the lung cancer biomarkers (FIG. 10D). For the neuron shown in FIG. 10D, nonanal showed a high spiking frequency with the onset of the VOC stimulus with a continued increase of spiking throughout the duration of the four second stimulus in comparison to baseline spiking while propylbenzene showed inhibition with the onset of the stimulus (FIG. 10D).TEC2025-0129 I TEC2025-0177 I TEC2026-00016550-000538-WO-POA

[0266] From 24 electrophysiological recordings a total of 69 spike sorted neurons were spike sorted. To validate the differences in stimulus-evoked spikes, a one-way ANOVA analysis with Bonferroni correction was performed, comparing the total spike counts over four seconds between two stimulus conditions for all individual neurons (n = 69; P < 0.05, d.f. = 4, 20, one-way ANOVA with Bonferroni correction). With this analysis, it is clear that several individual neurons respond significantly higher or lower in comparison to mineral oil. The ability of a single neuron to have significantly different responses to each VOC showcases how individual neurons can differentiate VOCs at this level, however it should be noted that it is presumed that in the locust antennal lobe odor identity is not encoded by single neurons but by spatiotemporal population neural responses (Broome, B.M., Jayaraman, V., Laurent, G., 2006. Encoding and Decoding of Overlapping Odor Sequences. Neuron 51(4), 467-482; Mazor, O., Laurent, G., 2005. Transient Dynamics versus Fixed Points in Odor Representations by Locust Antennal Lobe Projection Neurons. Neuron 48(4), 661-673; Saha, D., Leong, K., Li, C., Peterson, S., Siegel, G., Raman, B., 2013c. A spatiotemporal coding mechanism forbackground-invariant odor recognition. Nature Neuroscience 16(12), 1830-1839; Saha, D„ Mehta. D., Altan, E., Chandak, R., Trailer, M., Lo, R., Gupta, P., Singamaneni, S., Chakrabartty, S., Raman, B., 2020. Explosive sensing with insect-based biorobots. Biosensors and Bioelectronics: X 6, 100050; Stopfer, M„ Jayaraman, V., Laurent, G., 2003. Intensity versus identity coding in an olfactory system. Neuron 39(6), 991-1004).

[0267] Therefore, next, a population of antennal lobe projection neurons (n = 69) was examined, and their responses to each of the lung cancer VOC biomarkers over a specific time window showed the evolution of the population response over time. Using dimensionality reduction techniques such as PCA and LDA, the data were reduced from a high dimensional dataset (69 dimensions) to a low dimensional subspace to visualize the population neural responses in a 3-dimensional PCA or LDA subspace (FIG. 10E). Using these techniques, it can be qualitatively seen that unique and separate trajectories in 3-dimensional PCA subspace represent the population neural responses to each of the lung cancer biomarkers over 1.5 seconds during the odor stimulus (FIG. 10E). The angles in which these trajectories move in different directions signify that these biomarkers are eliciting unique and distinct responses in the neuronal population (FIG. 10E). Furthermore, using LDA, distinct clusters of 50 ms time bins (30 total time bins for each biomarker) were observed, displaying that the population neural responses were able to differentiate each of the lung cancer biomarkers tested.TEC2025-0129 I TEC2025-0177 I TEC2026-00016550-000538-WO-POA

[0268] It was next sought to predict the classification success of each of the lung cancer biomarkers using a high dimensional, leave-one-trial-out (LOTO) analysis based on 50 ms time bins or 1.5 second trials in high dimensional space (i.e.. 69 dimensions) (FIG. 10F). To achieve this, training templates were created that consisted of the average spiking events of 50 ms time bins across four trials with the 5thtrial being left out to be used as a test template for each odor. This resulted in five training templates and five testing templates due to the four VOCs and mineral oil used in this study. Both the average of the four trials (training template) and the 5thtrial (testing template) consisted of 50 ms time bins over a specified duration of stimulus exposure time. PN responses were segregated into 50 ms time bins due to the 20 Hz local field oscillation present in the locust AL (Perez-Orive, J., Mazor, O., Tu). The averaged 50 ms time bins from the training template were used to classify the 50 ms time bins from the test template in a time binned-matched manner by considering each time bin as a point in the high dimensional space.

[0269] The Euclidean distances between each test time bin and all five training time bins were calculated with the test time bin being assigned to the biomarker with the associated closest training time bin. This was repeated for the total number of time bins (1.5 seconds = 30 time bins. 50 ms each) and cycled through each of the five trials so that each trial (1-5) would have the chance of being the trial that is left out (or test trial). In this way, the test template (trials 1-5) was classified for each VOC and mineral oil in a bin- wise manner and summarized the percentages of time bins classified for each biomarker in a confusion matrix. A schematic representation of this analysis is present in (Parnas, M., McLane-Svoboda, A.K., Cox, E„ McLane-Svoboda, S.B., Sanchez, S.W., Farnum, A., Tundo, A., Lefevre, N., Miller, S., Neeb, E., Contag, C.H., Saha, D., 2024. Precision detection of select human lung cancer biomarkers and cell lines using honeybee olfactory neural circuitry as a novel gas sensor. Biosensors and Bioelectronics 261. 116466). Using this method, it was observed that most of the time bins were classified correctly with an accuracy of 58.93%. The entire test trial was then classified by assigning the whole trial to the biomarker associated with the mode of the time bin classification in a trial-wise manner as shown in FIG.10F. The trial-wise classification results achieved 100% accuracy (FIG. 10F). This trial-wise analysis is used for all consecutive LOTO analysis due to the high classification accuracy.TEC2025-0129 I TEC2025-0177 I TEC2026-00016550-000538-WO-POAEXAMPLE 11 : The biosensing platform can classify lung cancer biomarkers efficiently from a single MEA recording

[0270] The classification of lung cancer biomarkers from a single experimental recording was next investigated, as opposed to combining all neurons across recordings. From a single experiment.12 neurons were obtained after spike sorting and are shown in FIG. 11A. These 12 neurons showed varied responses to all 4 VOC biomarkers and mineral oil including excitatory and inhibitory responses both during stimulus presentation (“On response”) and after the stimulus presentation (“Off’ response) as shown in FIG.11A. By combining the “On” response and “Off’ response across neurons for population analysis via PCA, it was observed that neural trajectories for each VOC biomarker moved in distinct and separate directions in low-dimensional PCA subspace (FIG. 11 B).These unique trajectories indicated that while the VOC-evoked responses from 12 neurons were obtained, these 12 neurons were able to encode the identity of these four lung cancer biomarkers and mineral oil, which can be utilized for high-dimensional classification analysis. By employing a high dimensional LOTO classification analysis. 1.25 second trials encompassing both the “On” and “Off’ responses were able to correctly classify the biomarkers with an accuracy of 88% (FIG. 11C).EXAMPLE 12: Detection and classification of human lung cancer biomarkers at multiple concentrations using locust antennal lobe neurons

