Method and computer-implemented system for spectral analysis of a liquid substance mixture

Abstract

In a method for spectral analysis of a liquid substance mixture consisting of pre-known individual substances, a reference spectrum for the substance mixture is provided, wherein the reference spectrum represents a target state of the substance mixture in which each individual substance is contained in a target concentration. An actual spectrum of the liquid substance mixture to be analysed is created. Finally, concentrations of the individual substances in the substance mixture are determined on the basis of an approximate spectral composition of the actual spectrum from pre-known pure spectra of the pre-known individual substances and at least one compensation spectrum.

Description

[0001] Method and computer-implemented system for spectral analysis of a liquid mixture of substances

[0002] The invention relates to a method for the spectral analysis of a liquid mixture of substances and a computer-implemented system for a spectral analysis of a liquid mixture of substances.

[0003] The analysis of liquid mixtures, such as mixtures of liquid substances consisting of several individual substances, using vibrational spectroscopy, especially mid-infrared (midIR) and Raman spectroscopy, presents a significant challenge. Even small molecules exhibit several to many individual absorption bands and thus display a substance-specific spectrum. This is due to the width of the absorption bands in liquids, which encompass multiple wavenumbers and typically have a full width at half maximum (FWHM) of approximately 5 to 50 cm'. 1In the mid-infrared range, the individual bands typically overlap significantly. The mid-infrared range typically extends from 4000 to 400 cm⁻¹. 1 In certain wavenumber regions, such as the wavenumber region from approximately 1800 to approximately 900 cm⁻¹ 1 or approximately 3050 to approximately 2800 cm' 1 Many substances exhibit a continuous superposition of bands. Despite absorptions in the same wavenumber range, resulting in a certain overlap between the substance spectra, also known as the AOO (Area of ​​Overlap), substances nevertheless possess a unique spectrum, allowing them to be identified by their spectral pattern, also called their fingerprint.

[0004] Furthermore, almost all substances exhibit an IR spectrum. IR spectra of substance mixtures are therefore usually complex, as the spectra of the individual substances overlap to form a complete spectrum. An interpretation of the composition based on individual, non-overlapping absorption bands is generally not possible due to the high band overlap between the individual substance spectra. This complexity is further increased by the fact that the molecules, and thus their spectra, can be significantly influenced by various environmental factors, such as:

[0005] PH value,

[0006] ion concentrations,

[0007] Temperature,

[0008] Interactions with ingredients,

[0009] Solvent interaction

[0010] and / or

[0011] Redox potential.

[0012] These influences not only lead to shifts in bands in the spectrum, but can also cause new bands to appear or existing bands to disappear, for example through complex formation of individual substances or through conformational changes of at least one of the individual substances.

[0013] Therefore, the analysis of a liquid mixture and / or formulation generally requires a complex calibration process that accounts for these influences. This involves examining and measuring calibration samples of the mixture, which define a calibration chamber. The number of calibration samples depends on the number of individual substances and / or ingredients relevant to the spectrum within the mixture and / or formulation. The number of calibration samples can be determined, for example, by... nEstimate the number of relevant individual substances and / or ingredients. This suggests that even a formulation and / or mixture with only a few individual substances and / or ingredients requires a considerable number of calibration samples. Therefore, performing calibration as required for conventional IR-based analyses is often very time-consuming and prone to preparative errors.

[0014] Furthermore, the analytical models created through calibration are only valid for the specific calibration space used. Even if only a single influencing parameter changes within the calibration space, creating a new or expanded calibration space is often necessary. This typically means that each formulation and / or mixture of substances conventionally requires its own Design of Experiment (DoE) and its own individual analytical model. Due to the numerous possible combinations of relevant ingredients and the time-consuming nature of the respective calibrations, this approach is often economically unviable in terms of time, cost, and resources.

[0015] The invention is therefore based on the objective of enabling improved spectral analysis of a liquid mixture of substances. Furthermore, the invention may aim to reduce the calibration effort required for spectral analysis and / or to reduce the susceptibility to errors in spectral analysis.

[0016] The problem is solved by the subject matter of the independent claims. The dependent claims relate to preferred embodiments.

[0017] One aspect concerns a method for the spectral analysis of a liquid mixture of substances consisting of previously known individual substances, comprising the following steps:

[0018] - Providing a reference spectrum for the substance mixture, wherein the reference spectrum represents a target state of the substance mixture in which each individual substance is contained in a target concentration;

[0019] - Creating an actual spectrum of the liquid substance mixture to be analyzed; and

[0020] - Determining the concentrations of the individual substances in the substance mixture based on an approximate spectral composition of the actual spectrum from previously known pure spectra of the previously known individual substances and at least one compensation spectrum.

[0021] The method can be at least partially computer-implemented. As described earlier, the spectral analysis can be based on vibrational spectroscopy, in particular medium-infrared (midIR) and / or Raman spectroscopy. A spectrum of the substance mixture is recorded, specifically an IR spectrum. Therefore, the reference spectrum, the actual spectrum, and the pure spectra can all be IR spectra.

[0022] Within the scope of the invention, the terms "liquid mixture of substances," "mixture of substances," and "formulation" are used synonymously. Likewise, the terms "individual substance," "ingredient," and "component" can be used synonymously. Individual substances can refer to all substances comprising the mixture of substances that make a relevant contribution to the spectrum of the mixture. Individual substances that have no effect or only a negligible effect on the spectrum, particularly below the resolution of the measuring instrument used, can be disregarded.

[0023] Furthermore, the terms "spectrum" and "spectral pattern" can be used synonymously.

[0024] The liquid mixture of substances can be in the form of a liquid with ingredients, in particular as a solution such as a buffer, in which the individual substances are arranged, e.g. in dissolved form.

[0025] The substance mixture consists of known individual substances. This mixture could be, for example, a food product such as an energy drink, a pharmaceutical product such as a drug, or another industrial product that is intended to have a known composition of individual substances. This known composition corresponds to the target state of the substance mixture. In the target state, the known individual substances are present in the substance mixture at a known and / or predetermined target concentration. For some products, such as drugs, not only the concentrations of the individual substances are crucial for product quality, but also their states, such as conformational states. For such products, the target state can include not only the target concentrations of the individual substances, but also their target states.

[0026] This method can be used to determine how much a substance mixture currently being analyzed, for example, a sample from a production batch, deviates from the known target state of the substance mixture. The method can therefore be used in particular for quality control and / or quality assurance and / or within the framework of a stability study.

[0027] First, a reference spectrum for the substance mixture in its target state is prepared. The reference spectrum is preferably generated in the same manner as the actual spectrum of the substance mixture to be analyzed, recorded during spectral analysis, i.e., as a midlR spectrum. It is also preferably recorded with the same or at least an identical measuring instrument, e.g., to reduce the spectral influences of the instrument. When generating the reference spectrum, a spectrum of a reference substance mixture can be created, in which the individual substances are present in well-known concentrations and, optionally, well-known states.

[0028] The reference spectrum can be recorded and stored as a reference and / or as a reference point for further spectral analysis, in particular electronically on a storage medium.

[0029] To analyze the actual substance mixture present and to be analyzed, the actual spectrum of the liquid substance mixture to be analyzed is recorded, preferably in the same way as the reference spectrum, i.e., e.g., as a midlR spectrum, preferably over a wavelength range that is at least partially, largely, or completely overlapping. This allows for improved comparability of the actual spectrum with the reference spectrum.

[0030] Finally, the concentrations of the individual substances in the mixture are determined from the reference spectrum and the actual spectrum. For this purpose, the actual spectrum is considered as a spectral composition of the individual substances in pure form, with their respective concentrations and the equalization spectrum. To account for the individual substances in pure form, the corresponding known pure spectrum is considered for each individual substance. The pure spectrum of each individual substance is the spectrum that results from recording the individual substance in pure form, i.e., recorded separately from the mixture.

[0031] The pure spectrum of a single substance can be represented by one or more spectra of that substance, for example, from sub-state spectra to describe different states of the single substance within a mixture. These sub-state spectra are described in more detail below. The pure spectra can relate to the states of the individual substances as they exist within the mixture, such as concentration and / or pH value. This allows for the description and / or consideration of individual substances that do not exist in a single pure form within the mixture, but rather in a state that results from a mixture of different forms of the individual substance.

