Multiplex injection in liquid chromatography-mass spectrometry for analyte detection

Abstract

Disclosed are methods, and apparatuses containing a device, for analytical data acquisition, involving the application of Hadamard transform multiplexing to online (i.e., realtime) size exclusion chromatography (SEC) operably linked to a charge detection mass spectrometer. The apparatuses contain a device—microcontroller—specifically configured to deliver pulsed injections from a large sample loop onto an SEC for online charge detection-mass spectrometry (CD-MS) analysis. Data showed a series of peaks spaced according to the pseudorandom injection sequence, which were demultiplexed with a Hadamard transform algorithm. The demultiplexed data revealed improved CD-MS signals while preserving the retention time information. This multiplexing approach provides a general solution to the inherent incompatibilities of online separations and CD-MS detection, and can be used in a range of applications.

Description

[0001] MULTIPLEX INJECTION IN LIQUID CHROMATOGRAPHY-MASS SPECTROMETRY FOR ANALYTE DETECTION

[0002] CROSS REFERENCE TO RELATED APPLICATIONS

[0003] This application claims the benefit of and priority to U.S. Provisional Application No. 63 / 609,673, filed December 13, 2023, which is hereby incorporated by reference in its entirety.

[0004] STATEMENT REGARDING FEDERALLY SPONSORED RESEARCH

[0005] This invention was made with government support under Grant No. 1845230 awarded by NSF. The government has certain rights in the invention.

[0006] FIELD OF THE INVENTION

[0007] The field of the invention is generally related to methods, and apparatuses containing a device, for acquisition of analytical data, particularly multiplex injection in liquid chromatography (e.g., size exclusion chromatography) operably linked to charge detection mass spectrometry, and demultiplexing data from a mass spectrometer to detect analytes.

[0008] BACKGROUND OF THE INVENTION

[0009] Analysis of very large, heterogeneous biomolecules by native mass spectrometry (MS) is challenging when the charge states of ions cannot be determined. Charge detection-mass spectrometry (CD-MS) has been utilized in these settings by directly determining the charge states of individual ions (Fuerstenau, Rapid Commun. Mass Speclrom. 1995, 9 (15), 1528-1538; Pierson, et al. J. Am. Soc. Mass Spectrom. 2015, 26 (7), 1213-1220; Womer. T, et al. J. R. Nat. Methods 2020, 17 (4), 395-398; Kafader, J. O et al., Anal. Chem. 2019, 91 (4), 2776- 2783). CD-MS has found applications in the characterization of very large (up to tens of mega-Daltons (MDa)) biomolecules such as protein complexes (Womer, et al. J. R. Nat. Methods 2020, 17 (4), 395-398), intact virus capsids (Miller, et al., Anal. Chem. 2021, 93 (35), 11965-11972), vaccines (Miller, et al.. Anal. Chem. 2021, 93 (35), 11965-11972), and gene therapies (Kostelic, M. M et al., Anal. Chem. 2022, 94 (34), 11723-11727; Barnes, L. F et al., Mol. Ther.-Methods Clin. Dev. 2021, 23, 87-97). Growing needs in biotechnology and the introduction of commercial instruments have led to increased interest in analyzing large, heterogeneous biomolecules.

[0010] CD-MS is limited in that only a small number of ions can be measured at a time, which necessitates the acquisition of many hundreds or thousands of spectra to provide enough detection events for accurate mass measurements with acceptable signal-to-noise ratios (SNR). These long acquisitions generally prevent the coupling of CD-MS with online (i.e., real-time) separation techniques, such as size exclusion chromatography (SEC), which typically only provide enough time for 50-200 scans during the elution of a chromatographic peak. Progress has been made on this front by refining CD-MS instrumentation to enable the measurement of multiple ions simultaneously (Kafader, J. O et al., Anal. Chem. 2019, 91 (4), TI6- 2783; Harper, C. C et al., Anal. Chem. 2019, 91 (11), 7458-7465) and improve charge state assignments with deconvolution algorithms (Kostelic, M. M. et al., Anal. Chem. 2021, 93 (44), 14722-14729). However, a recent effort to evaluate the feasibility of online SEC-CD-MS produced deconvolved mass spectra that, while sufficient to measure accurate protein masses, had lower SNRs and contained significantly more erroneous peaks than spectra produced from infusion experiments (Strasser, L et al., J. Anal. Chem. 2023, 95 (40), 15118-15124). Accordingly, there remains a need to develop methods, and apparatuses containing a device, for improved acquisition of analytical data for very large, heterogeneous biomolecules and / or microorganisms.

[0011] It is therefore an object of the present invention to provide methods, and apparatuses containing a device, for improved acquisition of analytical data.

[0012] It is also an object of the present invention to provide methods, and apparatuses containing a device, involving multiplex injection in liquid chromatography (e.g., size exclusion chromatography) operably linked to charge detection mass spectrometry, and demultiplexing data from a mass spectrometer to detect analytes, wherein eluted samples from a liquid chromatograph are introduced into a mass spectrometer.

[0013] SUMMARY OF THE INVENTION

[0014] Described are methods, and apparatuses containing a device, for the acquisition of multiplexed analytical data for non-gas phase chromatographs operably linked to a mass spectrometer. The methods involve delivering two or more pulsed injections of a non-gas phase sample into a chromatograph. The analyte selectivity of the chromatograph is not based on gas phase ion mobility separation or gas chromatography. The device (e.g., a microcontroller) operates an injection valve to inject a sample directly or indirectly into a chromatograph. Preferably, the device is operably linked to a valve (e.g., an injection valve), such that it controls the valve (e.g., an injection valve), and an algorithm programs the device with a pseudo-random binary sequence to control injections as well as the duration of each injection (injection time) and delay time between injections (cycle time). The length of the pseudo-random binary sequence (PRBS) can be defined by 2n- 1, where n can be any positive integer > 2, such as 3 or 5. After flowing through the chromatograph, analytes are separated in the sample, and the eluted analyte is channeled into the mass spectrometer. The mass spectrometer generates convoluted time distribution data in real-time, which are demultiplexed by a processing algorithm, such as a Hadamard Transform. Reassembling the multiple demultiplexed datasets produces a second dataset containing mass, charge, and retention time of the analyte, preferably wherein the reassembling is performed using Hadamard Transform. In general the reassembling process involves first deconvolving the data from m / z and charge into mass and charge, such as by using UniDecCD (Kostelic, M. M. et al., Anal. Chem. 2021, 93 (44), 14722-14729), creating EICs at each mass and charge value, and then reassembling the demultiplexed data into an array of mass vs. charge vs. retention time.

[0015] It should be noted that EICs can also be created from m / z and charge selections, not just mass and charge selections. In other words, deconvolution is not a pre-requisite to Hadamard Transform, as Hadamard Transform can also be performed on extracted raw data prior to deconvolution. In the non-limiting case of charge detection mass spectrometry, files containing charge detection mass spectrometry data are initially processed, analyzed, and optionally visualized using a software package such as STORIboard (distributed by Proteinaceous). The data is further processed using a custom Hadamard Transform algorithm described here. Extracted ion chromatograms are generated by defining an m / z range and charge state range for each ion population of interest. Further, a kernel array is constructed based on the pseudorandom binary sequence (including zero pads when applicable) by substituting - Is for Os in the sequence and infilling with Os so that the array was the same size as the number of scans in the full Hadamard Transform cycle. The kernel array is next rotated to move the zero-padded section to the beginning of the sequence and smoothed by convolution with a Gaussian function with a width of 5 bins. The modified kernel is next convolved with each EIC, and the resulting demultiplexed EICs shifted back to the original time index.

[0016] The data described below show that using the disclosed methods and device, results in greatly improved signal-to-noise ratio in both the chromatographic and mass dimensions while preserving retention time information. The methods, and apparatuses containing a device, described herein are applicable in any chromatography system that uses isocratic gradients and that is capable of multiple injections, making it a generalized and versatile technique for generating high-quality data.

[0017] BRIEF DESCRIPTION OF THE DRAWINGS

[0018] FIG. 1 is a schematic of a Hadamard Transform-SEC-CD-MS workflow. The sample is loaded in a loop. A microcontroller operates the injection valve to inject sample onto the column with a defined pseudo-random binary sequence. After flowing through the column, the eluted sample is analyzed in real-time by CD-MS to produce a series of CD-MS scans. Extracting the intensity over time for specific regions of the scans produces an extracted ion chromatogram, which can be HT demultiplexed. Reassembling multiple demultiplexed chromatograms produces a dataset of mass vs. charge vs. retention time. In a non-limiting example, 200 mM ammonium acetate is supplied by the LC pump at 200 pL / min. The sample is loaded into a 100 or 500 pL sample loop connected to a 6-port valve. The 6-port valve is shown in the sample loading position. Counterclockwise from the pump inlet is the outlet to the column, sample loop entrance, injection port, waste outlet, and sample loop exit. The valve is controlled by a microcontroller and switches the 6-port valve between inject and load positions according to the PRBS. CD-MS spectra are acquired by a Thermo Q- Exactive HF UHMR Orbitrap mass spectrometer.

[0019] FIGs. 2A-2E show comparison of P-galactosidase SEC-CD-MS chromatograms generated by 1 -bit, 3-bit, and 5-bit injection sequences before (FIG. 2A) and after (FIG. 2B) HT demultiplexing. Data from three replicates are shown for each series to demonstrate reproducibility. The raw m / z vs charge distributions (FIGs. 2C-2E) produced by 1 -bit (FIG. 2C), 3-bit (FIG. 2D), and 5-bit (FIG. 2E) injection sequences for P-galactosidase. Ion counts from this data are reported in Table 1.

[0020] FIGs. 3A-3D show Hadamard transform SEC-CD-MS analysis of a mixture of P- galactosidase and GroEL using the 5-bit injection sequence. FIG. 3A is a line graph showing Raw TIC with overlaid EICs, colored, respectively. FIG. 3B is a line graph showing demultiplexed chromatogram with inset showing the two overlapping EIC peaks. FIG. 3C shows m / z vs charge heatmap after deconvolution with rectangles showing the m / z and charge ranges selected for the EICs of GroEL (upper-right rectangle) and P-galactosidase (lower-left rectangle), respectively. FIG. 3D is a deconvolved mass spectra showing both protein mass distributions, with extracted distributions: P-galactosidase, left peak; GroEL, right peak.