[0271] To probe the concentration-dependent neural activity and investigate the capacity of antennal lobe projection neurons to differentiate multiple biomarker concentrations, experiments were conducted by exposing the locust antenna to varying concentrations of nonanal and propylbenzene with mineral oil as a control. Three concentrations were created (i.e., 0.1% vol / vol, O.01% vol / vol, 0.0001% vol / vol) by diluting each biomarker in mineral oil and the gas phase concentration values in ppm and ppb were calculated (FIG. 12A). Raoult’s law is an approximation and deviations from Raoult’s law are an ongoing point of research (Bowler, M.G., Bowler, D.R., Bowler, M.W., 2017. Raoult's law revisited: accurately predicting equilibrium relative humidity points for humidity control experiments. Journal of Applied Crystallography 50(2), 631-638; Tzias, P., Treiner, C., Chemla, M., 1977. Applicability of Raoult's law in nonideal mixed solvents. Journal of Solution Chemistry 6(6), 393-402; Wexler, A.S., 2019. Raoult Was Right After All. ACS Omega 4(7), 12848-12852). It should be noted that these concentrations progress towards biologically relevant concentrations (Fuchs, P., Loeseken, C., Schubert, J.K., Miekisch, W., 2010. Breath gasTEC2025-0129 I TEC2025-0177 I TEC2026-00016550-000538-WO-POAaldehydes as biomarkers of lung cancer. International Journal of Cancer 126(11), 2663-2670; Poli, D., Carbognani, P., Corradi, M., Goldoni, M., Acampa, O., Balbi, B., Bianchi, L., Rusca, M., Mutti, A., 2005. Exhaled volatile organic compounds in patients with non-small cell lung cancer; cross sectional and nested short-term follow-up study. Respiratory Research 6(1), 71). Each concentration was delivered to the locust antenna in a pseudorandom order using a procedure controlling for stimulus delivery, timing, and airflow.

[0272] For each biomarker concentration, odor-evoked neural responses were recorded from multiple antennal lobe neurons. From the same voltage trace, a neuron was spike sorted and raster and peri-stimulus time histograms (PSTHs) were plotted to visualize the firing patterns of the neuron across five trials in which each biomarker concentration was delivered to the locust antenna (FIG.12B). Interestingly, neural responses are shown for multiple concentrations including the lowest concentration of 0.0001% vol / vol for both biomarkers. Overall, FIG. 12B captures spike frequencies across multiple trials, highlighting the responsiveness of the antennal lobe neurons to different concentrations of these biomarkers. LDA, a dimensionality reduction technique, was then employed to visualize the high dimensional population neural responses in LDA subspace (FIG. 12C). Using this analysis, clear separation of clusters was seen for all biomarker concentrations and mineral oil with the higher concentrations of nonanal (0.1% vol / vol and 0.01% vol / vol) separating farther away from the other four clusters (FIG. 12C). Notably, the lowest concentrations for both nonanal and propylbenzene (0.0001% vol / vol) cluster near each other.

[0273] PCA was also used to visualize the population neural responses as trajectories in PCA subspace where we see the trajectories move in distinct directionalities, suggesting adequate differentiation of all biomarker concentrations. To test the classification accuracy of the system, LOTO analysis was performed to obtain a classification accuracy of 86% (FIG.12D). In this analysis, 20% of the 0.0001% nonanal concentration got misclassified as 0.0001% propylbenzene. Additionally, while there is some misclassification for 0.01% propylbenzene, the misclassifications happen for the other two propylbenzene concentrations. To summarize the results of the LOTO analysis, the sensitivity and specificity of each biomarker concentration were calculated (FIG. 12E).The time window used for LDA, PCA, and LOTO analysis was from 0.5 to 1.25 seconds after stimulus onset.TEC2025-0129 I TEC2025-0177 I TEC2026-00016550-000538-WO-POAEXAMPLE 13: The insect-based bioelectronic sensing platform can detect and classify human lung cancer cell lines via the emitted gas mixture or “scent” of the cultured cells

[0274] Lastly, the ability of the sensing platform to detect and classify gas mixtures emitted from cultures of human lung cancer cell lines was investigated. To explore this, four different human lung cancer cell lines and a healthy lung fibroblast (HLF) cell line were cultured. In brief, cell lines were cultured for four days before cells were split and resuspended in modified sterile T-25 flasks the day before the electrophysiological recording. Furthermore, a flask containing the same media used to resuspend all the cell lines was used as a control. Two of the lung cancer cell lines were non-small cell lung cancer (NSCLC; NCI-H1437 stage 1. NCI-H1573 stage 4) and the other two were small cell lung cancer (SCLC; SHP-77, H69PR). Cell flask caps were closed airtight for 1-2 hours before each experiment in order to capture the gas mixture emitted from the cell lines before delivery to the locust antenna. Representative images of all the cell lines are shown in FIG. 13A. Cell culture images were taken the day of the experiment and cell flasks were affixed to the olfactometer in a pseudorandom order. During the experiment, cell lines were kept in an incubator at 37°C and were only transferred to the olfactometer for gas mixture delivery before returning to the incubator.

[0275] Extracellular voltage responses were recorded for all cell lines from the locust antennal lobe. From the same voltage trace, a neuron was spike sorted and was visualized using raster and PSTH plots (FIG. 13B). Distinct responses for multiple cell lines and media are shown at the individual neuron level (FIG. 13B). A population PSTH for all neural responses shows the change in voltage after baseline subtraction for all cell lines and media with the time window of 0.25 to 2.5 seconds being chosen for subsequent population analysis. To visualize the neural responses at the population level, LDA was utilized as a dimensionality reduction technique (FIG. 13C). In this analysis, the majority of the cell lines can be seen to separate while only NSC-NCI-H1573 stage 4 and SC-SHP-77 seem to cluster closely. An additional population analysis was done using PCA plotting the neural response dynamics over time where clear separation between media and all cell lines was seen.

[0276] Additionally, the trajectories for the two SCLC cell lines move in similar but separate directions from the NSCLC cell lines and HLF. Using the quantitative LOTO trial-wise analysis, an accuracy of 85% was obtained (FIG. 13D). Notably, the lowest classification was seen for NSC-NCI-H1573 stage 4 cell line that gets misclassified for SC-SHP-77 and HLF. Furthermore, the sensitivity and specificity of each cell line were calculated and summarized in a table (FIG. 13E).TEC2025-0129 I TEC2025-0177 I TEC2026-00016550-000538-WO-POACollectively, these results showcase the ability of this bioelectronic sensing platform to distinguish and classify with a high accuracy multiple human lung cancer cell lines based on the complex “scent” of cell cultures that get encoded within the locust olfactory system.EXAMPLE 14: General Methods for Detecting Bacterial Biofilms

[0277] Electrophysiology

[0278] In vivo extracellular neural recordings were conducted on post-fifth instar, but pre-mating stage, locusts (Schistocerca americand) of either sex raised in a crowded colony.