[0032] Thus, the evaluation in this procedure is based not only on the reference spectrum, but also on the pure spectra of the individual substances contained in the substance mixture.

[0033] The compensation spectrum allows the effects of interactions between individual substances and / or other variable states of individual substances, which are not separately considered in a compensation condition (see below), to be taken into account in the substance mixture.

[0034] Here, the actual spectrum of the substance mixture is described using the pure spectra of the individual substances it contains, which are present in their pure form. This method can be performed without additional calibration samples, thus eliminating the need for the numerous calibration samples of the substance mixture that are conventionally required and described earlier. This reduces the calibration effort, even with a large number of ingredients. It also reduces the method's susceptibility to errors.

[0035] For example, if one substance in a mixture of n individual substances is replaced, conventional recalibration methods may result in, for example, 2 n New calibration samples may be necessary for measurement in order to reconfigure the calibration chamber in which the spectral analysis is performed. In the inventive method, it may be sufficient to simply provide a new reference spectrum for the new substance mixture and to replace the pure spectrum of the modified individual substance with the pure spectrum of the new individual substance. A similar approach applies if, for example, an individual substance is added to or removed from the substance mixture.

[0036] This significantly reduces the calibration effort, especially when the substance mixture is changed.

[0037] The mixture of substances can be approximately spectrally composed and mathematically described as follows:

[0038] n comp

[0039] A jDnMix > — ^ jön I M nit x eraction -f,- V / -1 C; Mix AyiBn; Pure

[0040]

[0041] i = l

[0042] The following are listed:

[0043] - AWAY Mix for the (total) spectrum of the substance mixture; AB Mix Interaction for the compensation spectrum;

[0044] -

[0045]

[0046] AWAY i Pure for the respective pure spectrum of the individual substance i;

[0047] - c i Mix for the respective concentration of the individual substance i in the substance mixture;

[0048] and

[0049] n comp for the number of (relevant) individual substances in the substance mixture.

[0050] This spectral composition can be used for both the reference spectrum and the actually recorded spectrum, although the actual concentrations may differ slightly from the target concentrations.

[0051] In general, for each relevant individual substance in the mixture, at least one or exactly one pure spectrum can be considered in the approximation.

[0052] The preceding equation for the composition of the substance mixture can contain at least linear terms, i.e., it can be formed exactly as shown above. In many embodiments, this linear equation for the composition of the substance mixture may already be sufficient to solve the problem. However, in addition to the linear terms shown above, the equation can also contain nonlinear terms to describe the composition of the substance mixture even more precisely. In most applications, however, the contribution of the nonlinear terms is significantly smaller than that of the linear terms. Therefore, the nonlinear terms are negligible in several embodiments. In some embodiments, however, these nonlinear terms are also taken into account when solving the equation.

[0053] It can be advantageous to have very good instrumental interoperability between the measuring instruments and / or analyzer used for spectral analysis, so that the spectrum of a substance is largely independent of the individual instrument used, meaning that almost identical spectrums can be generated for this substance on different measuring instruments. In one embodiment, the compensating spectrum represents interaction effects of the individual substances and / or environmental influences in the substance mixture. Knowledge of these interaction effects and / or environmental influences in the substance mixture can increase the accuracy of the spectral analysis. The compensating spectrum can be composed of different interaction effects, for example, it can itself be a sum spectrum of different interaction spectra, which can also be called interaction spectra. Depending on their relevance, the interaction spectra can be scaled, e.g.,by means of their respective "interaction concentrations".

[0054] In one embodiment, the compensation spectrum is determined by spectral subtraction of the pure spectra of the individual substances at their respective target concentrations from the reference spectrum. In other words, it is assumed that the compensation spectrum in the actual spectrum essentially corresponds to the compensation spectrum in the reference spectrum. Thus, the compensation spectrum can be determined from the reference spectrum, the target concentrations, and the pure spectra of the individual substances by spectral subtraction, particularly by the following description, which is based on the specific case of the target state, i.e., the reference point, for the substance mixture:

[0055] n comp

[0056] XßMix Re / = + V Mixture

[0057] Interaction / , 1 1

[0058]

[0059] i = l

[0060] Analogous to the above case, the following apply here:

[0061] - AWAY Mix for the reference spectrum of the substance mixture;

[0062] - B. I M nt' X e R ra e / c,ti.on for the output c»equalization spectrum of the substance mixture in the target state;

[0063] - AWAY i Pure for the respective pure spectrum of the individual substance i;

[0064] - c i Mix for the respective target concentration of the individual substance i in the substance mixture in the target state; and

[0065] n comp for the number of individual substances in the substance mixture.

[0066] For example, the compensation spectrum can be determined using the preceding description.

[0067]

[0068] AWAY Mix Interaction of the substance mixture in the target state from the reference spectrum AB Mix , the target concentrations c i Mixand the pure spectra AB i Pure the individual substances i are determined.

[0069] The compensation spectrum X

[0070]

[0071] ß. M ' XRe / The spectrum of the substance mixture in its target state can be approximated as a compensation spectrum β. MiXs « m the substance mixture in its actual existing state.

[0072] This approximation can be based on the consideration that environmental influences in a mixture of substances typically lead to spectral deviations of only a few percent compared to the pure states of the contained substances. Thus, the actual substance spectra in the mixture largely correspond to the individual spectra of the individual substances in their pure form.

[0073] Nevertheless, environmental influences can significantly affect the determination of components that constitute only a small proportion of the total mixture. For example, if a mixture of substances consisting of 50 g / L of substance 1, 50 g / L of substance 2, and 1 g / L of substance 3 were present, and substances 1 and 2 interacted, resulting in approximately 1% spectral modification to their respective pure spectra, then the spectral area of ​​the interaction spectrum of these two substances would correspond to a generally unknown interfering potential with a concentration equivalent of approximately 1 g / L, roughly the same order of magnitude as substance 3. Consequently, the determination of substance 3 could be significantly affected if the interfering potential spectrum is unknown, potentially leading to a large analytical error. The degree of influence depends primarily on the extent to which the spectrum of substance 3 and the interfering potential spectrum overlap.The larger the AOO value, the greater the potential impact.

[0074] The balancing spectrum can also include and / or take into account theoretically determined components. For example, the magnitude of simple molecules such as hydrogen molecules can be calculated quantum mechanically and incorporated into the balancing spectrum. The balancing spectrum can therefore contain both experimentally determined and theoretically determined components.

[0075] As explained above, conventional analytical methods require calibration in which substance mixtures defined by the Design of Experiments (DoE) define a calibration space. The samples of the substance mixtures to be analyzed must conventionally be located within the calibration space to be accurately determined. This means that conventional interpolation into the measurement space is performed.

[0076] This can be illustrated, for example, with a three-component mixture, where the calibration space extends over, for example, 2 3 Eight calibration samples are set up. These eight samples represent the vertices of a cube in a three-dimensional concentration chamber (each component lies on an axis). The samples of substance mixtures to be analyzed must lie within the "cube" of the calibration chamber in terms of their constituents and concentrations.

[0077] In contrast, the method according to the invention does not define a typical calibration space, but rather uses a reference point, namely the reference spectrum of the substance mixture in the target state. In some embodiments, which will be described in more detail below, several reference points can also be used, e.g., in the form of mixture spectra in known states of the substance mixture, which are located as close as possible to the most probable states of the substance mixtures (more on this below).

[0078] Deviations of the actual state of the sample being analyzed—that is, the actual mixture of substances present—from the most probable state, i.e., the target state as the reference point, can be extrapolated using the pure spectra of the substances contained in the mixture and, if necessary, further compensation spectra. The more closely the actual state of a sample corresponds to the reference point or one of the reference points, the more closely all environmental influences and interactions inherent in the reference point are reflected in this sample. Assuming that the interactions of the constituents in the reference sample are equal to (or nearly equal to) those in the actual sample, these can be compensated for, for example, by scaled spectral subtraction, and the concentration differences of the constituents between the reference and the sample can be described solely by the pure spectra.Since in this approach the reference point essentially describes the state of the current sample, the pure spectra of the ingredients, which cannot reflect the environmental state in the sample and therefore involve a slight inaccuracy in the determination, are only needed to a small extent.