[0021] FIGs 4A-4D show high-mass fraction of E. coli cell lysate. FIG. 4A shows Raw TIC (top-most line) and EICs of the five most abundant observed proteins. FIG. 4B shows demultiplexed TIC and EICs from each of the chromatograms shown in FIG. 4A. FIG. 4C shows m / z vs charge heatmap after deconvolution with areas selected for EIC extraction shown in shaded polygons. FIG. 4D shows deconvolved mass spectra with colored traces indicating mass distributions from the selected regions from FIG. 4C.

[0022] FIG. 5 is a representative UniChromCD GUI (Graphical user interface) with example data displayed. The raw chromatogram and EICs are displayed in the top left plot. Demultiplexed chromatograms / arrival time distributions are displayed in the top right plot. A 2D m / z vs charge histogram of CD-MS data is displayed in the bottom left plot, and the mass distribution in the bottom right. Additional plots are not shown but can be accessed by scrolling down. The newly added green control panels include demultiplexing controls are shown.

[0023] FIG. 6 is a flowchart showing data flow and interactions between different exemplary programming classes of UniChromCD. Files are opened using the existing UniDecCD code, but new elements in UniChromCD and the demultiplexing engine (DM) set up the required timedomain information and prepare the demultiplexing kernels. After the data is processed in UniDecCD, UniChromCD parses the processed data into a histogram stack and creates the TIC. EICs can be extracted from the summed histogram, which can then be processed with UniDecCD to yield extracted mass distributions. The entire histogram stack can also be transformed to convert the m / z axis to mass. At any point, the TIC, EIC, m / z histogram stack, or mass histogram stack can be demultiplexed. The dashed arrow is only dashed to make it easier to follow.

[0024] FIGs. 7A-7K show an exemplary HT-SEC-CD-MS workflow. Raw TIC (FIG. 7A), 2D m / z vs charge histogram (FIG. 7B), and mass distribution plots (FIG. 7C) are generated when the data is opened. Scan compression of FIG. 7A produces a smoother chromatogram (FIG. 7D). Binning the m / z values in FIG. 7E smooths the histogram (FIG. 7E). Binning the mass values in FIG. 7C smooths the mass distribution (FIG. 7F). Selecting a region of the histogram (FIG. 7H)generates EICs (FIG. 7G)and the extracted mass distributions (FIG. 71) for each selected species. A split arrow is used to show that (FIG. 7G, 7I)proceed from the selections in (FIG. 7H). If the data is deconvolved by UniDecCD, the extracted masses will also be deconvolved (FIG. 7L). TICs and EICs can be demultiplexed (FIG. 7 J), which can then be further improved by masked demultiplexing (FIG. 7K). The inset in FIG. 7K shows a zoomed region around the major peaks to demonstrate the separation, with GroEL eluting around 8.9 min and P-galactosidase around 9.7 min.

[0025] DETAILED DESCRIPTION OF THE INVENTION

[0026] I. Definitions

[0027] “Non-gas phase,” as relates to a sample containing an analyte for detections, means that the sample is not in the gas phase.

[0028] “Gas phase ion mobility separation,” as relates to principles for separating analytes in a sample, means that the sample is in the gas phase and separation of ions occurs by collisions with one or more ionic states of the analytes. The gas can be helium, neon, argon, krypton, xenon, nitrogen, oxygen, methane, carbon dioxide, water, methanol, methyl fluoride, ammonia, deuterated analogs thereof, tritiated analogs thereof, and any combination thereof, and the collisions can be reactive collisions, non-reactive collisions, or a combination thereof. IL Methods, apparatuses, and non-transitory media for the acquisition of multiplexed analytical data a. Methods for the acquisition of multiplexed analytical data

[0029] Described are methods and apparatuses for the acquisition of multiplexed analytical data for non-gas phase chromatographs operably linked to a mass spectrometer. The methods involve delivering two or more pulsed injections of a non-gas phase sample into a chromatograph. Preferably, analyte selectivity of the chromatograph is not based on gas phase ion mobility separation. A device (e.g. , a microcontroller) operates an injection valve to inject a sample directly or indirectly into a chromatograph. An algorithm specifies a pseudo-random binary sequence (PRBS) of injections and controls the injections as well as the duration of each injection (injection time) and delay time between injections (cycle time). A user provides the algorithm with a PRBS, the injection time, and the cycle time. At the beginning of the algorithm, the mass spectrometry data acquisition is started. It then evaluates the first digit in the PRBS. If a 1 is present in the PRBS, the algorithm opens the valve (for example by switching to the “inject” position) for a length of time equal to the injection time. After the injection time is completed, the valve is closed (for example by switching to the “load” position), and the algorithm waits for a length of time equal to the cycle time minus the injection time. At the end of the cycle time, it evaluates the next digit in the PRBS. If a 0 is present in the PRBS, the algorithm simply waits for the full cycle time, leaving the injection valve closed. At the end of the cycle time, it evaluates the next digit in the PRBS. The length of the PRBS can be defined by 2n- 1 , where n is any positive integer > 2, The length of the pseudo-random binary sequence (PRBS) can be defined by 2n- 1, where n can be any positive integer > 2, such as between 2 and 100, inclusive, between 2 and 90, inclusive, between 2 and 80, inclusive, between 2 and 70, inclusive, between 2 and 60, inclusive, between 2 and 50, inclusive, between 2 and 40, inclusive, between 2 and 30, inclusive, between 2 and 20, inclusive, or between 2 and 10, inclusive, such as, 3 or 5. When necessary, additional zeros can be added to the end of the PRBS to ensure that all chromatographic peaks are observed prior to the end of the injection sequence. The sample loop in the injection valve must be sufficiently large to accommodate enough sample for all injections in a single sequence, based on product of the flow rate, sequence length, and injection time.

[0030] After flowing through the chromatograph, preferably, analytes are separated in the sample, and the eluted analyte is channeled into the mass spectrometer. The mass spectrometer generates convoluted time distribution data in real-time, which are demultiplexed by a processing algorithm. Real-time in this instance can be a time frame, such as within a millisecond, within a second, within 1 min, 2 mins, 3 mins, 4 mins, 5 mins, 10 mins, 15 mins, 20 mins, or 30 mins after receiving the eluted analyte. The demultiplexing can be performed via a Hadamard Transform or any other correlation chromatography techniques that involve a multiplexed binary injection sequence. Reassembling the multiple demultiplexed datasets produces a second dataset containing mass, charge, and retention time of the analyte. Preferably the reassembling is performed using Hadamard Transform. In the non-limiting case of charge detection mass spectrometry, files containing charge detection mass spectrometry data can be initially processed, analyzed, and optionally visualized using a software package such as STORIboard (distributed by Proteinaceous). This software converts the raw mass spectrometry transient data (e.g., the Orbitrap transient data) into charge states, returning a list of ion features with scan number, m / z, and charge for each feature. The data can then be further processed using a custom Hadamard Transform algorithm described here. Extracted ion chromatograms (EICs) can be generated by defining an m / z range and charge state range for each ion population of interest. Here, extracted ion chromatograms (EICs) can be generated for a subset of the data. This can be done for large ranges, where each EIC includes all charge states for an analyte or multiple analytes. EICs can also be constructed from very small ranges, such as a single data point. EICs can be created by selecting a specific m / z and / or charge region. They can also be created after deconvolution of m / z into mass to select a specific mass and / or charge region. Performing demultiplexing on small mass and charge regions provides for the creation of 3D data sets with mass, charge, and retention time as axes. Further, a kernel array can be constructed based on the PRBS (including zero pads when applicable) by substituting -Is for Os in the sequence and infilling with Os so that the array was the same size as the number of scans in the full Hadamard Transform cycle. The kernel array can then be rotated to move the zero- padded section to the beginning of the sequence and smoothed by convolution with suitable functions, such as a Gaussian function, having appropriate widths, such as a width of 5 bins. The modified kernel can then be convolved with each EIC, and the resulting demultiplexed EICs shifted back to the original time index.

[0031] By using this multiplexed injection and demultiplexing of mass spectrometer-generated data, the signal-to-noise ratio is greatly improved in both chromatographic and mass dimensions while preserving retention time information. Furthermore, the disclosed methods and apparatuses can be applied to any liquid chromatography system that uses isocratic gradients and that has been configured to perform two or more pulsed injections of samples. Accordingly, the disclosed methods and apparatuses are generalizable and versatile techniques for generating high-quality liquid chromatograph- mass spectrometer data (e.g., liquid chromatograph-charge detection mass spectrometer data).

[0032] The chromatograph used in the step for separating analytes is selected from a liquid chromatograph (such as size exclusion chromatograph, high performance liquid chromatograph, and reverse-phase liquid chromatograph), a column chromatograph, an ion-exchange chromatograph, a gel-permeation (molecular sieve) chromatograph, an affinity chromatograph, a hydrophobic interaction chromatograph, a pseudo-affinity chromatograph, or a combination thereof. Preferably the liquid chromatograph includes a size exclusion chromatograph. In some other preferred forms, the chromatograph is selected from the list above, such that the principle of analyte selectivity is not based on gas phase ion mobility separation.

[0033] As discussed above, the methods further involve introducing a sample (preferably an eluted sample, eluted from a chromatograph) into a mass spectrometer. In some forms, the mass spectrometer uses a single particle technique, wherein masses of individual ions are determined from simultaneous measurements of their mass-to-charge ratio (m / z) and charge. Preferably, the mass spectrometer includes a charge detection mass spectrometer, preferably an Orbitrap mass spectrometer. Charge detection mass spectrometry is discussed in Todd, et al. , Anal. Chem. 2020, 92, 11357-11364, the contents of which are herein incorporated by reference. In particularly preferred forms, the chromatograph is a liquid chromatograph, and the mass spectrometer is a charge detection mass spectrometer.

