(T1430-06-37) Quantum Electrochemical Spectroscopy (QES): A Novel AI-Driven Spectroscopic Approach to Simplify Bioanalytical Development in the Pharmaceutical and Biotech Industries

Purpose: Analytical assay development in the pharmaceutical industry is often hindered by complexity and cost due to the sequential and iterative approaches as well as the complexity of the technologies used. For example, using Immunoassays require the extensive development and validation of specific reagents, and using LC-MS or other spectroscopic techniques need sample preparation methods specific for each sample type assessed. In this study, we introduce Quantum Electrochemical Spectroscopy (QES), a novel technique that simplifies analytical testing and assay development: QES requires minimal sample volume (2-10µl), one pipetting step, and no reagents or sample preparation. QES employs a bench-top instrument to measure molecular vibrations, generating a digital fingerprint for each sample. This mathematical fingerprint allows the detection, quantification, and classification of biomarkers in a highly multiplexed and efficient manner, either a-priori or a-posteriori. In previous studies, QES has demonstrated sensitivity and specificity in differentiating mass isotopes and structural isomers¹. In this work, we showcase the differentiation and quantitation of protein isoforms: long-acting (Toujeo/ glargine) and short-acting (Humalog/ lispro) insulin in a mixture of the two molecules. We also extended this AI-driven analytical approach to detect and quantify low-abundance inflammatory proteins (e.g. IL-6 and TNF-alpha) in blood-derived samples. By reducing complexity and streamlining analytical development, QES can significantly reduce associated costs and increase the efficiency and transferability of the assays developed.

Methods: Data Acquisition: 4 ul of rat serum or 2ul of analyte standard solution in PBS were pipetted into a consumable electrochemical sensor coupled to a Probius™ QES instrument. The QES instrument performs a 30 min voltage scan of the sample to create a high dimensional vibrational signature of the molecular species therein. All samples and aliquots were assessed in triplicates. The high-dimensional dataset is uploaded to the software (Probius BCS) as a digital representation of the sample, a digital twin. Then, using reference standards (see below), the high-dimensional vibrational spectrum of the samples is deconvoluted to individual molecule contributions in the bulk signal and parsed for inferencing using previously trained models. Analyte calibration: Insulin glargine and insulin lispro (Sigma-Aldrich, USA) were solubilized in phosphate buffer saline, and serially diluted to produce standards from 25ag/ml to 125ag/ml (final concentration in the QES vial), with the two forms present in different ratios (0:1, 1:1, 1:2, 1:3). For TNF-alpha and IL-6 (Biotechne, USA) Standard solutions were prepared in PBS and Commercially available Serum ranging from 0-80 pg/mLand 0-5,000 pg/mL respectively. All stock solutions were analyzed by QES as described above. Electronic vibrational signal intensity was plotted vs standard concentration to create a calibration curve. Freshly prepared blinded standard aliquots were used to verify the method. (Figure 1)

Results: in this study, we demonstrate the ability of QES to identify, differentiate and quantify 2 protein isoforms where the two insulin-derived peptides differ in composition by three amino acids at positions B28, B29 and B30 of the B-chain. Both molecules can be differentiated to the ag/mL level, demonstrating the very high sensitivity of the QES approach. (Figure 2)
The quantitation-based machine learning model used to differentiate between the two insulin isoforms is extended to assay for the cytokines IL-6 and TNF-alpha in blood-derived samples, thereby not only demonstrating the ability to differentiate and quantitate peptides in a clean buffer-based matrix but also larger protein molecules in more complex matrices without the need for sample preparation or analyte-specific reagents.

Conclusion: Developing new assays in the bioanalytical lab for pharmaceutical development is an intensive and expensive process. The cost is commonly driven by the extensive validation required by reagent-centric methods (e.g. Immunoassays) or the extensive sample preparation techniques (e.g. LC-MS).
QES does not require reagents or sample preparation and only requires one pipetting step, dramatically reducing the complexity and the cost of method creation and validation. Despite being a simple and straightforward analytical method, QES does not compromise the sensitivity or specificity of the analyte detection, demonstrated in this work by the very low amount of insulin molecules identified and quantified (ag/mL) in a simpler background matrix. We also demonstrate the ability of QES to quantify a bigger protein (IL-6 and TNF-a) in a more complex matrix (blood-derived).

References: 1) Gupta, Chaitanya, et al. "Quantum tunneling currents in a nanoengineered electrochemical system." The Journal of Physical Chemistry C 121.28 (2017): 15085-15105.

Acknowledgements: All Authors of this abstract are either employees and/or shareholders of ProbiusDx Inc.

Figure 1: calibration curve generation method

Figure 2: Calibration Curves and UMap representation for Insulin Isoforms

Figure 3: calibration curves for TNF-alpha and IL-6 proteins and representation of electronic vibrational spectrum for the 2 different proteins and different concentrations (IL-6)