[0279] Surgery: Surgery to access the antennal lobes (AL) for neural recordings followed previously published methods (Saha, D., Leong, K., Katta. N. & Raman, B. Multi-unit recording methods to characterize neural activity in the locust (Schistocerca americana) olfactory circuits. J Vis Exp, doi: 10.3791 / 50139 (2013)). First, locusts were immobilized by amputating the legs, sealing the amputation sites with a tissue adhesive (Vetbond®, 3M), and securing the locust’s thorax to a custom surgical bed using electrical tape. Antennae were immobilized by threading them through small polyethylene tubes (Intramedic® Microbore Tubing, 427420) within a poly- vinyl chloride 18 American Wire Gauge (AWG) wire insulation jacket that were then secured to batik wax pillars. A batik wax bowl was then built connecting the pillars to the head, starting above the mandibles and encompassing the compound eyes, with the rim extending just above the top of the head. Prior to making any incision, the wax bowl is filled with physiologically balanced locust saline solution at room temperature (Laurent, G. & Naraghi, M. Odorant- induced oscillations in the mushroom bodies of the locust. Journal of Neuroscience 14, 2993-3004 (1994)). The exoskeleton and glandular tissue covering the brain was removed. The gut, which the brain rests on, was also removed to prevent brain movement and then the brain was supported with a batik wax-coated wire platform connected to the batik wax bowl (FIG. 14B). Finally, the thin membrane sheath covering the AL was removed following a short treatment with protease.

[0280] Recording: A commercial Neuronexus® 16-channel silicon probe (A2x2-tet-3mm-150-150-121) was electroplated to achieve impedances between 200 to 300 k . The electrode was inserted about 100 pm into the AL. A silver chloride ground wire was placed within the head of the locust but not touching the brain to complete the circuit through the saline solution. While recording, a small saline drip (10 mL / hr) was used to prevent loss of saline solution in the wax bowl due to evaporation.TEC2025-0129 I TEC2025-0177 I TEC2026-00016550-000538-WO-POADuring data collection, voltages were sampled at 20 kHz and digitized by an Tntan® pre-amplifier board (C3334 RHD 16-channel headstage). These digital signals were transferred to the Intan® recording controller (C3100 RHD USB interface board) and then stored on a PC using the Intan® GUI. Six locusts were used for electrophysiology experiments.

[0281] Odor stimulation: Odors were delivered to the insect antennae via a commercial olfactometer (Aurora Scientific®, 220A) (FIG. 14C, FIG. 20A-C). During trials, air flow through the three mass flow controllers, MFC1, MFC2, and MFC3, were held constant at 120, 80, and 200 seem, respectively. Total flow to the insect antennae was also held constant by controlling the final valve to deliver the 200 seem from MFC3 or the combined 200 seem from MFC1 and 2. The constant air flow to the insect antennae mitigates potential confounding neural responses from mechanosensory neurons. At the start of a trial, the final valve was configured to deliver clean air from MFC3. Five seconds prior to odor delivery, the final valve continues to deliver air from MFC3 while valves 1 and 2 open, and valve 3 closes, to divert the 80 seem of air from MFC2 through the bacterial culture flask and to prime the airlines up to the final valve. During odor delivery, the final valve switches and delivers the combined airflow from MFC 1 and 2 to the insect antennae for four seconds. After four seconds, the final valve reverts to the original position, delivering clean air from MFC3, valves 1 and 2 close, and valve 3 opens to reduce VOC loss from the bacterial culture headspace. A 6” diameter funnel pulling a slight vacuum was always behind the locust to clear the VOCs after odor delivery. All airlines use PTFE to reduce VOC interactions. Clean air was provided by a compressed air cylinder at 12 PSI. During each experiment, a replicate of the same odor panel was used: Pseudomonas aeruginosa biofilm (PAbio), Pseudomonas aeruginosa planktonic (PAplank), Staphylococcus aureus biofilm (SAbio), Staphylococcus aureus planktonic (SAplank), and lysogeny broth (LB). The order of odor delivery was pseudorandomized, and each odor was delivered five consecutive times with a 1 -minute inter-stimulus interval.

[0282] Bacteria culturing: For the preparation of biofilms, overnight cultures of Xen5 (Pseudomonas aeruginosa) and Xen36 (Staphylococcus aureus) were grown in lysogeny (LB) broth. The cultures were then diluted 1:100 in fresh LB broth, and 5 mL was seeded into 25 cm2customized tissue culture flasks. Prior to bacteria seeding, the flasks were altered by inserting inlet and outlet 20-gauge needles through the cap and vertically through one of the upper corners opposite the cap. Both needles were then secured using a low-volatile, two-part epoxy (Permatex®, 84101), which also sealed the holes around the needles. The epoxy was allowed to cure for longer than 24 hours. At allTEC2025-0129 I TEC2025-0177 I TEC2026-00016550-000538-WO-POAtimes, except during odor delivery, the inlet and outlet needles were capped to prevent VOC loss. Because Pseudomonas forms biofilms at the air-liquid interface, it needs an edge upon which to attach, so the flasks were tilted at 30°. The flasks were then placed in a 37°C incubator overnight. To verify that biofilms had formed, bioluminescent images were obtained with the IVIS® instrument (Perkin-Elmer Inc.). For planktonic cultures, 400 mL of overnight culture in LB broth was added to 20 mL of LB broth and grown in a 250 mL Erlenmeyer flask at 37°C shaking at 200 RPM for 1.5 hours. Optical density was taken, and 107colony-forming units were added to 1 mL in a tissue culture flask with a needle port in the cap. Immediately before experiments were performed, the planktonic bacteria were removed from the biofilms, and 1 mL of LB broth was added to prevent drying. All flasks contained 1 mL of LB broth, including the LB broth control.

[0283] Spike sorting: Intan® neural data was imported into MATLAB® for high-pass filtering using a Butterworth filter to remove components below 300 Hz. Data were then spike sorted in Igor Pro® using previously described methods (Pouzat, C., Mazor, O. & Laurent, G. Using noise signature to optimize spike-sorting and to assess neuronal classification quality. J Neurosci Methods 122, 43-57, doi:10.1016 / s0165-0270(02)00276-5 (2002)). Each tetrode from the Neuronexus electrode was considered an independent recording and spike sorted. The detection threshold to identify spiking events was maintained between 2.5-3.5 standard deviations (SD) of baseline fluctuations. Single PNs were incorporated into the dataset if they passed three criteria: 1) cluster separation > 5 SD, 2) interspike intervals < 10%, and 3) spike waveform variance < 10%. A total of 32 PNs from six locusts passed all the criteria.