[0079] Analogous to the two preceding equations, the following description results for the actual substance mixture present as a sample:

[0080] n comp ^ MiXS “ m = ^Cera7tion + £ C^ B^

[0081]

[0082] i = l

[0083] The following is analogous to the above:

[0084] - ß MiXs “ m for the actual spectrum of the substance mixture to be analyzed;

[0085] - ^interaction for the compensation spectrum of the substance mixture to be analyzed;

[0086] - AWAY iPure for the respective pure spectrum of the individual substance i;

[0087] > c Mix sam for r

[0088]

[0089] zu erm jtt e | nc | e Concentration of the individual substance i in the mixture of substances to be analyzed; and

[0090] n comp for the number of individual substances in the substance mixture.

[0091] By differentiating the description of the reference spectrum B MiXRe / (so) and the actual spectrum s MiXs “ m surrendered:

[0092] n comp

[0093] XßMix Sam _ Xß Mix Äe / = 2JXB^ x raction + Act AB? ure

[0094]

[0095] i = l Here, the following stands:

[0096] - AABjnieraction for the difference compensation spectrum between the substance mixture to be analyzed and the target state of the substance mixture; and

[0097] - Act for the difference concentration of the individual substance i in the substance mixture to be analyzed and in the target state of the substance mixture;

[0098] Assuming that the compensation spectra in the target state and in the substance mixture to be analyzed are essentially the same, i.e., that the following applies:

[0099] A / I pMix n

[0100] ^■^^Interaction u

[0101] This results in:

[0102] n comp

[0103] ^ßMixsam _ AB M iXRe f =

[0104]

[0105] Thus, the following condition results, which is also referred to below as the compensation condition:

[0106] n comp

[0107] AB^sam - AB^Ref > Ac t AWAY? uve= 0 Equilibrium condition

[0108]

[0109] The difference concentrations Ac can be determined from the equilibrium condition. t the individual substances / are determined, and from these the individual concentrations are derived. c 1Xsam in the actual mixture of substances to be analyzed:

[0110] Mix sam = c Mix Äe / +

[0111]

[0112] l

[0113] The equation for the equilibrium condition can be solved, for example, using a least-squares fit to determine the concentrations of the ingredients in the sample. Alternatively, the equation can also be solved using other approximations, such as at least one trained model and / or neural network. Particularly if nonlinear terms are considered in addition to linear ones, the equation can often be solved more reliably using a trained model, such as a neural network. A deep learning approach can also be used. However, if the equation is formulated purely linearly, a solution using a simpler method, such as a least-squares fit, is sufficient in many cases.

[0114] In the example above, if a reference point were used for the mixture of substances 1, 2, and 3, and the sample mixture to be analyzed deviated in its composition by, for example, 1% from substances 1 and 2, then the compensation spectrum of substances 1 and 2 would also be very similar to that of the reference point. In summary, the effects would almost completely cancel each other out, resulting in a significantly lower interference potential in the determination of substance 3. If one were to assume an approximate spectral deviation of 1% between the compensation spectra of the reference and sample, proportional to the concentration difference of the substances in the reference and sample, the resulting interference potential would also be only about 1% as large as the actual compensation spectrum.In the example above, the interference potential (expressed in concentration equivalents) for substance 3, calculated using the approximate assumption, would be only about 1 g / L * 0.01, or 0.01 g / L. Typically, the AOO values ​​of midlR spectra are significantly less than 100%, so the effective contribution of the interference potential error, even for substance 3, can be estimated at < 1%, which is below the several percent typical for achievable analytical errors. Therefore, the assumption made above that the compensation spectra are practically identical in the target state and for the substance mixture being analyzed leads only to an interference potential error that is usually smaller than the instrument's measurement accuracy and thus negligible.

[0115] The concentrations of individual substances determined by this method are more accurate the less the substance mixture being analyzed deviates from the target state. For example, the method can be used to analyze substance mixtures where the actual concentrations of the individual substances deviate by a maximum of approximately 50% from the respective target concentrations of the individual components, preferably by a maximum of approximately 20%, and particularly preferably by a maximum of approximately 10%. With such small deviations from the target state, determining the concentrations generally leads to a very reliable result.

[0116] The deviation can be a deviation from the respective target concentration. Alternatively, the deviation can also be an extrapolation from a range of target concentrations if more than one reference point is used for the procedure. In this case, each reference point can encompass target concentrations of the individual components specific to that reference point.

[0117] For some substance mixtures, even larger deviations from the target concentrations can be determined with sufficient accuracy, e.g., for some proteins, deviations of up to 100% are possible.

[0118] In one embodiment, the previously known pure spectra are stored in a spectral database and are retrieved from the database before determining the concentrations of the individual substances. The spectral database, which can also be abbreviated as SDS, can contain the pure spectra of several individual substances, particularly specific to predetermined buffers and / or typical pharmaceutical formulation components, and can be made available to different users of the method. In particular, the spectral database can be maintained and expanded by an instrument manufacturer and / or analysis software provider. Thus, the spectral database can be expanded and / or updated to include pure spectra of an increasing number of individual substances.This allows users to easily retrieve the relevant pure spectra of the modified individual substances from the spectral database and integrate them into their analytical procedure when the substance mixture is changed, without having to generate the pure spectrum themselves. The spectral database can be hosted on the internet, enabling access for different users from different locations. Each pure spectrum can then be used to describe and analyze many different formulations, further reducing the user's calibration effort.

[0119] The embodiment thus relates to a spectrum database-supported evaluation from which quantitative statements about the substance mixture can be derived. Unlike conventional methods, it is not necessary to measure a buffer beforehand on the same instrument and then subtract it during the analysis. For calibration measurements, e.g., to determine the target state, the entire formulation can initially be measured without a buffer.

[0120] Alternatively, the target state can be determined as a reference spectrum by measuring the entire formulation, including the buffer. If pure spectra of certain individual substances are not available, the complementary mixture of the formulation without the respective individual substance can be measured instead.

[0121] In one example, a formulation consists of an antibody and a buffer composed of several individual substances. A reference spectrum of the formulation can usually be generated, and the pure spectra of the antibody as well as all components of the buffer can be provided. If the pure antibody spectrum is unavailable, this can be compensated for by generating another reference spectrum that corresponds to the composition of the formulation without the antibody, i.e., a complementary antibody spectrum. In this case, a sample to be analyzed can be described by the reference spectrum of the entire formulation, the pure spectra of all components excluding the antibody, and the reference spectrum of the formulation without the antibody. The resulting missing spectrum of the pure antibody can, for example, be self-adjusted using a least-squares fit.The difference formulation minus the formulation without the antibody yields an approximate spectrum of the antibody in its pure form. This difference calculation can, but does not have to, be performed in a separate step to obtain the pure antibody spectrum. The fit calculates the antibody's latent spectrum using a least-squares fit.

[0122] If certain individual substances are not available separately when creating the reference spectra, the formulation mixture complementary to that individual substance can be measured. If two ingredients are not available in pure form, a formulation mixture complementary to each of the missing substances can also be measured. In this case, two complementary formulation mixtures are used as an alternative to the pure spectra of the two ingredients.

[0123] This method allows for the consideration of relevant environmental influences on liquid substance mixtures based on a spectral database-supported analysis strategy. This enables accurate analysis for all ingredients without the need for complex calibrations for the respective formulations.

[0124] In a training course, the pure spectra stored in the spectral database are generated for the solvent in which the substance mixture is dissolved. Solvent interactions with the solvent can be taken into account. The pure spectra stored in the spectral database can be generated for the specific solvent of the substance mixture being analyzed. This means that for an analysis of a substance mixture where water is the solvent, the pure spectra of all individual substances contained in the mixture are also recorded in aqueous medium. Solvent effects, such as the hydration of molecules, are thus represented in the safety data sheet (SDS) for the respective individual substances in almost the same way as they are present in the substance mixture. This further improves the accuracy of the analysis, as solvent interactions with the solvent are also considered.In one embodiment, when determining the concentrations of the individual substances in the mixture, deviations of the concentrations of the individual substances from their respective target concentrations are determined. The absolute concentrations can then be calculated from these deviations, e.g., using the equation already given above.