[0034] Several analytes are suitable for analysis employing the disclosed methods and apparatuses. In some forms, the analyte includes, but is not limited to, proteins; lipoproteins; glycosylated proteins; protein complexes; protein aggregates such as amyloid fibers; infectious viruses; non-infectious viruses; nanoparticles; glycans; gene therapies; vaccines; vesicles such as exosomes; nucleic acids; or a combination thereof. b. Apparatuses for the acquisition of multiplexed analytical data

[0035] Also disclosed is a device configured to and / or capable of delivering two or more pulsed injections of a sample into a chromatograph, preferably a liquid chromatograph. The injection is controlled by an algorithm that optionally starts the acquisition on the mass spectrometer and then injects samples for defined time periods based on a binary injection sequence detailing whether to inject or not inject for each cycle, a specific injection time, and a total cycle time specifying the length of the cycle. The algorithm injects sample for the defined injection time when called for by the sequence. It then waits for the remaining cycle time before evaluating the next digit in the sequence. If an injection is not called for in that digit of the sequence, it waits the full cycle time without injecting the sample. Preferably, the device is operably linked to a valve (e.g., an injection valve), such that it controls the valve (e.g., an injection valve), and the algorithm programs the device with a pseudo-random binary sequence to control injections as well as the duration of each injection (injection time) and delay time between injections (cycle time). The length of the pseudo-random binary sequence (PRBS) can be defined by 2n- 1 , where n can be any positive integer > 2, such as between 2 and 100, inclusive, between 2 and 90, inclusive, between 2 and 80, inclusive, between 2 and 70, inclusive, between 2 and 60, inclusive, between 2 and 50, inclusive, between 2 and 40, inclusive, between 2 and 30, inclusive, between 2 and 20, inclusive, or between 2 and 10, inclusive, such as, 3 or 5. When necessary, additional zeros can be added to the end of the PRBS to ensure that all chromatographic peaks are observed prior to the end of the injection sequence. Injections can be performed by switching the valve from a bypass position to an inject position for a few seconds, such as between 1-10 seconds, 1-5 seconds, etc. Flow of sample can then be directed to a chromatograph, and subsequently to another valve preferably adjusted to send all of the flow or a fraction of the flow (e.g., approximately between 1% and 50%, such as 1%, 5%, 10%, 15%, 20%, 25%, 30%, 35%, 50%, 45%, and 50%) to a mass spectrometer. It should be noted that there are no fundamental constraints to how much flow can be sent to the mass spectrometer. It is dictated by the flow rate of the liquid chromatograph system and ion source on the mass spectrometer.

[0036] Preferably, the device is operably linked to an apparatus described herein, such as a chromatogram, a mass spectrometer, or a combination thereof. Preferably, operably linking the device to the mass spectrometer can serve the purpose of triggering the start of data collection through a contact closure to a receiving electrical port on the mass spectrometer. Preferably, the device is a microcontroller. The device could also be controlled directly by software on the chromatograph instrument onboard or on external computers. The chromatograph can be selected from a liquid chromatograph (such as size exclusion chromatograph, high performance liquid chromatograph, and reverse-phase liquid chromatograph), a column chromatograph, an ion-exchange chromatograph, a gel-permeation (molecular sieve) chromatograph, an affinity chromatograph, a hydrophobic interaction chromatograph, a pseudo-affinity chromatograph, or a combination thereof. Preferably the liquid chromatograph includes a size exclusion chromatograph. Preferably, the mass spectrometer utilizes a single particle technique, wherein masses of individual ions are determined from simultaneous measurements of their mass-to- charge ratio (m / z) and charge. Preferably, the mass spectrometer contains a charge detection mass spectrometer, for example, an Orbitrap mass spectrometer. c. Non-transitory media

[0037] Also disclosed is a non-transitory computer-readable medium with computer executable instructions stored thereon executed by a processor and configured to perform a method of demultiplexing time-resolved mass spectrometry data in real-time obtain from a mass spectrometer. Preferably, the mass spectrometry data includes time-resolved charge detection data. In some forms, the non-transitory computer-readable medium is operably linked to the mass spectrometer. In some forms, the mass spectrometer is operably linked to a liquid chromatograph. Preferably, the mass spectrometer includes a charge detection mass spectrometer.

[0038] In some forms, of the non-transitory computer-readable medium, the executable instructions include instructions to perform:

[0039] (i) Hadamard transform demultiplexing,

[0040] (ii) Fourier transform demultiplexing,

[0041] (iii) removal of Hadamard transform spectral artifacts, or

[0042] (iv) a combination selected from two or more of (i) to (iii).

[0043] The disclosed methods, apparatuses, non-transitory media, and components thereof can be further understood through the following numbered paragraphs: Paragraph 1. A method for analyte detection, the method containing: delivering two or more pulsed injections of a non-gas phase sample into a liquid chromatograph operably linked to a mass spectrometer, wherein the injecting of the two or more pulses is controlled by an algorithm, preferably wherein analyte selectivity of the chromatograph is not based on (i) gas phase ion mobility separation or (ii) gas phase separation, preferably wherein the chromatograph contains a liquid chromatograph.

[0044] Paragraph 2. The method of paragraph 1 , further including sampling the analyte eluted from the chromatograph into the mass spectrometer.

[0045] Paragraph 3. The method of paragraph 2, further including generating a series of mass spectrometry scans for the sample.

[0046] Paragraph 4. The method of paragraph 3, further including extracting intensity over time for specific regions of the scans, optionally to produce an extracted ion chromatogram.

[0047] Paragraph 5. The method of paragraph 3 or 4, further including demultiplexing intensities in the scans. Paragraph 6. The method of paragraph 5, wherein demultiplexing is performed using a Hadamard Transform, or correlation techniques (such as correlation chromatography techniques preferably including a multiplexed binary injection sequence).

[0048] Paragraph 7 The method of paragraph 5 or 6, further including reassembling multiple demultiplexed datasets produces a second dataset containing mass, charge, and retention time of the analyte, preferably wherein the reassembling is performed using Hadamard Transform. Paragraph 8. The method of any one of paragraphs 1 to 7, wherein the chromatograph is selected from the group of a liquid chromatograph (such as size exclusion chromatograph, high performance liquid chromatograph, and reverse-phase liquid chromatograph), a column chromatograph, an ion-exchange chromatograph, a gel-permeation (molecular sieve) chromatograph, an affinity chromatograph, a paper chromatograph, a thin-layer chromatograph, a dye-ligand chromatograph, a hydrophobic interaction chromatograph, a pseudo-affinity chromatograph, or a combination thereof, preferably a liquid chromatograph, such as a size exclusion chromatograph.

[0049] Paragraph 9. The method of any one of paragraphs 1 to 8, wherein the mass spectrometer uses a single particle technique, wherein masses of individual ions are determined from simultaneous measurements of their mass-to-charge ratio (m / z) and charge.

[0050] Paragraph 10. The method of any one of paragraphs 1 to 9, wherein the mass spectrometer contains a charge detection mass spectrometer, preferably an Orbitrap mass spectrometer. Paragraph 11. The method of any one of paragraphs 1 to 10, wherein the sample contains an analyte selected from proteins; lipoproteins; glycosylated proteins; protein complexes; protein aggregates such as amyloid fibers; infectious viruses; non-infectious viruses; nanoparticles; glycans; gene therapies; vaccines; vesicles such as exosomes; nucleic acids; or a combination thereof.

[0051] Paragraph 12. A device configured with an algorithm to deliver two or more pulsed injections of a sample into a chromatograph, optionally wherein the sample contains an analyte to be detected. Paragraph 13. The device of paragraph 12, operably linked to an injection valve.

[0052] Paragraph 14. The device of paragraph 12 or 13, operably linked to a mass spectrometer. Paragraph 15. The device of any one of paragraphs 12 to 14, wherein the chromatograph is selected from the group of a liquid chromatograph (such as size exclusion chromatograph, high performance liquid chromatograph, and reverse-phase liquid chromatograph), a column chromatograph, an ion-exchange chromatograph, a gel-permeation (molecular sieve) chromatograph, an affinity chromatograph, a hydrophobic interaction chromatograph, a pseudo- affinity chromatograph, or a combination thereof, preferably a liquid chromatograph, such as a size exclusion chromatograph.

[0053] Paragraph 16. The device of paragraph 14 or 15, wherein the mass spectrometer uses a single particle technique, wherein masses of individual ions are determined from simultaneous measurements of their mass-to-charge ratio (m / z) and charge.

[0054] Paragraph 17. The device of any one of paragraphs 14 to 16, wherein the mass spectrometer contains a charge detection mass spectrometer, preferably an Orbitrap mass spectrometer. Paragraph 18. The device of any one of paragraphs 12 to 17, wherein a binary sequence of bits controls delivery of the sample, preferably wherein the binary sequence is a pseudo-random binary sequence.

[0055] Paragraph 19. The device of paragraph 18, wherein the binary sequence of bits controls injections, duration of each injection (injection time), delay time between injections (cycle time), and a combination thereof.

[0056] Paragraph 20. The device of any one of paragraphs 12 to 19, wherein the device is a microcontroller.

[0057] Paragraph 21. The device of any one of paragraphs 12 to 20, wherein the sample is a non-gas phase sample.

[0058] Paragraph 22. A non-transitory computer-readable medium with computer executable instructions stored thereon executed by a processor and configured to perform a method of demultiplexing time-resolved mass spectrometry data in real-time obtain from a mass spectrometer.

[0059] Paragraph 23. The non-transitory computer-readable medium of paragraph 22, wherein the non- transitory computer-readable medium is operably linked to the mass spectrometer.

[0060] Paragraph 24. The non-transitory computer-readable medium of paragraph 22 or 23, wherein the mass spectrometer is operably linked to a liquid chromatograph.

[0061] Paragraph 25. The non-transitory computer-readable medium of any one of paragraphs 22 to 24, wherein the mass spectrometer includes a charge detection mass spectrometer.

[0062] Paragraph 26. The non-transitory computer-readable medium of any one of paragraphs 22 to 25, wherein the mass spectrometry data includes time-resolved charge detection data.