[0284] Hierarchical clustering: Spikes were binned into 50 ms non- overlapping time bins. A baseline spike rate for each neuron was calculated as the average spike rate during the two seconds preceding the odor stimulus and then subtracted from the dataset. Then the Euclidean distance in high-dimensional space (32 neurons yield 32 dimensions) between observations (5 trials x 5 odors = 25 observations) was calculated for each time point during the stimulus presentation (4s time window / 50 ms time bins = 80 time points). The distances across time were averaged for each pair of observations to yield 300 distance values (25 observations choose 2 = 300). Linkages were calculated using Ward’s method to produce a dendrogram (FIG. 17C). Hierarchical clustering was done using custom written MATLAB® (R2020b) code.

[0285] Dimensionality reduction analyses: Principal component analysis (PCA) was performed to reduce the dimensionality of the dataset for visualization purposes (Farnum, A. et al. HarnessingTEC2025-0129 I TEC2025-0177 I TEC2026-00016550-000538-WO-POAinsect olfactory neural circuits for detecting and discriminating human cancers. Biosensors and Bioelectronics 219, 114814, (2023); Pamas, M. et al. Precision detection of select human lung cancer biomarkers and cell lines using honeybee olfactory neural circuitry as a novel gas sensor. Biosensors and Bioelectronics, 116466, (2024)). Spikes were binned into 50ms non-overlapping time bins. A baseline spike rate for each neuron was calculated as the average spike rate during the two seconds preceding the odor stimulus and then subtracted from the dataset. The five trials were then averaged together. Each of the neurons is a separate dimension (FIG.21A-C), so PCA dimensionality reduction was performed on the combination of all five odors (LB, PAplank, PAbio, SAplank, and SAbio) to reduce from 32 dimensions to the three principal components (PCs) that explained the largest variance (PCI, 2, and 3). Each of the odors was then plotted along these three PCs with the data connected in temporal order to create the neural trajectories (FIG. 17A). The trajectories were smoothed using a third-order IIR Butterworth filter (Half Power Frequency = 0.15) and shifted to begin at the origin. PCA was done using custom written MATLAB® (R2020b) code.

[0286] Quantitative classification analysis: To assess the bacterial odor classification performance, a leave -one -trial-out (LOTO) cross-validation analysis was performed (Farnum, A. et al. Harnessing insect olfactory neural circuits for detecting and discriminating human cancers. Biosensors and Bioelectronics 219, 114814, (2023); Parnas, M. et al. Precision detection of select human lung cancer biomarkers and cell lines using honeybee olfactory neural circuitry as a novel gas sensor. Biosensors and Bioelectronics, 116466, (2024)). Spikes were binned into 50ms nonoverlapping time bins. A baseline spike rate for each neuron was calculated as the average spike rate during the two seconds preceding the odor stimulus and then subtracted from the dataset. Four out of the five trials were averaged together to yield high-dimensional, temporal, training neural data (32 neurons x 80 time bins) for each of the odors (FIG. 17B). The left-out trials from each odor were used as testing data. Therefore, for each of the 50ms time bins, there were five points in highdimensional space for training and another five points for testing; one training and one testing point for each odor. The testing points were compared with all of the training points and assigned to the odor that minimized the Euclidean distance (FIG. 14D). Then, we iterated through each of the trials, each time leaving a different one out for testing and using the average of the remaining four as training data. So, for a four-second analysis window, there are 80 time bins x 5 odors x 5 trials = 2000 time bins for which classification was attempted. Additionally, a winner-take-all approach was implemented by using the results of the bin-wise classification to vote. Here, the most commonTEC2025-0129 I TEC2025-0177 I TEC2026-00016550-000538-WO-POAclassification assignment, or the mode, of each trial was used to assign the entire testing trial to one of the training odors (FIG. 14E).

[0287] Also, completely separate training and testing datasets were used by averaging two random trials together to create the training templates and using the remaining three trials as three separate testing templates (FIG. 22A). The robustness of these results was assessed by conducting this analysis on all possible combinations of two training trials with the accuracy results summarized in a histogram (FIG. 22B).

[0288] To investigate the impact of different analysis time windows the prior winner-take-all approach was applied across different time windows of the dataset (FIG. 18A-B). Here, three different time window sizes, 0.25s (5 time bins), 0.5s (10 time bins), and Is (20 time bins), were tested across a continuous moving window. The averages of the trial-wise classification accuracies were plotted with the S.E.M. indicated by the shaded region (FIG. 18A). Each line indicates the accuracy at the center of a given analysis window, (e.g., the green line, representing a Is window, at t=0 is the classification accuracy of a time window from -0.5 to 0.5 seconds). The time windows that achieved the highest accuracy for each window size are indicated by the shaded rectangles, and the individual confusion matrices are reported in FIG. 18B.

[0289] ROC analysis: Sensitivity is calculated as the ratio of true positive (TP) classifications to actual positive conditions, or the combination of TP and false negatives (FN).TPSensitivity = -yTP + FN

[0290] Specificity is calculated as the ratio of true negative (TN) classifications to actual negative conditions, or the combination of TN with false positives (FP).TNSpecificity = -1 yTN + FP

[0291] For non-binary classifications in which there are more than two possible outcomes we used a one-vs-rest method to convert into a binary classification. In short, the odor of interest is the positive odor, and all other odors are the combined negative (FIG. 17D).

[0292] The threshold for classification was varied by weighting the Euclidean distances calculated between testing and training templates (FIG. 17A-B). Distances between any testing template with the training template of the odor of interest were multiplied by the threshold weightingTEC2025-0129 I TEC2025-0177 I TEC2026-00016550-000538-WO-POAvalue. Then, the testing templates were assigned to the closest training template using the weighted distance values. Therefore, at a threshold weighting value of 1, this analysis is the same as using minimum Euclidean distance to assign the testing odor templates. Threshold weighting values greater than one reduce the chance for testing templates to be assigned to the odor of interest, and vice-versa for threshold weighting values less than one. The sensitivity and specificity were calculated for a range of thresholds and plotted on a receiver operating characteristic (ROC) curve (FIG.19A-B). The sensitivity and specificity across the range of thresholds of 100 random subsets of 10, 20, and 30 neurons from the recorded population of 32 PNs were averaged together to produce the ROC curves. This analysis was replicated for subsets of neurons ranging from 5 to 31. The area under the curve (AUC) for each subset of neurons is calculated and averaged. An exponential model fits the data with the y-intercept at 0.5 because no neural data leads to a completely random binary classifier, or an AUC of 0.5 (FIG. 19C-D).