[0125] c Mix Sam _

[0126]

[0127] + Z1C

[0128] Alternatively or additionally, the deviations Ac can also be used. tThe results can be evaluated immediately, e.g., as part of quality control and / or quality assurance. If the deviations are sufficiently small, e.g., within a tolerance range and / or tolerance space, the quality control can be classified as passed and / or sufficient. This implementation can be used, for example, in the context of drug testing, e.g., to check the quality of a drug batch using at least one sample from the batch.

[0129] One embodiment of the method relates to the spectral analysis of the substance mixture in which at least one of the individual substances in the mixture assumes different forms, in particular different protonation forms and / or different redox states and / or different conformational states. When determining the concentration, the following are used:

[0130] the different forms of the individual substance are taken into account as sub-substances, and the sub-concentrations of the sub-substances are determined using previously known sub-substance pure spectra of the sub-substances;

[0131] and / or

[0132] several previously known sub-state spectra are taken into account, in which the individual substance exists in different previously known states and thus in differently proportioned forms; and / or

[0133] A pure spectrum is used for this individual substance, which is designed as a sub-model that describes the state of the individual substance in the mixture of substances via latent quantities, in particular its equilibrium and / or its protonation state and / or its redox state and / or its conformational state;

[0134] and / or

[0135] For this single substance, several sub-state spectra are used, each of which is designed as a sub-model, so that the sub-models together can describe the respective state of the single substance in the mixture of substances via their respective latent quantities, in particular their equilibrium and / or their protonation state and / or their redox state and / or their conformational state.

[0136] The individual substance, which is referred to below as a variable individual substance, can exist in different forms within the mixture, particularly simultaneously. This means that the individual substance can, for example, exist partly as a positively charged molecule and partly as a neutral molecule. The positively charged molecules can be considered the first sub-substance, and the neutral molecules the second sub-substance.

[0137] In a mixture of substances, a substance can exist in a certain state, which is composed of different forms. For example, in the case of an acid, the state can be 50% protonated and 50% deprotonated, where the purely protonated and the purely deprotonated molecule each represent a (protonation) form.

[0138] The protonation state of a single substance can therefore be composed of the mixture ratio of two protonation forms of that single substance. These forms are also referred to as sub-synthetic pure substances of the mixture, with each form exhibiting its own associated sub-synthetic pure spectrum. A sub-synthetic can thus be, for example, a pure protonation form of a single substance; for instance, H₂PO₄' is a protonation form of phosphoric acid and therefore a sub-synthetic. H₂PO₄' does not exist as a single substance in aqueous solution at any pH; rather, the other protonation forms, such as PO₄, are also present.3 ', HPO4 2 ' and H3PO4 are also present in proportions. In general, a protonation state is thus a mixture of several sub-substances (protonation forms) that occupy a specific mixing ratio to each other at a given pH. Spectral database entries can, for example, directly represent protonation states (as mixtures of protonation forms).

[0139] The single substance in a given state can be composed of at least a first component (i.e., a first subconcentration) of a first form and a second component (i.e., a second subconcentration) of a second form. In its pure form, the single substance can also exist entirely in a single form.

[0140] The sub-synthetic spectra cannot necessarily be measured in their pure form, i.e., in a pure protonation form, as illustrated by the example of phosphoric acid above. Therefore, the pure sub-synthetic spectra may not be known beforehand. Rather, different proportions of the various protonation forms, i.e., different protonation states, can be measured at different pH values. Mathematically, theoretical sub-synthetic spectra can then be calculated for each protonation form, which cannot be directly measured.

[0141] Thus, the sub-individual pure spectrum can, for example, be formed as a theoretically calculated pure spectrum of a protonation form. The sub-individual pure concentration can, for example, be formed as a concentration of a protonation form.

[0142] The state, in particular the protonation state, of a single substance results as a mixture of several forms of the single substance, i.e., e.g., as a mixture of several protonation forms.

[0143] A sub-state concentration can be the sum concentration of all partial forms of a state, particularly a protonation state. The sub-state concentration can thus result from the sum of all the pure sub-state concentrations.

[0144] The sub-state spectrum can be formed as the spectrum of a single substance in a specific state. Thus, the sub-state spectrum can, for example, be composed partly of sub-state pure spectra.

[0145] In the equilibrium condition, either the sub-individual pure spectra of the corresponding pure forms, or sub-individual state spectra of corresponding states, or both, can be considered for a single substance. If the sub-individual state spectra are considered, then instead of sub-concentrations, the corresponding sub-individual state concentrations can be taken into account, i.e., the concentrations of the proportionally present states.

[0146] Thus, for example, not just a single pure spectrum is used for the variable single substance, but several sub-single pure spectra are used, in particular a dedicated sub-single pure spectrum for each different form of the variable single substance. Or several sub-state spectra are used, in particular a dedicated sub-state spectrum for each different state of the variable single substance considered. Instead of a single pure spectrum, the several sub-single pure spectra are considered for the variable single substance in the spectral composition, e.g., in the balancing condition, in particular with their associated sub-single pure concentrations, or several sub-state spectra are considered, in particular with their associated sub-state concentrations.Particularly if measuring sub-state spectra is difficult and / or impossible, the embodiment with sub-state spectra may be preferred. This makes it possible to depict interaction effects of a single substance with itself when multiple protonation states are present, at a desired level of detail. For example, acetic acid has two protonation states, so only two sub-state spectra are available. An interaction effect that occurs in a 50 / 50 mixture between the two protonation states cannot be described by the sub-state spectra. However, if any number of sub-state spectra of a single substance are stored, for example, in a spectral database, one can select, for instance, the states that are closest to the 50 / 50 state. For example, the 53 / 47 state and the 45 / 55 state.The corresponding sub-state spectra of these two closely related states already intrinsically reflect the interaction effects of the protonation modes in the respective state, which is very similar to the state at 50 / 50. The slight deviations here can potentially even be neglected.

[0147] For a single substance, a sub-state spectrum can be systematically recorded at constant, predetermined intervals, e.g., every 5% change in state. Each of these sub-state spectra can be stored in a spectral database. From all these sub-state spectra, a partial model can be created as a spectral generator, which can be interpolated with sufficient accuracy to any desired state of the single substance. This allows the interaction effects present in this interpolated state to be adequately described.

[0148] From the sub-state spectra, a state interaction spectrum can be calculated depending on the current state of the individual substance.

[0149] A state interaction spectrum of a variable single substance allows the interaction and / or equilibrium state between the different states of the variable single substance to be taken into account. This can significantly improve the accuracy of the spectral composition, especially the equilibrium condition, and / or the spectral analysis, as known properties of the variable single substance are better considered.

[0150] In this embodiment, determining the concentrations allows not only the concentrations of the individual substances but also the sub-concentrations of the sub-substances of the variable individual substances. This also enables conclusions to be drawn about the state of the variable individual substance, thereby increasing the informative value of the spectral analysis.

[0151] In other words, the variable single substance in pure form can be described, for example, by its forms and corresponding sub-concentrations of the forms, i.e., its protonation forms and corresponding fraction coefficients of the protonation forms.

[0152] The variable single substance in its pure form can also be described by at least two states, e.g., two protonation states, as depicted, for example, in a spectral database, and corresponding sub-state concentrations of those states. A first protonation state might, for example, be 41% protonated and 59% deprotonated, and a second, different state, for example, 55% protonated and 45% deprotonated.

[0153] Furthermore, the variable individual substance in pure form can be described by a protonation state and a corresponding sub-state concentration, which can be optimally adapted to the degree of protonation of the acid in the substance mixture via a partial model. For this purpose, the spectral database entries that depict the various protonation states of the acid in pure form can be used to achieve a desired state, e.g.,

[0154] to interpolate 50% / 50%. Finally, the variable single substance in pure form can be described by two protonation states and corresponding fraction coefficients (i.e., sub-state concentrations), each interpolated from a sub-model (see preceding paragraph). In one example, a first protonation state is 45% protonated and 55% deprotonated, and a second protonation state is 55% protonated and 45% deprotonated. Using this approach, a protonation state of an acid in the substance mixture that is not known exactly, but only approximately, can be taken into account.

[0155] The explanations regarding states and forms were illustrated using protonation examples. These explanations also apply analogously to redox processes and other conceivable processes in substances listed above.

[0156] The following are different examples of how to consider variable individual substances that can assume different states in the substance mixture.