[0063] Paragraph 27. The non-transitory computer-readable medium of any one of paragraphs 22 to 26, wherein the executable instructions include instructions to perform:

[0064] (i) Hadamard transform demultiplexing,

[0065] (ii) Fourier transform demultiplexing,

[0066] (iii) removal of Hadamard transform spectral artifacts, or (iv) a combination selected from two or more of (i) to (iii).

[0067] The methods, apparatuses, and components thereof herein described are further illustrated in the following examples, which are provided by way of illustration and are not intended to be limiting. It will be appreciated that variations in proportions and alternatives in elements of the components shown will be apparent to those skilled in the art and are within the scope of disclosed forms. All parts or amounts, unless otherwise specified, are by weight.

[0068] Examples

[0069] Example 1: Coupling online size exclusion chromatography with charge detection-mass spectrometry using Hadamard transform multiplexing

[0070] Analysis of large, heterogeneous biomolecules by native mass spectrometry (MS) is challenging because the charge states of ions often cannot be determined. Charge detection-mass spectrometry (CD-MS) solves this challenge by directly measuring the charge states of individual ions (Fuerstenau, Rapid Commun. Mass Spectrom. 1995, 9 (15), 1528-1538; Pierson, et al. J. Am. Soc. Mass Spectrom. 2015, 26 (7), 1213—1220; Womer, T, et al. J. R. Nat. Methods 2020, 17 (4), 395-398; Kafader, J. O et al., Anal. Chem. 2019, 91 (4), 2776- 2783). By avoiding the need to resolve charge states or isotopes, CD-MS has enabled characterization of very large assemblies such as protein complexes (Wbrner, T, et al. J. R. Nat. Methods 2020, 17 (4), 395-398), intact virus capsids (Miller, L. M.; et al., Anal. Chem. 2021, 93 (35), 11965-11972),5vaccines (Miller, et al., Anal. Chem. 2021, 93 (35), 11965-11972), and gene therapies (Kostelic, M. M et al., Anal. Chem. 2022, 94 (34), 11723-11727; Barnes. L. F et al., Mol. Ther.-Methods Clin. Dev. 2021, 23, 87-97). Growing needs in biotechnology to characterize these heterogeneous assemblies has led to increased interest in CD-MS.

[0071] One requirement of CD-MS is that only a limited number of ions can be measured at a time, which requires acquisition of hundreds or thousands of spectra to provide enough detection events for mass measurements with acceptable signal-to-noise ratios (SNR). The need for these long acquisitions generally limits the coupling of CD-MS with online separation techniques, such as size exclusion chromatography (SEC), which typically only provides enough time for 50-200 scans during the elution of a chromatographic peak.

[0072] Attempts have been made to speed up CD-MS by refining instrumentation to enable the measurement of multiple ions simultaneously (Kafader, J. O et al., Anal. Chem. 2019, 91 (4), 2776- 2783; Harper, C. C et al., Anal. Chem. 2019, 91 (11), 7458-7465) and improve charge state assignments with deconvolution algorithms (Kostelic, M. M. el al., Anal. Chem. 2021, 93 (44), 14722-14729). Research to evaluate the feasibility of online SEC-CD-MS produced deconvolved mass spectra that, while capable of measuring protein masses, had lower SNR and contained significantly more erroneous peaks than spectra produced from infusion experiments. 11 Other research using online buffer exchange (VanAemum, Z. L et al.,. Nat. Protoc. 2020, 15 (3), 1132-1157; Liu, W et a / ., Anal. Chem. 2023, 95 (47), 17212-17219) with CD-MS for the analysis of IgM assemblies was also able to produce mass distributions after alteration of MS conditions but was still limited to only several hundred ions per run (Yin, V.; et al., J. Am. Soc. Mass Spectrom. 2024, 35 (6), 1320-1329). Yet other research explored the use of capillary electrophoresis (CE) and other fluidic devices with CD-MS and observed that short elution profiles often produced low-quality CD-MS spectra (McGee, J. P et al., Anal. Chem. 2022, 94 (48), 16543-16548). Hadamard transform multiplexing provides a powerful strategy for improving throughput and SNR when scan / separation speeds are slower than detection speeds. It has been applied to a range of separation techniques, including gas chromatography (Fan, G.-T et al., Taianta 2014, 120, 386—390; Fan, Z.; et al., J. Chromatogr. A 2010, 1217 (5), 755-760), liquid chromatography (LC)( Lin, C.-H et al., Anal. Chem. 2008, 80 (15), 5755-5759; Siegle, A. F. Anal. Chem. 2014, 86 (21), 10828- 10833), and CE (Kaneta, T.; et al.,. Anal.

[0073] Chem. 1999, 71 (23), 5444-5446; Kaneta. T et al., Anal. Chem. 2002, 74 (10), 2257-2260) and to enhance a range of analytical techniques, including optical spectroscopy (Marshall, et al., Anal. Chem. 1975, 47 (4), 491A-504A. 91 A-504A; Harwit, M.; Sloane, N. J. A. The Basic Theory of Hadamard Transform Spectrometers and Imagers. In Hadamard Transform Optics; Harwit, M., Sloane, N. J. A., Eds.; Academic Press, 1979; Chapter 3, pp 44-95), time-of-flight MS (Brock, A et al., Anal. Chem. 1998, 70 (18), 3735-3741; Brock, N et al.,. N. Rev. Sci. Instrum. 2000, 71 (3), 1306—1318; Brais, C. J et al., J. Mass Spectrom. Rev. 2021, 40 (5), 647-669) and ion mobility spectrometry (Clowers, B. H et al., Anal. Chem. 2006, 78 (1), 44-51; Szumlas, A. W et al., Anal. Chem. 2006, 78 (13), 4474—4481; Ibrahim, Y. M et al., Anal. Chem. 2016, 88 (24), 12152-12160; Reinecke, T.; et al., Trends Anal. Chem. 2019, 116, 340-345; Clowers, B. H. et al., Anal. Chem. 2021, 93 (14), 5727-5734). With conventional single-plex methods, a sample is injected, and the second sample must wait until the first injection is completed before it can be injected. With Hadamard transform multiplexing, multiple overlapping injections are performed in rapid succession as dictated by a pseudorandom binary sequence (PRBS). The resulting chromatogram contains multiple peaks for each analyte in the sample, which can be demultiplexed by multiplication with a matrix derived from the PRBS. This results in a single peak for each analyte with accurate retention times and improved SNR.

[0074] Here, Hadamard transform multiplexing was applied to SEC coupled with CD-MS for the acquisition of spectra from multiple chromatographic peaks. A simple multi-injection strategy using a standard LC pump and 6-port injection valve is described. An open-source Hadamard transform CD-MS demultiplexing algorithm built on UniDec software platform to analyze the data is used.10,32,33Finally, it is demonstrated the Hadamard transform SEC-CD-MS method on a mixture of model proteins and on a complex cell lysate mixture. Together, this experimental design provides a general solution to the problem of coupling CD-MS detection with isocratic separations that can be easily adopted on a range of platforms for a range of applications.

[0075] Materials and methods

[0076] Materials

[0077] P-galactosidase and GroEL were purchased from Sigma or expressed and purified as previously described34. All protein samples were buffer exchanged and purified using either Biospin P6 (BioRad) size exclusion spin columns or a Superdex 6 Increase 10 / 300 (Cytiva) size exclusion column. Protein samples were diluted to 0.1 - 0.5 mg / ml in 200 mM ammonium acetate prior to analysis.

[0078] To produce Escherichia coli cell lysate, E. coli OverExpress C43 (DE3) cells were cultured in 1 L flasks of Terrific Broth media, as previously described.35Cells were then harvested by centrifugation at 5000 rpm for 10 min at 4 °C. Cells were lysed in 20 mM Tris, 100 mM NaCl, pH 7.4 with protease inhibitor using an LM20 microfluidizer (Microfluidics International Corporation) at 20,000 psi. The cell lysate was clarified by centrifugation at 20,000g for 25 min at 4 °C, and membranes were removed at 100,000g for 2 h and 10 min at 4 °C. The supernatant with soluble proteins was diluted lOx in 200 mM ammonium acetate and reconcentrated in a 300 kDa molecular weight cutoff filter five times to isolate high-mass proteins.

[0079] (i) Multi-injection online size exclusion chromatography (SEC)

[0080] A constant flow of 200 mM ammonium acetate was supplied by an Agilent 1100 quaternary pump at a flow rate of 0.2 ml / min. This flow was directed to a 6-port switching valve with either a 100 pL or a 500 pL sample loop containing the protein sample (FIG. 1). The valve was controlled by an Arduino Uno microcontroller, which was programmed with a pseudorandom binary sequence (PRBS) to control injections as well as the duration of each injection (injection time) and delay time between injections (cycle time). Injections were performed by switching the valve from the bypass position to the inject position for 1-5 seconds. Flow was then directed to an Agilent AdvanceBio SEC column (4.6 X 300 mm, 2.7 pm particle size, 300 A pore size) and then to a Idex Micro- Splitter valve adjusted to send approximately 25% of the flow to the mass spectrometer. Code for the Arduino is provided at github.com / michaelmarty / UniDec / blob / master / PublicScripts / Multiplexing / HTswitch2.ino. Injections were performed according to a pseudo-random binary sequence (PRBS) with a length of 2” - 1 where n = 1, 3, 5, or 6. The specific sequences used were obtained from Harwit and Sloane (Harwit, M.; Sloane, N. J. A. The Basic Theory of Hadamard Transform Spectrometers and Imagers. In Hadamard Transform Optics; Harwit, M., Sloane, N. J. A., Eds.; Academic Press, 1979; Chapter 3, pp 44-95) and are “1110100”, “0000100101100111110001101110101”, and “0000010000110001010011 11010001 11001001011011 1-011001 101010111111” for the 3-, 5-, and 6-bit sequences, respectively. A one-bit sequence is simply “1”, a standard single-plex chromatographic injection. The 3 -bit sequence was chosen because it had a quick analysis time and contained enough injections to see the injection pattern. The 5-bit sequence was chosen because it was longer than the 3-bit, with more injections to demonstrate the effects of scaling the PRBS length. The longer 6-bit sequence was used for complex lysate data, where the analysis benefitted from additional injections. When necessary, additional zeros were added to the end of the PRBS to ensure that all chromatographic peaks were observed prior to the end of the injection sequence.