[0293] Youden’s Index was calculated as a combination of sensitivity and specificity to find the best threshold weighting to use (FIG. 19E-F).Youden’s Index — Sensitivity + Specificity — 1

[0294] Photoionization detector: A commercial photoionization detector (PID) (Aurora Scientific®, miniPID model 200B) with a 10.6 eV, RF-excited, lamp was used to check the VOC delivery profile of the olfactometer to the insect (FIG. 14C). The built-in pump on the PID was set to the lowest setting (750 seem) to pull air into the PID. After allowing the lamp to warm up for 30 minutes, the PID was calibrated to 0.1 pV using clean air with a gain of 5x. 80 seem of headspace from the bacterial odor flasks were diluted in 720 seem (10% dilution) for a total of 800 seem of air being delivered to the PID inlet via the olfactometer. Odors were delivered using the same timing (four seconds) of odor stimulation as the electrophysiology experiments for five trials each with 1-minute interstimulus intervals. Voltage outputs were collected at 100 Hz and saved as LVM files to a PC for subsequent analysis. Each trial was normalized by subtracting the baseline, calculated as the average voltage output during the first 30 seconds of each trial, from that trial. The odor was released 35-39 seconds after each trial started. The five trials were averaged together, and then a three-point moving smoothing filter was applied.EXAMPLE 15: Bacterial biofilm odors were classified using a biological olfactory brain.TEC2025-0129 I TEC2025-0177 I TEC2026-00016550-000538-WO-POA

[0295] The locust olfactory system was used to differentiate between biofilm and planktonic bacterial cultures from two different species. The bacteria were cultured in customized air-tight T25 flasks to produce five different odor flasks: P. aeruginosa biofilm (PAbio), P. aeruginosa planktonic (PAplank), S. aureus biofilm (SAbio), S. aureus planktonic (SAplank), and lysogeny broth (LB) as a control (FIG. 14A). Meanwhile, a locust was prepared for extracellular electrophysiology recordings from the antennal lobe (AL) (FIG. 14B). After all preparations were completed, the headspace from the odor flasks were presented to the locust antennae in a pseudorandom order using an olfactometer (FIG. 14C. FIG. 20A-C). A pseudorandom order was used to eliminate temporal biases, such as the possibility of a stronger response to the first odor presented than the last odor. The locust antenna, analogous to vertebrate noses, interacts with VOCs within the surrounding gaseous environment. Sensilla, tiny (about 10 pm long) hair-like follicles that cover the exterior of the antennae, have pores for VOCs to enter the antennal lymph (Ochieng, S. A., Hallberg, E. & Hansson, B. Fine structure and distribution of antennal sensilla of the desert locust, Schistocerca gregaria (Orthoptera: Acrididae). Cell and tissue research 291, 525-536 (1998)). There, VOCs can interact with odorant receptors (ORs) on the membranes of olfactory receptor neurons (ORNs) to initiate an electrical signal down the antennal nerve and into the AL. Individual ORNs express one of the over 100 ORs, but each OR can be sensitive to multiple VOCs (Pregitzer, P. et al. In search for pheromone receptors: certain members of the odorant receptor family in the desert locust Schistocerca gregaria (Orthoptera: Acrididae) are co-expressed with SNMP1. International Journal of Biological Sciences 13, 911 (2017)). Locusts, and most biological olfactory systems including vertebrates, combine the different response profiles of many ORs to enable a drastic increase in the number of unique odors they can detect. As an example, locusts using combinatorial coding with a binary, on / off, paradigm for each OR, could theoretically encode over 1.2 x IO30(2n- 1) unique odors from just 100 OR (Hallem, E. A. & Carlson, J. R. Coding of odors by a receptor repertoire. Cell 125, 143-160 (2006)). Once the neural signal reaches the AL, it is processed by a dense interconnected network of excitatory projection neurons (PNs) and inhibitory local neurons. The PNs then output to higher-order brain centers (e.g., Mushroom Body) for learning and decision making. Interestingly, there are over 100,000 ORNs within each antennae but only -800 PNs within an AL, connecting with over 300,000 neurons in the mushroom bodies (Laurent, G. Olfactory network dynamics and the coding of multidimensional signals. Nature reviews neuroscience 3, 884-895 (2002); Sachse. S. & Galizia, C. G. Topography and dynamics of the olfactory system (2006)). Due to this bottleneck, all the olfactory information is carried by only 800 PNs, which is why the PNs are targeted for data collection. Voltage traces, theTEC2025-0129 I TEC2025-0177 I TEC2026-00016550-000538-WO-POAraw data, were collected that contain responses from multiple PNs using an electrode carefully placed in the AL (FIG. 14B-C). The voltage traces were spike sorted to isolate the response pattern of each PN. The PN responses were then combined for analysis and high-dimensional “fingerprints” were created for each of the odors based on the population neural response (FIG. 14D). Using relatively simple pattern matching and Euclidean distance metrics the classification performance of the locust olfactory system was then quantified for the tested odors (FIG. 14E).EXAMPLE 16: Neural responses can differentiate between biofilm and planktonic cultures of Pseudomonas aeruginosa

[0296] The initial investigation focused on the responses of single PNs within the AL to bacterial odors. This was to determine whether the locust olfactory system could detect these bacteria. The raw voltage traces acquired from extracellular electrophysiology recordings are the amalgamation of neural responses from a small region of the brain, which can include multiple neurons or specifically, for the brain region we recorded from, multiple PNs (FIG. 15A). In this representative voltage trace from a single recording location, there are spikes caused by neuron action potentials. These spikes could be attributed to a single neuron or to many different neurons. However, the voltage spike shapes of individual neurons are unique, based on proximity to the electrode and neuron morphology (Buzsaki, G., Anastassiou, C. A. & Koch, C. The origin of extracellular fields and currents — EEG, ECoG, LFP and spikes. Nature reviews neuroscience 13, 407-420 (2012)), which can be leveraged to spike sort, e.g., assign each of the voltage spikes to individual neurons. Using a spike sorting algorithm (Pouzat, C., Mazor, O. & Laurent, G. Using noise signature to optimize spike-sorting and to assess neuronal classification quality. J Neurosci Methods 122, 43-57, doi: 10.1016 / s0165-0270(02)00276-5 (2002)), models can be created for each of the spike shapes within a voltage trace and apply these models to all of the voltage traces obtained during that recording. The output is the spike times assigned to individual PNs (32 PNs identified). To visualize a single PN’s responses, the spike times for each of the five trials of odor presentation were plotted as a raster, and the trials were averaged together into a peri-stimulus time histogram (PSTH) (FIG. 15B). For this specific neuron, notice that prior to the odor presentation (boxed area) there was very little spiking that then increased after odor presentation started and rapidly returned to the pre-stimulus baseline after odor presentation stopped. This PN had a slightly larger steady-state (~2-4 seconds after stimulus onset) and off (immediately after stimulus offset) response to the biofilm than the planktonic odor and both of theseTEC2025-0129 I TEC2025-0177 I TEC2026-00016550-000538-WO-POAresponses were much larger than the response to the LB control. These differences are important for classification analyses.