[0157] pH-dependent individual substances

[0158] In mixtures of substances in which at least one variable individual substance can undergo an acid-base reaction, e.g., protolysis, the acid-base reaction and / or the acid / base pairs of the respective individual substance can be represented, in particular, completely represented. For example, in aqueous media, this means that the at least one variable individual substance is represented in both its protonated and deprotonated forms, or in one or more protonation intermediates, when determining concentrations, especially in the spectral database. Different proportions of dependent acid / base pairs can be described spectrally.

[0159] For example, acetic acid, which can exist in predominantly protonated or deprotonated forms depending on the pH, can be fully described for any desired degree of protolysis by the presence of the protonated and deprotonated states in the spectral database and / or by the presence of several different sub-state spectra of multiple states of acetic acid and / or by a suitable sub-model. A suitable description could, for example, be achieved via a least-squares fit, in which at least two arbitrary but distinct protonation states of acetic acid from the spectral database or the sub-model are used to describe any third, usually unknown, state.

[0160] In general, a sub-spectrum for the profaned state as well as a sub-spectrum for the deprotonated state of the pH-dependent variable single substance can be considered. Alternatively, at least two or more different sub-state spectra from at least two or more different states (i.e., different degrees of protolysis) of the variable single substance can be considered.

[0161] In this method, at least as many or more sub-state spectra, in particular SDS entries, can be used to fully describe acid / base-induced equilibrium settings as there are independent acid / base-induced effects that can occur in the respective state of the variable individual substance in the substance mixture.

[0162] For citric acid as a variable single substance with three pKa values ​​and thus four different protonation forms of citric acid, at least four different sub-state spectra, especially entries from the SDS, can be used to fully describe the state of citric acid at any arbitrary or unknown degree of protolysis in its spectral composition.

[0163] In general, for pH-dependent single substances, preferably but not necessarily at least as many different sub-spectra, especially SDS entries, can be used as the number of different protonation forms the pH-dependent single substance can assume.

[0164] After solving the spectral composition, i.e., the equilibrium condition and especially after determining the concentrations and sub-concentrations, predictions can also be made and / or output regarding the pH value and / or the protonation capacity in the substance mixture.

[0165] From the safety data sheets (SDS) entries for the respective individual substances and sub-substances, sub-models can be developed that describe the equilibrium state of the variable individual substance via latent quantities. This can be done, for example, using principal component analysis (PCA) or principal component regression (PCR) loadings. The number of necessary loadings does not necessarily have to match or exceed the number of independent, acid / base-related effects. Sub-models can pre-interpolate and / or extrapolate the state of an individual substance in a mixture and thus incorporate only one or more optimally approximated pure spectra of the individual substance into the description of the overall spectrum of the mixture, e.g., into the equilibrium condition.

[0166] Preferably, the interaction effects of the various sub-substances of the respective single substance are approximated almost completely, particularly using spectral database entries or sub-models, so that all protonation states and ratios of the variable single substance can be described almost without error and / or small errors are significantly below the detection limit of the measuring instrument used. This can be described for the variable single substance i as follows:

[0167] "pH

[0168] oPure > V pH Pure pH pH

[0169] ~ ' c i,x n D \,x + Interaction

[0170]

[0171] X=1

[0172] This states: - ßf ure for the respective pure spectrum of the individual substance i;

[0173] - cf" for the sub-state concentration of state x to be determined in the substance mixture to be analyzed, where state x describes a protonation state of the individual substance i;

[0174] - ß P x urepH for the sub-state spectrum of state x of the individual substance i in the substance mixture to be analyzed;

[0175] n pH for the number of protonation states of the single substance i; and

[0176] - ^^interaction f ür e ' n Single substance equalization spectrum between the protonation states of the single substance i.

[0177] It can be approximately assumed that the interaction of the different protonation states with each other is intrinsically contained in the spectral database entries, and therefore the following applies:

[0178] A _ n

[0179]

[0180] Interaction.

[0181] Initially, only coefficients can be used as sub-state concentrations cf". If the sub-state spectra β P x urepH e.g., all normalized to 1 g / L, these coefficients then correspond to the concentrations and can be added without further scaling to calculate the final (total) concentration of the individual substance i.

[0182] Individual substances with different redox states

[0183] Analogous to pH-dependent individual substances, redox reactions of variable individual substances can also be represented for all or all relevant forms that can occur in the redox reaction by means of a dedicated sub-individual pure spectrum and / or a corresponding entry in the SDS and taken into account when describing substance mixtures. For example, the following description can be used for the redox states:

[0184] nredox

[0185] pPure > \ ^redox pP ure redox i predox

[0186] Ati i - / | c i,x + Atl i, Interaction

[0187]

[0188] X=1

[0189] This states:

[0190] - AWAY i Pure for the respective pure spectrum of the individual substance i;

[0191] - c x edox for the sub-state concentration of state x to be determined in the substance mixture to be analyzed, where state x describes a redox state of the individual substance i;

[0192] - ß i p x ureredox for the sub-state spectrum of state x of the individual substance i in the substance mixture to be analyzed;

[0193] n redox refers to the number of redox states of the single substance i; and

[0194] - ^ ßunt°raction f ü r ein Einzelsubstanzausgleichsspektrum zwischen den redox- Staaten der Einzelsubstanz i.

[0195] Here too, it can be approximately assumed that the interaction of the different redox states is intrinsically contained in the spectral database entries, and therefore the following applies:

[0196] predox n

[0197] Interaction.

[0198] Partial models can also be used to describe the redox states of an individual substance within the mixture via latent quantities and / or to pre-interpolate and / or extrapolate them. Thus, for example, only an already approximated and / or optimized pure spectrum of the individual substance can be included in the description of the overall spectrum of the mixture, i.e., in the equilibrium condition.

[0199] Consideration of Conformational Changes of Individual Substances: When determining concentrations, conformational changes of the variable individual substances can also be considered and / or represented, particularly analogous to the pH or redox states of the individual substances. This representation can be achieved using sub-individual pure spectra and / or entries in the safety data sheet (SDS) to describe substance mixtures. Sub-models can also describe the state of an individual substance within the mixture via latent quantities, and / or interpolation and / or extrapolation can be performed beforehand. Thus, an optimized and / or approximated pure spectrum of the individual substance can be incorporated into the description of the overall spectrum of the substance mixture.

[0200] For example, the following description can be used for the conformational states:

[0201] n Conformity

[0202] pPure > \ "Conformation / | pP ureConformation i pConformation Ati i - / | c i,x + Ab i, Interaction

[0203]

[0204] X = 1

[0205] This states:

[0206] - AWAY i Pure for the respective pure spectrum of the individual substance i;

[0207] - c

[0208]

[0209] ^° n f° rmation fQ r the zu Determining sub-state concentration of state x in the substance mixture to be analyzed, where state x describes a conformational state of the individual substance i;

[0210] - B i p x Pure conformation for the sub-state spectrum of state x of the individual substance i in the substance mixture to be analyzed;

[0211] «Conformation f ü T Number of conformational states of the individual substance i; and - ^ ß i Kinte?a™ton n for a single substance compensation spectrum between the conformational states of the single substance i.

[0212] Here too, it can be approximated that the interaction of the different conformational states with each other is intrinsically contained in the spectral database entries, and thus the following holds: oConformation^ Q

[0213]

[0214] Interaction.

[0215] Furthermore, latent quantities can also be derived from substance classes to describe conformational changes in individual substances. For example, partial models can derive the secondary structure of a protein from a protein spectral database and thus approximate changes in a protein in a substance mixture compared to its pure form, without requiring the protein to be measured in different conformational states and / or to be available as a complete database entry.

[0216] Consideration of complex formation between individual substances

[0217] According to one embodiment, the substance mixture contains complexable individual substances which form complexes with each other, whereby when determining the concentration of these complexable individual substances:

[0218] - the complexes of the complexable individual substances are taken into account in the approximate spectral composition of the actual spectrum by means of pure complex spectra and associated complex concentrations,

[0219] and / or

[0220] - Pure spectra are used for the complexable individual substances, which are designed as a partial model that describes the complex formation of the complexable individual substances in the substance mixture via latent quantities.