[0081] In addition to the PRBS, the two important parameters to define the injection sequence were the injection time (how long the valve stayed in the inject position before being switched back to load during each cycle) and the cycle time (how long the code waited before proceeding to the next digit of the PRBS). At the beginning of each cycle, the PRBS was evaluated. If the digit was a 1, the valve was switched to inject for as long as specified by the injection time. After the injection time was done, it was switched back to load for the remainder of the cycle time. If the digit was a 0, the valve remained in load for the entire duration of the cycle time. After one cycle was completed, it proceeded to evaluate the next digit in the sequence and continues until the sequence was completed. Here, the injection time was set to 1.5 s (5 pL injection), and the cycle time was 1 min. A trigger was sent from the Arduino to the MS to start acquisition at the start of the sequence, and the acquisition time on the MS was set to the total sequence length multiplied by the cycle time. No other coordination is performed to synchronize MS scans and LC injections.

[0082] After the loop, flow was directed to a Superose 6 Increase 5 / 150 column (Cytiva) and then to the heated electrospray ionization (HESI) source. Approximately 10 ft. of 0.005 in. PEEK “resistor tube” was placed between the grounding union and ESI needle to reduce spray current while spraying high ionic strength mobile phase at high flow rates (VanAernum, Z. L el al.,. Nat. Protoc. 2020, 15 (3), 1132-1157). All data was collected with at least three replicate injection sequences, and representative data was provided. (ii) P-galaclosidase Static nESI CD-MS

[0083] Static nano-electrospray ionization (nESI) charge detection-mass spectrometry (CD-MS) was performed on the same Thermo Fisher Q-Exactive HF Ultra High Mass Range Orbitrap mass spectrometer using Direct Mass Technology (DMT). Borosilicate glass capillaries, 100 mm length, 1.2 mm outer diameter, 0.68 mm inner diameter (World Precision Instruments) were pulled by a P-1000 Micropipette Puller (Sutter Instruments) to form nESI needles. Ions were generated by a Thermo Fisher Nanospray Flex source using a spray voltage of 1.2 kV and capillary temperature of 200 °C. The m / z range was set to 5,000-20,000 m / z and extended trapping voltage set to 25 V. CD-MS data was collected with DMT enabled, at a resolution setting of 240,000, 1 microscan, and ion injection times fixed to 20 ms. 0.1 mg / mL P- galactosidase in 200 mM ammonium acetate was analyzed in triplicate with static electrospray CD-MS.

[0084] To directly compare the charge and mass distribution of the 3-bit Hadamard transform SEC-CD-MS sequence, the number of scans analyzed with static electrospray CD-MS was limited to match the number of ions in the 3-bit sequence used to generate the charge histogram and mass distribution. When compared with the 3-bit sequence (data not shown), static electrospray CD-MS yielded higher charged ions with lower m / z (data not shown). However, both methods yielded a similar mass distribution to the 3-bit sequence. Slight differences in mass distribution are likely due to the increased difficulty in desolvating ions generated from the much higher flow rate of the HESI source used for online SEC experiments. The lower charge states observed with online SEC CD-MS are similarly likely due to charge stripping during desolvation. Similar spectra were obtained in direct infusion experiments on the HESI source without separations, but the large sample consumption precluded collecting enough data to compare with the same number of ions.

[0085] (iii) fi-galactosidase / GroEL Mixture nESI CD-MS

[0086] A mixture of 0. 1 mg / mL -galactosidase, and 0.2 mg / mL GroEL in 200 mM ammonium acetate was analyzed in a similar fashion to above with static electrospray ionization CD-MS. This yielded 48,588, 48,735, and 46,297 ions in the triplicate analyses, which was compared to a 5-bit Hadamard transform SEC-CD-MS sequence of 47,592 ions.

[0087] Like the lone P-galactosidase sample, the mixture yielded higher charge states and lower m / z for both the P-galactosidase and GroEL (data not shown). The deconvolved mass distribution of the mixture analyzed with static electrospray CD-MS (data not shown) is similar with the 5-bit sequence (data not shown) for P-galactosidase and GroEL. Slight differences in relative intensity between the two species are due to the fact that different samples were used and were prepared several months apart.

[0088] (iv) Mass spectrometry

[0089] All experiments were performed on a Thermo Fisher Scientific Q-Exactive HF Ultra High Mass Range (UHMR) Orbitrap mass spectrometer using the Direct Mass Technology (DMT) acquisition mode, which implements the STORI processing method (Kafader,; et al., J. Am. Soc. Mass Spectrom. 2019, 30 (11), 2200-2203). Ions were generated by a HESI source using a spray voltage of 3.8 kV, a sheath gas setting of 20, and an auxiliary gas setting of 5. The m / z range was set to 3000-12,000 m / z, and extended trapping settings of 25-100 V were used to maximize desolvation while minimizing fragmentation of ions. CD-MS spectra with DMT were collected with a resolution setting of 240,000, 1 microscan, and ion injection times were fixed between 50 and 200 ms to maintain ion populations of 10-100 ions per scan. All 1 -, 3-, and 5-bit sequences were collected with 200 ms ion injection times. The 6-bit sequence for E. coli lysate was collected with 50 ms ion injection times, likely due to a higher concentration of protein in the sample. It may also be possible to use automatic injection control (AIC) acquisition features (McGee, J. P et al., Anal. Chem. 2022, 94 (48), 16543-16548), with further tuning to adjust the AIC algorithm to respond on a chromatographic time scale.

[0090] (v) Data Processing

[0091] Raw files containing CD-MS data were initially processed in STORIboard (Proteinaceous) and then further processed using a custom Hadamard transform algorithm implemented in the UniChromCD window of UniDec, which is described briefly here and in more detail in a companion manuscript (Sanders, J.D., Anal Chem., 2024, 96(37), 15014-15022). Extracted ion chromatograms (EICs) were generated by defining an m / z range and charge state range for each ion population of interest. A kernel array was constructed based on the PRBS (including zero pads when applicable) by substituting - Is for 0s in the sequence and infilling with 0s so that the array was the same size as the number of scans in the full HT cycle. The kernel array was then rotated to move the zero-padded section to the beginning of the sequence and smoothed by convolution with a Gaussian function with a width (fwhm) of 2 scans. The modified kernel was then convolved with each EIC, and the resulting demultiplexed EICs were shifted back to the original time index. The compiled program and source code can be found at: github.com / michaelmarty / UniDec. Additional details on deconvolution are provided below.

[0092] ( vi) Deconvolution Settings

[0093] Deconvolution of the data was performed in UniDecCD for static infusions and in UniChromCD for SEC-CD-MS data. To generate the raw m / z vs. charge histograms, the data was binned with 10 m / z bins and a scan compression of 10. Mass was binned every 1000 (Da). Deconvolution was performed with a m / z spread of 20 and a charge spread of 5. The charge states were smoothed with default settings, and the smoothing of nearby points was set to 1. The beta value from the SoftMax function was set to 0.01 to reduce charge mis-assignments. The m / z to mass transformation was performed by integration, and data was plotted in the reconvolved / profile mode.

[0094] Results

[0095] Comparison of Injection Sequence Lengths

[0096] To evaluate the Hadamard transform multiplexed injection strategy, injection sequences with different lengths were compared. A 0.1 mg / mL (~0.2 pM) solution of P-galactosidase was analyzed using 1 -bit (1 injection), 3-bit (4 injections), and 5-bit (16 injections) sequences in triplicate. The 3- and 5-bit sequences were each padded with 10 zeros, resulting in 17 and 41 min run times, respectively. The single injection ( 1 -bit) chromatogram was recorded for 12 min, which was the minimum required to elute the sample. The multiplexed total ion chromatograms (TICs) are similar in intensity and retention time to the 1 -bit (single-plex) injection, as seen in FIG. 2A. However, the areas of the demultiplexed peaks scale with the number of injections (FIG. 2B), demonstrating improved signal for the multiplexed data. Importantly, even though the 5-bit PRBS used in this experiment starts with 4 zeros (i.e., no injection occurs for the first 4 min), the retention time is correctly reported as 9.5 min in the demultiplexed chromatogram. Together, these data demonstrate the ability of the Hadamard transform algorithm to sum the areas of all individual peaks in the multiplexed chromatogram into a single peak at the correct retention time.

[0097] As shown in Table 1, the Hadamard transform multiplexed data yield a significantly higher number of total ions collected, peaks that were broader, likely due to more challenging desolvation, and at lower average charge states, likely due to charge stripping during desolvation. However, these differences are not intrinsic to the separation, and good data can be obtained with either method.

[0098] Table 1. Total Number of Injections, Total Run Times, Number of Ions, and Average Number of lons / min for 1 -, 3-, and 5-Bit Injection Sequences for P- Galactosidase Data Presented in FIGs. 2A-2E and FIG. 3A. aFor 1 -bit data, the retention time of 12 min defines the total sequence runtime. For 3- and 5 -bit data, the runtime is defined by a 1 min cycle time times the sequence length, including zero-padding.

[0099] Separation of a Mixture of Proteins

[0100] Next, to test the ability to separate two proteins, a simple mixture containing P- galactosidase and GroEL was prepared and injected using the 5-bit sequence. Compared to the single protein sample, the TIC (FIG. 3A) shows broader features where the injection sequence pattern is not obvious. EICs were created for each protein by selecting both an m / z and charge range that encompassed the full charge state distributions, as seen in FIG. 3C. The resulting EICs (FIG. 3A) reveal a notable shift between the two multiplexed chromatograms. Hadamard transform demultiplexing produced a TIC and EICs that were well-defined and with minimal artifacts (FIG. 3B). Although the resolving power of the short SEC column used here was not sufficient to resolve the two species in the TIC, the EICs show clear separation and the expected elution order with the larger GroEL eluting first.