[0297] To examine the entire population of 32 identified PNs, a principal component analysis (PCA) was conducted to visualize the population neural trajectories (FIG. 15C). PCA, an unsupervised learning algorithm, was used to reduce 32-dimensional PN responses to the three principal components (PCs) that explain the largest variance in the dataset. PN spike times are binned into 50ms non-overlapping time bins. This matches the 20 Hz oscillations found within the locust AL, which is thought to allow for information integration prior to projection to higher-order brain centers (Laurent. G. Olfactory network dynamics and the coding of multidimensional signals. Nature reviews neuroscience 3, 884-895 (2002)). This 50ms binned data is used in all subsequent analyses. The time-binned data of all 32 PNs are transformed using PCA, and then the data points are connected in sequential order to show the temporal evolution of the neural response. To better understand the relationship between the temporal neural responses and the PCA neural trajectories, the responses from two neurons to two different odors were checked (FIG.21A). These spike timings were plotted with the neurons on separate axes (FIG. 21B). Using just two neurons, the relationship between temporal odor-evoked responses can be visualized without using dimensionality reduction. This visualization can then be extended out to all 32 neurons using PCA discussed above (FIG. 21C).

[0298] Trajectories that travel further from the origin (e.g., PAbio and PAplank) indicate larger neural responses with higher spiking rates and differences in the trajectory angles indicate different sets of neurons are responding (Stopfer, M., Jayaraman, V. & Laurent, G. Intensity versus identity coding in an olfactory system. Neuron 39, 991-1004, doi:10.1016 / j.neuron.2003.08.011 (2003)). Within most olfactory responses there is a transient phase associated with odor onset, which is the most discriminatory, followed by a steady state that is still different from the baseline but not as discriminatory as the transient phase. Here, the analysis on the transient response was focused during the first two seconds of odor stimulus. Clear differences in the trajectories for all the odors were observed, revealing that the population PN responses are unique for each bacterial odor. This PCA transformation explains over 50% of the variance in the dataset (obtained by summing PCI, 2, and 3) and therefore is used for qualitative purposes only.

[0299] A leave-one-trial-out (LOTO) analysis was used to quantitatively assess the classification accuracy of the PN responses using previously described methods (Farnum, A. et al. Harnessing insect olfactory neural circuits for detecting and discriminating human cancers. Biosensors andTEC2025-0129 I TEC2025-0177 I TEC2026-00016550-000538-WO-POABioelectronics 219, 114814, (2023); Pamas, M. et al. Precision detection of select human lung cancer biomarkers and cell lines using honeybee olfactory neural circuitry as a novel gas sensor. Biosensors and Bioelectronics, 116466, (2024)). As there are three odors, this analysis is a three-way test in which each testing template is compared to all three training templates and assigned using minimum Euclidean distance; completely random classification should yield 33% accuracy. Spikes were binned into non-overlapping 50ms time bins and then classified using time-matched training and testing templates. In a winner-take-all (WTA) approach, the mode of time bin classifications was used to classify each trial, and achieved 100% accuracy (FIG.15D). The WTA approach avoids the stochastic trial to trial variation within the 50ms bins that can be observed in the raster plots (FIG. 15B). Using neural information from only 32 PNs, discrimination between biofilm and planktonic cultures of P. aeruginosa was possible.EXAMPLE 17: Staphylococcus aureus biofilm and planktonic cultures were also differentiated

[0300] A key strength of biological olfaction is its sensitivity to a broad swath of odors. The locust sensor, with no genetic or chemical manipulation, not only responded to P. aeruginosa but also to S. aureus odors. There was an odor-evoked response at the single PN level to S. aureus biofilm and planktonic odors as seen in a representative neuron (FIG. 16A-B). At the population neural response level, SAbio and SAplank project in different directions within the neural response space indicating distinct responses (FIG. 16C). And, using a LOTO analysis, differentiation between the neural responses to these odors was possible (FIG. 16D). Only one of the five trials, 20%, of SAplank was misclassified to SAbio. The misclassification between LB and both S. aureus odors could be due to some of the odor-evoked response being from the control.EXAMPLE 18: Locust projection neurons can encode the odors of biofilm cultures from two different bacterial species, Pseudomonas aeruginosa and Staphylococcus aureus

[0301] The neural responses to both bacterial species were combined into a single analytical test to classify biofilms. PC A of the transient odor response, the first two seconds of odor stimulus, shows distinct trajectories for each of the five odors with the P. aeruginosa odors (red) projecting towards the top and 5. aureus odors (blue) projecting towards the bottom (FIG. 17A). Over 50% of the variance was explained in the first three PCs. A quantitative classification using LOTO achieves 88%TEC2025-0129 I TEC2025-0177 I TEC2026-00016550-000538-WO-POAaccuracy of the entire four-second odor stimulus (FIG. 17B). PAbio and SAbio were both classified with 100% accuracy. One trial (20%) of SAplank was misclassified as SAbio.

[0302] To further understand the neural encoding of bacterial odors, a hierarchical clustering analysis was performed (FIG. 17C). The time-matched 50ms bins were compared across the entire four second stimulus duration using Euclidean distance in 32 dimensions. The largest difference was seen between bacterial species, with a sub-splitting of the data between biofilm and planktonic cultures. These results suggest that the locust VOC sensor can more easily differentiate between bacterial species, however, it can still detect the more subtle differences between biofilm and planktonic cultures.