[0221] Similar to the analysis of individual substances in different states, complexes between individual substances within a mixture can also be considered and / or modeled when determining concentrations. This modeling can be achieved using complex pure spectra and complex concentrations, and / or entries in the safety data sheet (SDS), to describe the mixture. Sub-models can also describe the complex state of the individual substances within the mixture using latent quantities, and / or interpolation and / or extrapolation can be performed beforehand. Thus, an optimized and / or approximated pure spectrum of the relevant complex-capable individual substances can be incorporated into the description of the overall spectrum of the mixture.

[0222] Condition analysis

[0223] In a further training course, after determining the concentrations of the individual substances, a state analysis of at least one selected of the individual substances is carried out, in particular an analysis of a secondary structure and / or a charge and / or a deamidation and / or a glycosylation and / or an oxidation and / or an aggregation and / or other molecular change of the selected individual substance is determined from:

[0224] - determined subconcentrations of sub-substances of the selected individual substance and associated sub-species pure spectra in the actual spectrum of the substance mixture;

[0225] and / or

[0226] - determined proportions of the different previously known states in which the individual substance is present in each proportion;

[0227] and / or

[0228] - determined latent quantities of the submodel describing the state of the individual substance in the substance mixture, which in particular describes the equilibrium and / or the redox state and / or the conformational state of the individual substance;

[0229] and / or

[0230] - the determined latent quantities of the sub-models, which together can describe the respective state of the individual substance in the substance mixture via their respective latent quantities, in particular their equilibrium and / or their protonation state and / or their redox state and / or their conformational state;

[0231] and / or

[0232] - determined complex concentrations of complexes of the selected individual substances and associated complex spectra in the actual spectrum of the substance mixture,

[0233] and / or

[0234] - at least a partial model of the state variance of the selected single substance.

[0235] In this context, spectral analysis not only includes the determination of the individual concentrations of the individual substances, but also a state analysis of at least one of the individual substances is carried out.

[0236] For example, if a single substance i has a pH-dependent value, the overall protonation state of the individual substance i present in the mixture can be deduced from the determined sub-concentrations c for the different protonation states. Thus, the protonation state of the individual substance i can be analyzed in addition to, or as part of, spectral analysis.

[0237] Similarly, for example, in the case of a single substance i, which can assume different redox states, the determined sub-concentrations c can be used to determine x edoxThe different redox states allow conclusions to be drawn about the redox state of the individual substance i present in the substance mixture. Thus, the redox states of this individual substance i can be analyzed in addition to, or as part of, the spectral analysis.

[0238] The same applies to a single substance i with conformational changes. Here, the determined sub-concentrations can be used to... c ^° n f° rmatlon The different conformational states allow conclusions to be drawn about the conformational state of the individual substance i itself, as it exists in the substance mixture. Thus, in addition to or within the framework of spectral analysis, the conformational state of this individual substance i can be analyzed.

[0239] In addition to or within the framework of spectral analysis, complex formations between individual substances can also be analyzed, in particular by means of the respective complex concentrations.

[0240] Typically, changes in the state of a variable individual substance within a mixture must also be considered in the Design of Experiments (DoE) and thus (conventionally) represented in the calibration space in order to be analyzed using an evaluation model. This can be very complex or even impossible, as the different states of an individual substance are not always available in pure form and / or the effort required for preparative isolation of the individual substance is not economically viable compared to the benefit of a single analytical model, which is generally only applicable to a specific mixture of substances. This is the case, for example, with proteins / antibodies in different states of secondary structure, charge, or aggregation.

[0241] For the inventive method, knowledge of various state changes of a single substance and / or its representation in the spectral database is also useful or even necessary. Thus, the variance of a variable single substance in a mixture can also be described by the spectral composition described above, in particular by the equation of the equilibrium condition, in which the sub-species spectra of the state changes are also known. An advantage of this approach is that the description of the single substance is no longer limited to the specific mixture for which it was calibrated. Rather, the description can also be applied to other mixtures in which the state changes of the single substance can also occur.

[0242] An advantage of the method according to the invention is that by varying and / or changing the spectral composition of the actual spectrum and / or the reference spectrum, in particular by changing the equation for the equilibrium condition, the pure spectrum of an individual substance can be extracted from the total spectrum, i.e., the actual spectrum, of the substance mixture. The extracted pure spectrum can then provide information about the current state of the individual substance in the substance mixture in a further analysis, e.g., within the framework of the state analysis discussed above.

[0243] The method can also be used to extract state changes of a single substance from the actual spectrum of the substance mixture, particularly in different mixed states, which are then available as pure forms. Pure forms of state changes of a single substance, which may be difficult to isolate preparatively, can thus be obtained via spectral extraction. This is of interest, for example, for proteins / antibodies in formulations, in order to derive states from the sample under investigation via the secondary structure, charge, or aggregation of the protein / antibody.

[0244] Furthermore, measured or extracted pure spectra, in particular spectral database entries from the SDS, can be used to derive typical and / or generally valid statements for substance classes, which can be applied to unknown individual substances that nevertheless belong to the substance class.

[0245] For example, a general model for predicting the secondary structure of proteins can be created from the SDS database entries of the substance class proteins, based on their pure or extracted spectra. This model can be used to describe the secondary structure of unknown proteins in substance mixtures after approximating the corresponding pure spectrum, i.e., by means of spectral extraction. This can be particularly advantageous when changes in the protein's state are not available for the calibration of conventional methods.

[0246] A substance class is not necessarily described solely by a single spectrum, but can also be described by multiple spectra, with these spectra representing the variance of the substance class. Examples of substance classes include proteins, RNA, and / or polypeptides. Raw material classes that exhibit variance, such as polysorbate 20 or 80, can also be considered substance classes. Furthermore, the aging and / or rearrangement of a substance can be understood as a substance class, as seen in the Lobry-de Bruyn-Alberda-van Ekenstein rearrangement of glucose or the epimerization of a molecule. In these cases, various sub-spectra can describe the rearrangement effects of the substance.

[0247] In other words, even for a single substance for which the pure spectrum is not precisely known, especially for which no correct SDS entry exists, but which belongs to a known class of substances such as proteins, a partial model and / or dummy model can be used instead of the pure spectra. This allows for better elimination of variances in a mixture of substances.

[0248] The subsequent state analysis following the determination of concentrations can increase and thus improve the informative value of the spectral analysis. This can be particularly crucial in the testing of medications, where the effect often depends not only on the concentrations of the individual substances, but also, for example, on the conformational state of the individual substance(s) assumed within the mixture. The quality of such state-dependent medications can also be reliably verified using this method.

[0249] In one embodiment, exactly one reference spectrum is used to determine the concentrations of the individual substances in the substance mixture, and in particular, a quality and / or stability control is carried out.

[0250] Interaction effects between molecules and / or ions can be of a diverse nature. Therefore, the method according to the invention can generally distinguish between several cases with regard to whether one, several, or no reference point is required and used for spectral analysis. The embodiment described here relates to the variant using a single reference point. This method is particularly well-suited for analyzing products where a consistent composition and / or state with only minor batch variations is desired. This quality attribute is commonly found in many manufacturing processes, especially in the food and / or pharmaceutical industries.

[0251] The use of a single reference point represents a significant reduction in calibration effort compared to conventional methods, where a calibration space with, for example, 2 n Calibration points must be defined within which the substance mixture must be arranged. In contrast to this conventional interpolation within the calibration space, the method according to the invention extrapolates away from the reference point.

[0252] One variant involves a method that dispenses with even a single reference point. This variant thus relates to a method for the spectral analysis of a liquid mixture of substances consisting of known individual substances, in which an actual spectrum of the liquid mixture to be analyzed is generated and approximate concentrations of the individual substances in the mixture are determined based on an approximate spectral composition of the actual spectrum from known pure spectra of the known individual substances.

[0253] This variant of the method can be sufficiently accurate if the individual substances do not interact with each other in any relevant ways. This is the case, for example, when the concentrations of the ingredients are low.