[0101] As with the isolated P-galactosidase data above, the mass distributions were similar between the SEC-CD-MS and conventional static infusion (data not shown). After UniDec deconvolution to reduce the spread in the charge dimension, the deconvolved mass spectrum (FIG. 3D) shows extracted mass assignments for both species. Somewhat broader peaks were observed from GroEL, likely due to poorer desolvation on the HESI source. Overall, these data demonstrate that Hadamard transform SEC-CD-MS can be used to analyze mixed samples and extract the chromatograms and mass distributions of isolated species.

[0102] Separation of a Complex Mixture

[0103] Finally, the performance of Hadamard transform SEC-CD-MS on a complex mixture of E. coli cell lysate was tested. To deplete smaller species, a 300 kDa molecular weight cutoff filter was used. Here, 500 pL of E. coli cell lysate was diluted 5x to a total volume of 2500 pL, then centrifuged back down to 500 pL. This process was repeated a total of 5 times. However, initial experiments were still dominated by low mass proteins, so the instrument was tuned to further reject low mass ions by setting the injection flatapole, interflatapole lens, and bent flatapole to 9, 8, and 10 V, respectively, similar to the “voltage rollercoaster filtering” method described by McGee, et al. (McGee, et al., J. Am. Soc. Mass Spectrom. 2020, 31, 763-767). This tuning resulted in a mass distribution spanning the 400 kDa to 1 MDa range seen in FIG. 4D. To collect more scans on this complex mixture, it was injected using a 6-bit injection sequence (63 segments, 32 injections) for a total run time of 78 min including 15 zero pads.

[0104] Even after deconvolution, the m / z vs charge distribution (FIG. 4C) contains numerous species that were close in m / z and charge, which hindered our ability to cleanly select protein signals using the simple m / z and charge state range method in FIGs. 3A and 3B. Thus, a method to select protein signals along predicted m / z vs charge curves was developed. Using curve selection to produce the EICs of the 5 most prominent features (FIG. 4A), the demultiplexed chromatograms (FIG. 4B) were further processed using a masked multiplex (Clowers, B. H. et al., Anal. Chem. 2021, 93 (14), 5727-5734) method to remove artifacts. Here, it was observed that the masses generally correlated with the retention time, as expected. Attempts to extract less prominent features produced demultiplexed EICs with poor SNR and peak shapes, indicating that even longer injection sequences or more precise extraction settings could be implemented to fully resolve lower abundance species in mixtures such as this.

[0105] Although it is difficult to confidently identify proteins from this mixture by intact mass alone, a few observations can be made about the potential of Hadamard transform SEC-CD-MS for the analysis of complex mixtures. First, even when five or fewer ions are observed in each retention time bin of an EIC, the Hadamard transform algorithm is still able to produce good quality chromatographic peaks that allow retention times to be accurately measured. Second, in this example both the lightest blue and light green EICs originate from relatively crowded areas of the m / z vs charge distribution and may contain contributions from two or more proteins that could have different retention times. In FIG. 4B, multimodal retention time distributions were observed, demonstrating that Hadamard transform demultiplexing can resolve complex chromatographic features that are not simple Gaussian distributions. In any case, having the ability to isolate data in mass, charge, and / or retention time helps in analysis of these peaks that would be otherwise difficult to discern. Also contemplated are the use of higher resolution SEC columns, longer injection sequences, and additional prefractionation to probe the capacity of this method to characterize complex mixtures with increased dynamic range.

[0106] Conclusions

[0107] Performing CD-MS on a chromatographic time scale is challenging because many ions are needed to produce quality CD-MS spectra, but only a few ions can be collected in each scan. By stacking together multiple injections in short succession, Hadamard transform demultiplexing improves SEC-CD-MS by increasing SNR in both chromatographic and mass dimensions. It enhances the duty cycle relative to simply averaging repeated single-plex injections, enabling more efficient use of instrument time without sacrificing retention time resolution or information.

[0108] Hadamard transform multiplexing can be broadly applied to any chromatography system that uses isocratic gradients and that is capable of handling multiple injections, making it a generalized and versatile technique for generating high-quality LC-CD-MS data. For example, isocratic HIC has been applied to antibodies (Wei, B et al., Anal. Chem. 2019, 91 (24), 15360-15364), and CE-MS has been attempted with proteins as large as GroEL (Marie, et al., Adv. Sci. 2024, 11 (11), 2306824), both of which would be compatible with multiplexed injections and CD-MS data acquisition.

[0109] Finally, presented here is an improved and cheaper design of a system using open-source software and hardware that can be readily adapted to different instrument systems. The system can find relatively easy implementation in most laboratories. Also contemplated is an adaptation on automated SEC-MS systems by reprogramming the autosampler to introduce multiple injections in a predefined sequence. Combining these advanced separation strategies with improved instrumental methods (Pierson, et al. J. Am. Soc. Mass Spectrom. 2015, 26 (7), 1213—1220; Goodwin, et al., J. Am. Soc. Mass Spectrom. 2024, 35 (4), 658—662; Parikh, et al., Anal. Chem. 2024, 96 (7), 3062-3069) for collecting CD-MS data improves the characterization of heterogeneous assemblies for a range of applications.

[0110] Example 2: UniChromCD for Demultiplexing Time-Resolved Charge Detection Mass Spectrometry Data

[0111] Charge detection-mass spectrometry (CD-MS) is uniquely powerful for studying large and / or heterogeneous biomolecules where conventional MS techniques fail due to lack of well- defined charge state distributions or isotopic envelopes typically used to identify charge states (Pierson, et al., J. Am. Soc. Mass Speclrom. 2015, 26 (7), 1213—1220; Desligniere, el al., J. R. Acc. Chem. Res. 2023, 56 (12), 1458-1468; Kafader, el al., Anal. Chem. 2019, 91, 2776-2783; Harper, et al., Anal. Chem. 2019, 91 (11), 7458-7465). In particular, CD-MS has found applications in the analysis of virus capsids (Miller, et al., Essays Biochem. 2023, 67 (2), 315— 323; Kostelic, et al., Anal. Chem. 2022, 94 (34), 11723— 11727), lipoproteins (Lutomski, et al., Anal. Chem. 2018, 90 (11), 6353-6356), exosomes (Brown, et al., Anal. Chem. 2020, 92 (4), 3285-3292), nanoparticles (Harper, et al., ACS Nano 2023, 17 (8), 7765-7774), and heavily modified proteins (den Boer, et al., Anal. Chem. 2022, 94 (2), 892-900). By analyzing only a few ions at a time, CD-MS takes advantage of the fact that the magnitude of an ion’ s induced current signal scales directly with the charge state of the ion, allowing charge states to be assigned for each ion based on a calibration of intensity vs charge.

[0112] Because CD-MS only analyzes a few ions at a time, hundreds or thousands of scans are often needed to produce high-quality data. Thus, coupling CD-MS to time dispersive separation techniques such as liquid chromatography (LC) presents a challenge due to an inherent duty cycle mismatch. Previous efforts to couple CD-MS with LC (Strasser, et al., J. Anal. Chem. 2023, 95, 15118-15124; Yin, V et al., J. Am. Soc. Mass Spectrom. 2024, 35, 1320-1329) and capillary electrophoresis (CE)( McGee, Anal. Chem. 2022, 94 (48), 16543-16548) have achieved some success, although spectra had relatively low signal-to-noise ratios (SNR).

[0113] One strategy for addressing duty cycle mismatch is with multiplexed injections. Multiplexing of gas chromatography (GC)( Fan, Z.; et al., J. Chromatogr. A 2010, 1217 (5), 755—760; Fan, et al., Taianta 2014, 120, 386-390), LC (Lin, et al., Anal. Chem. 2008, 80 (15), 5755-5759; Siegle, et al., Anal. Chem. 2014, 86 (21), 10828- 10833), and CE (Kaneta, et al., Anal. Chem. 1999, 71 (23), 5444—5446; Kaneta, T.; et al., Anal. Chem. 2002, 74 (10), 2257-2260) injections using Hadamard transform (HT) can improve throughput and SNR. Using a variety of injection apparatuses, a sample is injected onto the column multiple times in close succession according to a pseudorandom binary sequence (PRBS). The resulting chromatogram, which contains multiple peaks for each analyte in the sample, is then demultiplexed using an inverse Hadamard transform, yielding a single peak for each analyte with a demultiplexed area equal to the sum of the areas of all corresponding peaks in the multiplexed chromatogram.

[0114] HT multiplexing strategies offer a substantial improvement in duty cycle, defined here as the fraction of total MS scans that contain useful ion signals, and are thus attractive options to improve the throughput of time-dispersive separation methods with CD-MS. However, the incorporation of CD-MS data into demultiplexing workflows adds significant complexity because each scan must be processed to extract m / z and charge information, and the high scan- to-scan variability inherent to CD-MS data often produces noisy and inconsistent features in time-domain data.

[0115] This study describes a software package, UniChromCD, to process time-resolved CD-MS data, perform both HT demultiplexing operations, and perform many postprocessing tasks such as removal of HT spectral artifacts.

[0116] Materials and Methods

[0117] Instrumentation, Sample Preparation, and Data Collection

[0118] The instrumentation used to generate the example SEC-MS data presented in this manuscript is described in detail in Example 1 (Sanders, J.; et al., ChemRxiv 2024 DOI: 10.26434 / chemrxiv-2024-h08x5), so only a brief description is provided here. 0-galactosidase and GroEL were purchased from Sigma or expressed and purified as described previously (Grason, et al., Proc. Natl. Acad. Sci. U.S.A. 2008, 105(45), 17339-344). Glutamate dehydrogenase (GDH) was purchased from Sigma. All samples were buffer exchanged into 0.2 M ammonium acetate (Sigma) using Biospin P6 (BioRad) spin columns and diluted to 0.1 -0.5 mg / mL. The sample concentration was optimized by diluting the sample to achieve the right level of ions in the CD-MS measurement. Measurements were collected in triplicate, and a single representative spectrum is usually shown. Only a single replicate of the Shift 15 data was collected because it did not yield usable data, as discussed below.