[0303] The sensitivity, calculated as the ratio of true positive test results (TP) over the presence of the odor (TP + FN), and specificity, the ratio of true negative test results (TN) over the absence of the odor (TN + FP), for each of the individual bacterial odors was high (FIG. 17D). PAbio, PAplank, and SAbio all achieved 1.0 sensitivity and both PAbio and PAplank had 1.0 specificity. The locust sensor reliably detected each of the bacterial odors.EXAMPLE 19: The spatiotemporal classification occurs rapidly and robustly after odor delivery

[0304] The classification ability was then tested using different subsets of the data to investigate the transient response period. The WTA analysis was applied to different time window sizes: 0.25s (5 time bins), 0.5s (10 time bins), and Is (20 time bins) (FIG. 18A). The classification accuracy was low during the pre-stimulus time period, indicating that there are no biases. However, after odor-onset during the transient response, the accuracy rapidly increased to 82-91% before slightly decreasing during the steady state response. These results are robust across different time windows and mirror the high overall accuracy of 88% achieved using the entire 4s window (FIG. 17B). The highest classification accuracies all center on ~ls after stimulus onset (FIG. 18B).EXAMPLE 20: Modeling the capabilities of the locust olfactory system for biofilm detection

[0305] In any classification model, thresholds between positive and negative classifications need to be properly and conscientiously chosen. Previously, the minimum Euclidean distance was used to classify odors; this effectively places the threshold exactly in the middle between any two odors (FIG.TEC2025-0129 I TEC2025-0177 I TEC2026-00016550-000538-WO-POA23A-B). Here, this threshold was varied by biasing the model towards or away from the PAbio (FIG.19A) and SAbio (FIG. 19B) odors. At each threshold, the sensitivity and specificity can be calculated to create a receiver operating characteristic (ROC) curve. Random subsets were also sampled (10, 20. or 30 PN) of the population of 32 recorded PNs and saw that the sensor performance trended upwards with the inclusion of more neurons for both biofilm odors. This trend was modeled as an exponential by comparing the area under the curve (AUC) with the number of neurons used in the model (FIG.19C-D). It was assumed that zero neurons would yield a completely random binary classifier and that as the number of neurons approaches infinity, the classification should approach perfection. The exponential regression models fit the data with R2values over 0.95. Finally, to determine the best classification thresholds to use for model generation, the Youden’s Index (YI) was calculated, a combination of sensitivity and specificity (FIG. 19E-F). The maximum YI, or best classification performance, for both biofilm odors occurs at a threshold very close to 1, which is the same as the minimum Euclidean distance. Therefore, using the minimum Euclidean distance in future odor classification problems would be a good approximation for the optimal classification threshold.

Claims

TEC2025-0129 I TEC2025-0177 I TEC2026-00016550-000538-WO-POAWHAT TS CLAIMED IS:

1. A method of detecting a presence or absence of a per- or polyfluoroalkyl substance (PFAS), the method comprising:exposing a biological chemosensory array to 1) at least one test gas-phase PFAS mixture and 2) at least one control gas-phase PFAS mixture, at least one control gas-phase non-PFAS mixture, or a combination thereof;obtaining at least one test neuronal response and at least one control neuronal response from the biological chemosensory array;comparing the at least one test neuronal response to the at least one control neuronal response; andoutputting a result based on the comparison, wherein the result indicates a presence or absence of a PFAS or a probability of a presence or absence of a PFAS.

2. The method of claim 1, wherein the test gas-phase PFAS mixture is emitted from one or more environmental sample.

3. The method of claim 2, wherein the environmental sample comprises air, water, soil, sediment, dust, a biosolid, or a biological sample.

4. The method of any one of the previous claims, wherein the biological chemosensory array is one or more insect antenna.

5. The method of claim 4, wherein the at least one test neuronal response and the at least one control neuronal response are obtained from one or more insect antennal lobes.

6. The method of claim 4 or claim 5, wherein the insect belongs to the order Orthoptera, Hymenoptera, or Diptera.

7. The method of claim 6, wherein the insect belonging to the order Orthoptera is a locust.TEC2025-0129 I TEC2025-0177 I TEC2026-00016550-000538-WO-POA8. The method of claim 6, wherein the insect belonging to the order Hymenoptera is a honeybee or an ant.

9. The method of any one of claims 4-8, wherein the insect has at least about 100 olfactory receptors.

10. The method of any one of the previous claims, wherein the at least one test neuronal response and the at least one control neuronal response are extracellular neuronal voltage signals.

11. The method of any one of the previous claims, wherein the at least one test neuronal response and the at least one control neuronal response are obtained from an insect in vivo.

12. The method of any one of claims 1-10, wherein the at least one test neuronal response and the at least one control neuronal response are obtained from an insect ex vivo.

13. The method of any one of the previous claims, further comprising determining the concentration or concentration range of the PFAS.

14. The method of any one of the previous claims, wherein the PFAS is detected at an environmental concentration, such as about 0.01% vol / vol to about 0.0001% vol / vol.

15. The method of any one of the previous claims, wherein the PFAS comprises a perfluoroalkyl acid (PFAA), perfluoroalkyl sulfonamide (FOSA), perfluoro alkyl sulfonamidoethanol (FOSE), fluorotelomer alcohol (FTOH), fluorotelomer-based PFAS, or polyfluoroalkyl precursor.

16. The method of any one of the previous claims, wherein the PFAS is a PFAA, and the PFAA is perfluorooctanesulfonic acid (PFOS).

17. The method of any one of claims 1-15, wherein the PFAS is a PFAA, and the PFAA is PFOA.

18. The method of any one of claims 1-15, wherein the PFAS is fluorotelomer alcohol (FTOH), fluorotelomer acrylate (FTAcr), methyl perfluorooctane sulfonamide (MeFOSA), perfluorooctane sulfonamide (PFOSA), ethyl perfluorooctane sulfonamidoethanol (EtFOSE), methylTEC2025-0129 I TEC2025-0177 I TEC2026-00016550-000538-WO-POAperfluorooctane sulfonamidoethanol (MeFOSE), perfluorooctane sulfonic acid (PFOS), perfluorooctanoic acid (PFOA), or a combination thereof.

19. A system for detecting the presence or absence of a per- or polyfluoroalkyl substance (PFAS), the system comprising:one or more biological chemosensory array; an odor stimulus delivery component for delivering 1) at least one test gas-phase PFAS mixture and 2) at least one control gas-phase PFAS mixture, at least one control gas-phase non-PFAS mixture, or a combination thereof to the one or more biological chemosensory array;wherein the one or more biological chemosensory array is stabilized by a stabilizing component;a neuron probe for detecting at least one test neuronal response and at least one control neuronal response from the one or more biological chemosensory array; andat least one processor which stores the at least one test neuronal response and the at least one control neuronal response in memory, wherein the at least one test neuronal response and the at least one control neuronal response are one or more neuronal voltage signals.

20. The system of claim 19, wherein the test gas-phase PFAS mixture is emitted from one or more environmental sample.

21. The system of claim 20, wherein the environmental sample comprises air, water, soil, sediment, dust, a biosolid, or a biological sample.

22. The system of any one of claims 19-21, wherein the biological chemosensory array is one or more insect antenna.

23. The system of claim 22, wherein the at least one test neuronal response and the at least one control neuronal response are obtained from one or more insect antennal lobes.