[0254] In one embodiment, the substance mixture clusters into different cluster states. For each cluster state, a corresponding cluster reference spectrum is used to determine the concentrations of the individual substances in the mixture. This method thus uses not just a single reference spectrum, but several cluster reference spectra. During the determination process, not only can the concentrations of the individual substances in the entire mixture be determined, but also, or alternatively, the cluster concentrations of the individual substances in the different possible clusters. This results, for example, in the following spectral composition of the reference spectrum:

[0255] ApMiXRef _ y n ciuster _ y n C luster ( iy ra comp ^Mix^gPure \

[0256]

[0257] ^-'1 = 1 °1 ^1=1 y u |, Interaction l,i 14 J'

[0258] This states:

[0259] - ß MiXRe / for the reference spectrum of the substance mixture;

[0260] -

[0261]

[0262] for the cluster reference spectra of cluster 1 of the substance mixture; - n ciuster for the number of clusters 1 of the substance mixture

[0263] - Aß interaction

[0264]

[0265] c * as Compensation spectrum of cluster 1 of the substance mixture in the target state;

[0266] - AB1 p i ure for the respective sub-state spectrum of the individual substance i in cluster 1;

[0267] - c ( M ( ix for the respective target concentration of the individual substance i in cluster 1 in the substance mixture in the target state; and

[0268] n CO mp represents the number of individual substances i in the substance mixture.

[0269] In this case, several compensation spectra AB can be used. MiXSam The compensation condition, derived from the cluster reference spectra by subtracting the individual substances, takes into account the differences. The compensation spectra can each be used proportionally to describe a mixture of substances via a factor. This factor can be determined using a model, such as a least-squares fit.

[0270] For the compensation spectra AB MiXSam applies: n Cluster ra comp

[0271] AB Mix Sam = V x A B^ Ref .. + yc^

[0272] / , 1 1, Interaction / , 1 ix XB

[0273] 1Pure

[0274]

[0275] 1=1 i=l

[0276] Alternatively, the compensation spectra of the clusters can also be combined.

[0277] Multiple reference points, such as cluster reference points with cluster reference spectra, may be necessary if the substance mixture clusters into several highly probable states. This could, for example, also include the description of a reaction or process where the reference points represent several typical states of the process.

[0278] All variants have in common that the method always describes the deviation of the substance mixture from the reference point matrix via the pure spectra of the relevant ingredients and / or via their sub-models, taking environmental influences into account.

[0279] Description of unknown state changes or interactions

[0280] In one embodiment, residual variance in the approximate spectral composition of the actual spectrum is compensated for by Voigt profiles. Thus, another variant of the method according to the invention is the description of the individual bands in spectra using Voigt profiles. For example, changes in the state of a single substance or interactions between single substances that cannot be represented, or cannot be represented sufficiently, by the aforementioned methods so that the compensation condition is met, can also be approximated by Voigt profiles. Here, any indescribable residual variance (residuals) that may still be present of a relevant magnitude when solving the equation for the compensation condition is compensated for by the Voigt profiles.This can be achieved, for example, by a least-squares fit, in which the parameters describing one or more Voigt profiles, such as wavenumber, peak amplitude, width, and shape, are determined iteratively. In one embodiment, the approximate determination is performed using a least-squares fit and / or extrapolation. In particular, the spectral composition of the actually recorded spectrum and / or the equation of the regression condition can be solved using such an approximation method. Alternatively or additionally, neural networks and / or deep learning methods can be used to solve the spectral composition of the approximation and / or the equation of the regression condition. Such approximation methods are well-established and can often be solved using standardized mathematical procedures.

[0281] One aspect concerns a computer-implemented system for the spectral analysis of a liquid mixture of substances consisting of known individual substances. This system includes a reference module containing a reference spectrum for the mixture, where the reference spectrum represents a target state of the mixture in which each individual substance is present at a target concentration. The system also features a spectral module configured to receive an actual spectrum of the liquid mixture to be analyzed. Furthermore, it includes an evaluation module configured to approximate the concentrations of the individual substances in the mixture based on an approximate spectral composition of the actual spectrum derived from known pure spectra of the known individual substances and at least one compensation spectrum.

[0282] The system can be configured to carry out the procedure described above. Therefore, all statements regarding the procedure also apply to the system, and vice versa.

[0283] The modules can each be designed as a software module and / or a computer program product. They can each have one and / or a shared computer-readable storage medium on which one or more computer programs are stored. The respective computer program can include instructions which, when executed by a programmable processor, perform the respective process steps. In particular, the evaluation module can be programmed to determine the concentrations of the individual substances approximately by means of a spectral composition of the actual spectrum from the previously known pure spectra of the previously known individual substances and the at least one compensation spectrum.

[0284] In one embodiment, the computer-implemented system includes a spectral database containing previously known pure spectra, interaction spectra, pattern spectra, compensation spectra, and / or sub-state spectra. The evaluation module retrieves these spectra from this database. The spectral database allows for easy access to the spectra. In one variant, the output of the pure spectra can be achieved via a sub-model that derives them from latent quantities. In particular, the spectral database can enable access to the same pure spectrum of a single substance for the analysis of different substance mixtures, thus allowing the information contained in the pure spectrum about the single substance to be used for the analysis of different substance mixtures.

[0285] Additional aspects and / or embodiments of the invention are described in more detail below with reference to the figures. These show:

[0286] Figure 1 shows a schematic representation of an embodiment of a computer-implemented system for spectral analysis of a liquid mixture of substances.

[0287] Figure 1 shows an exemplary computer-implemented system for realizing the invention, comprising a general-purpose computing device in the form of a conventional computing environment 20 (e.g., a personal computer). The conventional computing environment 20 includes an arithmetic unit 22, a system memory 24, and a system bus 26. The system bus 26 connects various system components, including the system memory 24, to the arithmetic unit 22. The arithmetic unit 22 can perform arithmetic, logical, and / or control operations by accessing the system memory 24. The system memory 24 can store information and / or instructions for use in combination with the arithmetic unit 22. The system memory 24 can contain volatile and non-volatile memory, such as random-access memory (RAM) 28 and read-only memory (ROM) 30.ROM 30 can contain a basic input / output system (BIOS) that includes the basic routines that assist in transferring information between elements within the computing environment 20, for example, during startup. The system bus 26 can be one of several types of bus structures, including a memory bus or memory controller, a peripheral bus, and a local bus, which uses one of several bus architectures.

[0288] The computing environment 20 can also include a drive 32 for reading and writing to a hard disk and / or an SSD (not shown) and an external disk drive 34 for reading and writing to a removable disk 36. The removable disk 36 can be, for example, a magnetic disk for a magnetic disk driver, an optical disk such as a CD-ROM for an optical drive, or an SSD (Solid State Drive). The drive 32 and the external disk drive 34 are connected to the system bus 26 via a disk drive interface 38 and an external disk drive interface 40, respectively. The drives and their associated computer-readable media serve for the non-volatile storage of computer-readable instructions, data structures, program modules, and other data for the computing environment 20.The data structures can contain relevant data for carrying out the spectral analysis procedure of a liquid mixture consisting of known individual substances, as described above. The relevant data can be organized in a database, e.g., in a relational database management system or an object-oriented database management system.

[0289] Although the exemplary environment described here uses a hard disk (not shown) and a removable disk 36, other types of computer-readable media capable of storing data accessible to a computer can also be used, such as magnetic cartridges, flash memory cards, digital video discs, random access memory, read-only memory, solid-state disks, and the like.

[0290] A number of program modules can be stored on the hard disk, the removable disk 36, the ROM 30 or the RAM 28, including an operating system (not shown), one or more application programs 44, other program modules (not shown) and program data 46. The application programs can contain at least part of the functionality of the method for the spectral analysis of a liquid mixture of substances consisting of previously known individual substances.

[0291] A user can input commands and information, as described below, into the computing environment 20 via input devices such as the keyboard 48 and the mouse 50. Other input devices (not shown) may include a microphone (or other sensors), a joystick, a gamepad, a scanner, or similar devices. These and other input devices can be connected to the computing unit 22 via a serial interface 52 coupled to the system bus 26, or they can be acquired via other interfaces, such as a parallel interface 54, a game port, or a universal serial bus (USB). Furthermore, the information can be printed via a printer 56. The printer 56 and other parallel input / output devices can be connected to the computing unit 22 via the parallel interface 54. A monitor 58 or other type of display device can also be connected via an interface, such as...A video input / output 60, connected to the system bus 26. In addition to the monitor, the computing environment can include 20 further peripheral output devices (not shown), such as loudspeakers or other acoustic outputs.