[0119] Online SEC used an Agilent 1100 quaternary pump connected to a 6-port switching valve, which was equipped with a 100 pL sample loop and controlled by an Arduino microcontroller. The column used was a Superose 6 Increase 5 / 150 (Cytiva), and the flow of 200 mM ammonium acetate was maintained at 0.2 mL / min. The resin bed dimensions are 5 mm in diameter and 150 mm in height, with a total volume of approximately 3 mL. This column was used to demonstrate the approach because it has a broad fractionation range (from 5 kDa to 5 MDa), a fast elution time to allow rapid testing, and excellent biocompatibility. Longer columns with higher resolution and more carefully tuned parameters can be used instead to improve the chromatographic resolution.

[0120] Each injection was for 1.5 s, which yielded a 5 pL injection, and the total cycle time was 1 min. These parameters were optimized based on the resolution of the column to achieve good peak density, distinct peaks, and minimal total analysis time. CD-MS data were acquired in direct mass technology (DMT) mode on a Q Exactive HF UHMR Orbitrap mass spectrometer (Thermo Scientific Instruments) using a resolution setting of 240,000 at m / z 400 (512 ms transient length) and fixed ion injection times between 50 and 200 ms, which were optimized to achieve a single ion level. Additional details can be found in the materials and methods of Example 1.

[0121] Software Overview

[0122] UniChromCD was programmed as a new module within the UniDec software package. It is built entirely in Python, using libraries such as numpy (Harris, et al., Nature 2020, 585 (7825), 357-362), scipy (Virtanen et al., Nat. Methods 2020, 17 (3), 261-272), matplotlib (Hunter, et al., Anal. Chem. 2021, 93 (44), 14722-14729), and wxPython (version 4.2). It is designed with a model-presenter- view architecture, where the model is a backend application programming interface (API) that can be accessed through Python scripting (HTEng.py), the view is a wxPython frontend GUI, and the presenter (UniChromCD. -py) controls the overall program flow by connecting the front and backends.

[0123] UniChromCD inherits many of the same functions and GUI elements as UniDecCD, the module for analysis of individual CD-MS spectra (Kostelic, et al., Anal. Chem. 2021, 93 (44), 14722-14729). For example, the frontend GUI is the same as UniDecCD but includes added panels for demultiplexing and additional plotting panels, as shown in FIG. 5. The presenter (UniChromCDApp) inherits the presenter from UniDecCD (UniDecCDApp) and includes additional func-tions to make use of the new frontend features. The backend engine (UniChromCDEng) inherits the engine from Un-iDecCD (UniDecCD) and the additional demultiplexing class (HTEng). Thus, the UniChromCD engine contains three primary components, the prior UniDecCD functions, new time-domain UniChromCD functions, and new demultiplexing functions. The relationship between these different types of code is shown in FIG. 6 and described below.

[0124] From UniDecCD, UniChromCD inherits the existing tools for opening, processing, transforming, and deconvolving CD-MS data of various types (FIG. 6, left). The new UniChromCD engine (FIG. 6, center) class adds time-domain processing of CD-MS scans to generate histogram stacks, which are three-dimensional (3D) data sets of m / z vs charge vs time. It includes new tools for creating total ion chromatograms (TICs) and extracted ion chromatograms (EICs) from the CD-MS data. The term “chromatogram” is used to refer to either a chromatogram or mobiligram when talking generally about the algorithm because the structure and handling of the data is the same. It also includes new functions to transform the m / z axis on histogram stacks to a mass axis. Existing UniDecCD tools are repurposed to transform extracted histograms in isolation.

[0125] Finally, UniChromCD also inherits a new demultiplexing engine (FIG. 6, right). The demultiplexing engine includes tools to set up both HT and FT demultiplexing. It operates on individual TICs or EICs. However, the UniChromCD engine can also pass each pixel from the histogram stack (either the m / z or mass histogram stack) to demultiplex the entire stack.

[0126] This overall architecture allowed reuse of the same code from UniDecCD while providing a flexible platform to build new tools into it. For example, as new demultiplexing options were added, they were simply added to the demultiplexing class and did not have to make any edits to the UniChromCD class.

[0127] Data Import and Processing

[0128] UniDec can open a range of different file types, including Thermo Raw, DMT, mzML, and several text / binary formats. UniChromCD uses the same algorithm as UniDecCD for data processing and charge assignment, as previously described.37When the data is imported, a TIC is automatically created. EICs can be created by the user selecting a square or “swoop” region (FIG. 5) of the m / z vs charge histogram.

[0129] Although the demultiplexing aspects of the software are the focus of this study, UniChromCD provides a user-friendly interface for analysis and visualization of conventional time-resolved CD-MS.

[0130] HT Demultiplexing

[0131] The theory behind HT demultiplexing has been described in detail previously (Lin, et al., Anal. Chem. 2008, 80 (15), 5755-5759; Siegle, et al., Anal. Chem. 2014, 86 (21), 10828- 10833; Clowers, et al., Anal. Chem. 2006, 78 (1), 44-51; Reinecke, et al., Trends Anal. Chem. 2019, 1 16, 340-345; Naylor, et al., J. Am. Soc. Mass Spectrom. 2023, 34 (7), 1283-1294; Pallmann, S et al., Anal. Chem. 2018, 90 (14), 8445-8453). The core of the implementation used here is a convolution of the HT kernel with the multiplexed chromatogram, which uses the Fourier convolution theorem: where Y is the demultiplexed chromatogram, X is the multiplexed chromatogram, K is the HT kernel, F denotes a Fourier transform, F-1denotes an inverse Fourier transform, and F(K')* is the complex conjugate of F(JC) .

[0132] Because retention times in LC are typically much longer than the desired cycle time (the time interval between injections), it can be necessary to extend the acquisition beyond the end of the injection sequence. For convenience, this is accomplished by appending zeros to the end of the sequence so that the end of the expanded sequence occurs after the last peak has fully eluted. It is important to note that these added zeros are not used in the construction of the HT kernel and therefore do not violate the requirement (Harwit, M.; Sloane, N. J. A. The Basic Theory of Hadamard Transform Spectrometers and Imagers. Academic Press, 1979; Chapter 3, pp 44-95) that the number of zeros in the kernel PRBS be 2m-;- 1. Prior to demultiplexing, the appended zeros are moved from the end of the sequence to the beginning, which in turn shifts the original PRBS to the end of the chromatogram, where ideally it completely overlaps with the portion of the chromatogram that contains peaks. In cases where the difference in retention time between two or more peaks in the chromatogram are greater than the cycle time, it may be necessary to manually adjust this time shift during processing of chromatograms to ensure complete overlap of the HT kernel with the data. To help check alignment, the kernel can be plotted over the raw chromatograms using Tools > Plot Kernel.

[0133] Results and Discussion

[0134] HT-SEC-CD-MS Workflow

[0135] To illustrate a typical HT-SEC-CD-MS workflow, data was collected for a simple mixture of P-galactosidase and GroEL. The sample was injected according to a 5-bit sequence as described in a companion paper (Sanders, J.; et al., ChemRxiv 2024 DOI: 10.26434 / chemrxiv- 2024-h08x5). Here, the focus was on data with a 5-bit injection sequence because it produced excellent quality data in a relatively short time, but additional bit lengths are shown in that paper (Sanders, I.; et al., ChemRxiv 2024 DOI: 10.26434 / chemrxiv-2024-h08x5). When raw CD-MS data is imported for the first time, the three plots shown in FIGs. 7A-7C were automatically generated. The raw chromatogram (FIG. 7A) appeared noisy because each individual scan contained only a few ions. This low number of ions caused inherent shot noise variability between scans. Similarly, the m / z (FIG. 7B) and mass (FIG. 7C) distributions are too finely sampled and did not contain enough ion counts per bin. Setting the Scan Compression to 10, m / z Bin Size to 10, and Mass Bin Size to 20 kDa produced smoother distributions, shown in FIGs. 7D-7F.

[0136] Note, the scan compression and mass bin size parameters were independent of the others. In other words, adjusting the scan compression will only affect chromatography data (FIGs. 7A and 7D) and did not affect the total number of ions in the m / z vs charge histograms (FIGs. 7B and 7E) or the transformed mass distributions (FIGs. 7C and 7F). Changing the m / z binning affected the potential resolution of the mass bins, but neither binning parameter affected the chromatography data because the total number of ions per scan were preserved.

[0137] Next, EICs were generated by either simple range selection (generating a square selection window in m / z and charge) or “swoop” selection (generating a diagonal selection window that follows the natural m / z trend lines), as shown in FIG. 7H. After selecting the m / z distribution, UniChromCD produced the raw EICs (FIG. 7G) and extracted mass distributions of the selected ranges (FIG. 71). Currently, per-scan UniDecCD deconvolution (Kostelic, et al., Anal. Chem. 2021, 93 (44), 14722-14729) of each histogram in the stack was not implemented. However, deconvolution of the summed histogram can be performed, and extracted mass distributions can be generated from this deconvolved summed distribution (FIG. 7L).

[0138] After generating the TIC or EICs, HT demultiplexing can be applied to the chromatograms to produce demultiplexed chromatograms (FIG. 7J). Although the two species were not resolved in the demultiplexed TIC, the demultiplexed EICs showed separation with the expected elution order. The baseline illustrates the application of the Time Shift parameter discussed above. The baseline from the demultiplexed data was bracketed by flat sections that were automatically added to compensate for the zero pads used in the acquisition. Thus, the total chromatogram with included zero pads is the same length as the original data. However, there is an option in UniChromCD to define the time range to truncate the chromatograms to remove the unneeded baseline beyond when the peaks elute.

[0139] In practice, it was found that TIC and EIC demultiplexing were most natural to use. However, there is an option to perform per-pixel demultiplexing. Here, an EIC was generated for each m / z vs charge pixel in the histogram stack. Each pixel EIC was demultiplexed independently and then reassembled into a 3D demultiplexed histogram stack (FIG. 6). The same process can also be performed on a transformed mass histogram stack, which has been converted first into mass vs charge vs time. Both operations allow users to produce an EIC by selecting a mass range. Interestingly, summing each time point of the per-pixel demultiplexing perfectly reproduced the demultiplexed TIC, revealing that the total number of ions are conserved in the demultiplexing operation. It may be useful to isolate regions of 3D per-pixel demultiplexed data to simplify complex data with overlapping species, as opposed to simply selecting from the two-dimensional (2D) data with the EICs.