24. The system of claim 22 or claim 23. wherein the insect belongs to the order Orthoptera, Hymenoptera, or Diptera.TEC2025-0129 I TEC2025-0177 I TEC2026-00016550-000538-WO-POA25. The system of claim 24, wherein the insect belonging to the order Orthoptera is a locust.

26. The system of claim 24, wherein the insect belonging to the order Hymenoptera is a honeybee or an ant.

27. The system of any one of claims 22-26, wherein the insect has at least about 100 olfactory receptors.

28. The system of any one of claims 19-27, wherein the at least one test neuronal response and the at least one control neuronal response are obtained from an insect in vivo.

29. The method of any one of claims 19-27, wherein the at least one test neuronal response and the at least one control neuronal response are obtained from an insect ex vivo.

30. The system of any one of claims 19-29, wherein the PFAS comprises a perfluoro alkyl acid (PFAA), perfluoroalkyl sulfonamide (FOSA), perfluoroalkyl sulfonamidoethanol (FOSE), fluorotelomer alcohol (FTOH), fluorotelomer-based PFAS, or polyfluoroalkyl precursor.

31. The system of any one of claims 19-30, wherein the PFAS is a PFAA, and the PFAA is perfluorooctanesulfonic acid (PFOS).

32. The system of any one of claims 19-30, wherein the PFAS is a PFAA, and the PFAA is PFOA.

33. The system of any one of claims 19-30, wherein the PFAS is fluorotelomer alcohol (FTOH), fluorotelomer acrylate (FTAcr), methyl perfluorooctane sulfonamide (MeFOSA), perfluorooctane sulfonamide (PFOSA), ethyl perfluorooctane sulfonamidoethanol (EtFOSE), methyl perfluorooctane sulfonamidoethanol (MeFOSE), perfluorooctane sulfonic acid (PFOS), perfluorooctanoic acid (PFOA), or a combination thereof.

34. The system of any one of claims 19-33, for use in the detection of PFAS in an environmental sample.TEC2025-0129 I TEC2025-0177 I TEC2026-00016550-000538-WO-POA35. A method of detecting a presence or absence of lung cancer, the method comprising:exposing a biological chemosensory array to 1) at least one test volatile organic compound (VOC) mixture and 2) at least one control VOC mixture, at least one control non-VOC mixture, or a combination thereof;obtaining at least one test neuronal response and at least one control neuronal response from the biological chemosensory array;comparing the at least one test neuronal response to the at least one control neuronal response; andoutputting a result based on the comparison, wherein the result indicates a presence or absence of lung cancer or a probability of a presence or absence of lung cancer.

36. The method of claim 35, wherein the at least one test VOC mixture and the at least one control VOC mixture are VOC gas mixtures.

37. The method of claim 35 or claim 36, wherein the at least one test VOC mixture is emitted from one or more biological sample.

38. The method of claim 37, wherein the biological sample comprises breath, urine, sweat, or blood.

39. The method of claim 37 or claim 38, wherein the biological sample is obtained from a mouse, primate, or human.

40. The method of any one of claims 35-39, wherein the biological chemosensory array is one or more insect antenna.

41. The method of claim 40, wherein the at least one test neuronal response and the at least one control neuronal response are obtained from one or more insect antennal lobes.

42. The method of claim 40 or claim 41, wherein the insect belongs to the order Orthoptera, Hymenoptera, or Diptera.TEC2025-0129 I TEC2025-0177 I TEC2026-00016550-000538-WO-POA43. The method of claim 42, wherein the insect belonging to the order Orthoptera is a locust.

44. The method of claim 42, wherein the insect belonging to the order Hymenoptera is a honeybee or an ant.

45. The method of any one of claims 40-44, wherein the insect has at least about 100 olfactory receptors.

46. The method of any one of claims 35-45, wherein the at least one test neuronal response and the at least one control neuronal response are extracellular neuronal voltage signals.

47. The method of any one of claims 35-46, wherein the at least one test neuronal response and the at least one control neuronal response are obtained from an insect in vivo.

48. The method of any one of claims 35-46, wherein the at least one test neuronal response and the at least one control neuronal response are obtained from an insect ex vivo.

49. The method of any one of claims 35-48, further comprising classifying the type of lung cancer.

50. The method of any one of claims 35-49, wherein the lung cancer is non-small cell lung cancer (NSCLC) or small cell lung cancer (SCLC).

51. A system for detecting the presence or absence of lung cancer, the system comprising:one or more biological chemosensory array;an odor stimulus delivery component for delivering 1) at least one test volatile organic compound (VOC) mixture and 2) at least one control VOC mixture, at least one control non-VOC mixture, or a combination thereof to the one or more biological chemosensory array;wherein the one or more biological chemosensory array is stabilized by a stabilizing component;a neuron probe for detecting at least one test neuronal response and at least one control neuronal response from the one or more biological chemosensory array; andTEC2025-0129 I TEC2025-0177 I TEC2026-00016550-000538-WO-POAat least one processor which stores the at least one test neuronal response and the at least one control neuronal response in memory, wherein the at least one test neuronal response and the at least one control neuronal response are one or more neuronal voltage signals.

52. The system of claim 51, wherein the at least one test VOC mixture and the at least one control VOC mixture are VOC gas mixtures.

53. The system of claim 51 or claim 52, wherein the at least one test VOC mixture is emitted from one or more biological sample.

54. The system of claim 53. wherein the biological sample comprises breath, urine, sweat, or blood.

55. The system of claim 53 or claim 54, wherein the biological sample is obtained from a mouse, primate, or human.

56. The system of any one of claims 51-55, wherein the biological chemosensory array is one or more insect antenna.

57. The system of claim 56, wherein the at least one test neuronal response and the at least one control neuronal response are obtained from one or more insect antennal lobes.

58. The system of claim 56 or claim 57. wherein the insect belongs to the order Orthoptera, Hymenoptera, or Diptera.

59. The system of claim 58. wherein the insect belonging to the order Orthoptera is a locust.

60. The system of claim 58, wherein the insect belonging to the order Hymenoptera is a honeybee or an ant.

61. The system of any one of claims 56-60, wherein the insect has at least about 100 olfactory receptors.TEC2025-0129 I TEC2025-0177 I TEC2026-00016550-000538-WO-POA62. The system of any one of claims 51 -61 , wherein the at least one test neuronal response and the at least one control neuronal response are obtained from an insect in vivo.

63. The system of any one of claims 51-62, wherein the at least one test neuronal response and the at least one control neuronal response are obtained from an insect ex vivo.

64. The system of any one of claims 51-63, for use in diagnosis, prevention, and / or treatment of lung cancer.

65. The system of any one of claims 51-64, wherein the lung cancer is non-small cell lung cancer (NSCLC) or small cell lung cancer (SCLC).

Images