[0292] The computing environment 20 can communicate with other electronic devices such as a computer, a (wired or wireless) telephone, a personal digital assistant, a television, or similar devices. In particular, the computing environment 20 can communicate with a measuring device (not shown) that sends measurement data of the actual spectrum of the liquid substance mixture to be analyzed to the computing environment 20. The measuring device can be connected to the system bus 26 via one of the interfaces of the computing environment 20. Alternatively, the measurement data can also be sent to the computing environment via a remote computer 62 and / or a network.

[0293] To communicate, the computing environment 20 can operate in a networked environment that utilizes connections to one or more electronic devices. Figure 1 shows the computing environment networked with the remote computer 62. The remote computer 62 can be another computing environment, such as a measuring device with a processor, a server, a router, a network PC, a peer device, or another shared network node, and can include many or all of the elements described above with respect to the computing environment 20. The logical connections shown in Figure 1 include a local area network (LAN) 64 and a wide area network (WAN) 66. Such network environments are common in offices, enterprise computer networks, intranets, and the Internet, and can, in particular, be encrypted.

[0294] When used in a LAN network environment, the computing environment 20 can be connected to the LAN 64 via a network I / O 68. When used in a WAN network environment, the computing environment 20 can include a modem 70 or other means for establishing communication over the WAN 66. The modem 70, which can be internal or external to the computing environment 20, is connected to the system bus 26 via the serial interface 52. In a networked environment, the program modules or parts thereof represented in relation to the computing environment 20 can be stored on a remote storage device located on or accessible to the remote computer 62. Furthermore, other data relevant to the procedure for optimizing the evaluation of a policy (as described above) can be stored on or accessible via the remote computer 62.The network connections shown are examples only, and alternative or additional means can be used to establish a communication connection between the electronic devices.

[0295] The computing environment described above is only one example of the type of computer system that can be used to carry out the method for the spectral analysis of a liquid mixture of substances consisting of previously known individual substances.

[0296] In particular, the computing environment 20 can be configured as a computer-implemented system for the spectral analysis of a liquid mixture of substances consisting of known individual substances. The reference module can, for example, be stored on the hard drive 32. The spectral module can receive the actual spectrum of the liquid mixture to be analyzed via at least one of the interfaces 52 / 54, the network 68, the modem 70, and / or the system bus 26. The evaluation module can be configured as one of the application programs 44 of the computing environment 20.

[0297] Reference symbol list

[0298] 20 Computing environment

[0299] 22 computing units

[0300] 24 GB system memory

[0301] 26 System bus

[0302] 28 Random Access Memory (RAM)

[0303] 30 Read-only memory (ROM)

[0304] 32-bit drive

[0305] 34 External disk drive

[0306] 36 Interchangeable plates

[0307] 38 Interface for disk drive

[0308] 40 Interface for external disk drives

[0309] 44 one or more application programs 46 program data

[0310] 48 keyboard

[0311] 50 mice

[0312] 52 serial interface

[0313] 54 parallel interfaces 56 printers

[0314] 58" Monitor

[0315] 60 Video input / output 62 Remote computer

[0316] 64 Local Area Network (LAN) 66 Wide Area Network (WAN) 68 Network I / O

[0317] 70 Modem

Claims

Applicant: Clade AG MB&P symbol: C05254WO - hy / mu Patent claims 1. Method for the spectral analysis of a liquid mixture of substances consisting of previously known individual substances, comprising the following steps: Providing a reference spectrum for the substance mixture, wherein the reference spectrum represents a target state of the substance mixture in which each individual substance is contained in a target concentration; Creating an actual spectrum of the liquid substance mixture to be analyzed; and Determining the concentrations of the individual substances in the substance mixture based on an approximate spectral composition of the actual spectrum from previously known pure spectra of the previously known individual substances and at least one compensation spectrum.

2. Method according to claim 1, wherein the compensation spectrum represents interaction effects of the individual substances in the substance mixture.

3. Method according to claim 1 or 2, wherein the compensation spectrum is determined by a spectral subtraction of the pure spectra of the individual substances at the respective target concentration from the reference spectrum.

4. Method according to one of the preceding claims, wherein the previously known pure spectra are stored in a spectral database and are read out from the spectral database before determining the concentrations of the individual substances.

5. Method according to claim 4, wherein the pure spectra stored in the spectral database are created for the solvent in which the substance mixture is also dissolved.

6. Method according to one of the preceding claims, wherein, when determining the concentrations of the individual substances in the substance mixture, deviations of the concentrations of the individual substances from their respective target concentrations are determined.

7. Method according to one of the preceding claims, wherein, when determining the concentration, at least one of the individual substances, which assumes different forms in the substance mixture, in particular different protonation forms and / or different redox states and / or different conformational states, is: the different forms of the individual substance are taken into account as sub-substances, and the sub-concentrations of the sub-substances are determined using previously known sub-substance pure spectra of the sub-substances; and / or Several previously known sub-state spectra are taken into account, in which the individual substance is present in different previously known states and thus in differently proportionate forms; and / or For this single substance, a pure spectrum is used, which is designed as a sub-model that describes the state of the single substance in the mixture of substances via latent quantities, in particular its equilibrium and / or its redox state and / or its conformational state; and / or For this single substance, several sub-state spectra are used, each of which is designed as a sub-model, so that the sub-models together can describe the respective state of the single substance in the mixture of substances via their respective latent quantities, in particular their equilibrium and / or their protonation state and / or their redox state and / or their conformational state.

8. A method according to any of the preceding claims, wherein the mixture of substances contains complexable individual substances which form complexes with each other, wherein, when determining the concentration of these complexable individual substances: The complexes of the complexable individual substances are taken into account in the approximate spectral composition of the actual spectrum by means of pure complex spectra and associated complex concentrations. and / or Pure spectra are used for the complexable individual substances, which are designed as a partial model that describes the complex formation of the complexable individual substances in the substance mixture via latent quantities.

9. A method according to claim 7 or 8, wherein, after determining the concentrations of the individual substances, a state analysis of the state of at least one selected of the individual substances is carried out, in particular an analysis of a secondary structure and / or a charge and / or a deamidation and / or a glycosylation and / or an oxidation and / or an aggregation and / or other molecular change of the selected individual substance, wherein the state of the selected individual substance is determined from: determined subconcentrations of sub-individual substances of the selected individual substance and associated sub-individual pure spectra in the actual spectrum of the substance mixture; and / or determined proportions of the different previously known states in which the individual substance is present in each proportion; and / or Determined latent quantities of the submodel describing the state of the individual substance in the substance mixture, which in particular describes the equilibrium and / or the redox state and / or the conformational state of the individual substance; and / or the determined latent quantities of the sub-models, which together can describe the respective state of the individual substance in the substance mixture via their respective latent quantities, in particular their equilibrium and / or their protonation state and / or their redox state and / or their conformational state; and / or determined complex concentrations of complexes of the selected individual substances and associated complex spectra in the actual spectrum of the substance mixture, and / or a sub-model of the state variance of the selected single substance.

10. Method according to any of the preceding claims, wherein exactly one reference spectrum is used to determine the concentrations of the individual substances in the substance mixture and in particular a quality and / or stability control is carried out.

11. Method according to any one of claims 1 to 9, wherein the substance mixture clusters into different cluster states and for each cluster state an associated cluster reference spectrum is used to determine the concentrations of the individual substances in the substance mixture.

12. Method according to one of the preceding claims, wherein a residual variance of the approximate spectral composition of the actual spectrum is compensated by Voigt profiles.

13. Method according to one of the preceding claims, wherein the determination is carried out by means of a least-squares fit and / or by means of extrapolation.

14. Computer-implemented system for spectral analysis of a liquid mixture of substances consisting of previously known individual substances with: a reference module in which a reference spectrum for the The substance mixture is stored, with the reference spectrum representing a target state of the substance mixture in which each individual substance is contained in a target concentration; a spectral module configured to receive an actual spectrum of the liquid mixture of substances to be analyzed; and an evaluation module configured to determine concentrations of the individual substances in the substance mixture based on an approximate spectral composition of the actual spectrum from previously known pure spectra of the previously known individual substances and at least one compensation spectrum.

15. Computer-implemented system according to claim 14 with a spectral database in which the previously known pure spectra and / or previously known interaction spectra and / or previously known pattern spectra and / or previously known compensation spectra and / or previously known sub-state spectra are stored, and from which the evaluation module reads these spectra.

Images