[0140] HT Artifact Removal

[0141] In addition to the core demultiplexing features, tools to remove HT demultiplexing artifacts were also included. These artifacts are caused by noise and imperfections between the peaks, which should ideally be perfectly repeated copies of each other. For example, if one peak is randomly larger than the others, this will cause baseline artifacts during demultiplexing. Systematic drifts, such as a gradual decrease in signal, also cause similar artifacts.

[0142] To remove these baseline artifacts, a masked multiplexing strategy was employed for HT-IM data.44This option was activated by changing the Demultiplex Mode parameter from “HT” to “mHT”. The masked multiplexing was defined by setting the number of iterations, which defines how many sets of random masks are created (data not shown), and the number of masks per iteration, which defined how many digits are masked in each set of masks. It was found that 50 iterations and 5 masks per iteration significantly cleaned up demultiplexing artifacts without distorting the primary peaks (FIG. 7K, additional data not shown). Together, these tools provide a robust and user-friendly workflow for processing HT-SEC-CD-MS data.

[0143] HT-SEC-CD-MS Sequence Shifting

[0144] Many of the PRBSs from Harwit, et al. (Harwit, M.; Sloane, N. J. A. The Basic Theory of Hadamard Transform Spectrometers and Imagers. Academic Press, 1979; Chapter 3, pp 44-95) place the largest continuous string of zeros at the beginning of the sequence. The leading zeros have the unfortunate effect of delaying the first injection by several minutes, creating unnecessary dead time at the beginning of the chromatogram. Because the Hadamard transform demultiplexing operation is circular in nature, it was contemplated that it should be possible to rotate these sequences so that the longest stretch of zeros falls at the end of the sequence. Putting the zeros at the end reduces initial dead time and reduces the number of zero pads that must be used because the last injection occurs earlier in the sequence. Therefore, the total analysis time can be shortened significantly.

[0145] To evaluate this approach, the 5-bit injection of P-galactosidase and GroEL using two modified sequences was repeated: the first rotating the sequence four positions to the left to move the first four zeros to the end of the sequence (termed “shift 4”) and the second rotating the original sequence 15 positions (“shift 15”) to split the longest stretch of ones:

[0146] Standard 5-bit: 0000100101100111110001101110101

[0147] Shift 4: 1001011001111100011011 101010000

[0148] Shift 15: 1110001101110101000010010110011

[0149] To form the HT kernel used in the demultiplexing operation, the injection sequence is modified by replacing 0s with -Is and expanded by superimposing the modified sequence onto an array of zeros the same length as the section of the chromatogram being demultiplexed (see discussion of zero padding and time shifting). Optionally, the kernel can then be smoothed by convolution with a Gaussian function, as specified by the user. Using a smoothed kernel will smooth the data during demultiplexing with minimal additional processing time, which is important in per-pixel demultiplexing.

[0150] The shift 4 chromatogram (data not shown) is like the original but with everything 4 min earlier, including the end of the acquisition. The shift 4 sequence produced a demultiplexed chromatogram (data not shown) that is nearly identical to the one produced with the original sequence (data not shown). In contrast, the demultiplexed chromatogram from the shift 15 sequence (data not shown) contained significantly more artifacts than either the original or the shift 4 sequence. These artifacts were caused by an inability to fully overlap the HT kernel and raw chromatogram. In the data produced by the original and shift 4 sequences (data not shown), the kernel completely overlapped all the peaks in the chromatogram. However, the shift 15 sequence started and ended with an injection, producing a chromatogram (data not shown) where the first and last peaks were separated by an additional 4 min compared to the other two sequences, which have four zeros at the beginning or end. Because the kernel does not fully overlap with all chromatographic peaks, not all peaks were included in the demultiplexing, which caused the artifacts (data not shown).

[0151] Although incomplete kernel overlap can be easily avoided by rotating the sequence to place the largest block of zeros at the beginning or end, it illustrated a potential limitation of HT demultiplexing method that could arise if the separation of two analytes is greater than the maximum number of consecutive zeros in the sequence times the cycle time. In these cases, a longer sequence or cycle time would be required to ensure that the kernel can overlap the sequence completely.

[0152] Conclusions

[0153] UniChromCD provides a comprehensive platform for analyzing time-resolved CD-MS data with or without multiplexing. The program allows users control of their data and the ability to customize processing for new applications as they arise. The use of UniChromCD for demultiplexing and analysis of HT-SEC-CD-MS data of a mixture of proteins was demonstrated. The embedded masked multiplexing algorithm removed artifacts from HT demultiplexed chromatograms while preserving peak fidelity. Although the focus of this study was on well- defined protein mixtures to illustrate how the software works, more complex samples of Escherichia coli lysate are shown in Example 1 demonstrating that the software performs well for complex and heterogeneous mixtures.

[0154] UniChromCD demultiplexing recovers retention information while significantly increasing CD-MS throughput. This software can be applied to any separation type coupled with CD-MS, as long as the injection strategy is HT, FT, or a single injection, which would not need demultiplexing. For HT and FT separations, an important limitation of multiplexing is that the separation needs to be isocratic. Thus, gradient solvent conditions may not work with multiplexed injections. Instead, single injections may be needed. However, adding additional demultiplexing algorithms is simple in the workflow, and it may be possible to add an algorithm to sum multiple sequential single injections by creating an HT sequence of all l's. These new computational tools provide a range of emerging applications coupling CD-MS with timedomain separations, especially to characterize complex biotherapeutics and heterogeneous protein complexes.

[0155] Those skilled in the art will recognize, or be able to ascertain using no more than routine experimentation, many equivalents to the specific embodiments of the invention described herein. Such equivalents are intended to be encompassed by the following claims.

Claims

We claim:

1. A method for analyte detection, the method comprising: delivering two or more pulsed injections of a non-gas phase sample into a liquid chromatograph operably linked to a mass spectrometer, wherein the injecting of the two or more pulses is controlled by an algorithm, preferably wherein analyte selectivity of the chromatograph is not based on (i) gas phase ion mobility separation or (ii) gas phase separation, preferably wherein the chromatograph comprises a liquid chromatograph.

2. The method of claim 1, further comprising sampling the analyte eluted from the chromatograph into the mass spectrometer.

3. The method of claim 2, further comprising generating a series of mass spectrometry scans for the sample.

4. The method of claim 3, further comprising extracting intensity over time for specific regions of the scans, optionally to produce an extracted ion chromatogram.

5. The method of claim 3 or 4, further comprising demultiplexing intensities in the scans.

6. The method of claim 5, wherein demultiplexing is performed using a Hadamard Transform, or correlation techniques (such as correlation chromatography techniques preferably comprising a multiplexed binary injection sequence).

7. The method of claim 5 or 6, further comprising reassembling multiple demultiplexed datasets produces a second dataset comprising mass, charge, and retention time of the analyte, preferably wherein the reassembling is performed using Hadamard Transform.

8. The method of any one of claims 1 to 7, wherein the chromatograph is selected from the group consisting of a liquid chromatograph (such as size exclusion chromatograph, high performance liquid chromatograph, and reverse-phase liquid chromatograph), a column chromatograph, an ion-exchange chromatograph, a gel-permeation (molecular sieve) chromatograph, an affinity chromatograph, a paper chromatograph, a thin-layer chromatograph, a dye-ligand chromatograph, a hydrophobic interaction chromatograph, a pseudo- affinity chromatograph, or a combination thereof, preferably a liquid chromatograph, such as a size exclusion chromatograph.

9. The method of any one of claims 1 to 8, wherein the mass spectrometer uses a single particle technique, wherein masses of individual ions are determined from simultaneous measurements of their mass-to-charge ratio (m / z) and charge.

10. The method of any one of claims 1 to 9, wherein the mass spectrometer comprises a charge detection mass spectrometer, preferably an Orbitrap mass spectrometer.

11. The method of any one of claims 1 to 10, wherein the sample comprises an analyte selected from proteins; lipoproteins; glycosylated proteins; protein complexes; protein aggregates such as amyloid fibers; infectious viruses; non-infectious viruses; nanoparticles; glycans; gene therapies; vaccines; vesicles such as exosomes; nucleic acids; or a combination thereof.

12. A device configured with an algorithm to deliver two or more pulsed injections of a sample into a chromatograph, optionally wherein the sample comprises an analyte to be detected.

13. The device of claim 12, operably linked to an injection valve.

14. The device of claim 12 or 13, operably linked to a mass spectrometer.

15. The device of any one of claims 12 to 14, wherein the chromatograph is selected from the group consisting of a liquid chromatograph (such as size exclusion chromatograph, high performance liquid chromatograph, and reverse-phase liquid chromatograph), a column chromatograph, an ion-exchange chromatograph, a gel-permeation (molecular sieve) chromatograph, an affinity chromatograph, a hydrophobic interaction chromatograph, a pseudoaffinity chromatograph, or a combination thereof, preferably a liquid chromatograph, such as a size exclusion chromatograph.

16. The device of claim 14 or 15, wherein the mass spectrometer uses a single particle technique, wherein masses of individual ions are determined from simultaneous measurements of their mass-to-charge ratio (m / z) and charge.

17. The device of any one of claims 14 to 16, wherein the mass spectrometer comprises a charge detection mass spectrometer, preferably an Orbitrap mass spectrometer.

18. The device of any one of claims 12 to 17, wherein a binary sequence of bits controls delivery of the sample, preferably wherein the binary sequence is a pseudo-random binary sequence.

19. The device of claim 18, wherein the binary sequence of bits controls injections, duration of each injection (injection time), delay time between injections (cycle time), and a combination thereof.

20. The device of any one of claims 12 to 19, wherein the device is a microcontroller.

21. The device of any one of claims 12 to 20, wherein the sample is a non-gas phase sample.

Images