HRP20210509A1 - PROCEDURE FOR MENOPAUSE DIAGNOSIS AND FOR PERIMENOPAUSE PERIOD DETERMINATION BASED ON BLOOD PLASMA IgG GLYCOMA COMPOSITION - Google Patents

Abstract

The subject invention discloses a procedure for diagnosing menopause and passing through perimenopause through the quantitative analysis of N-glycans bound to immunoglobulin G from the blood plasma of women. Quantitative amounts of one or more IgG glycans are included in one or more numerical models previously obtained by statistical processing of the results of a study of variations in the quantitative IgG glycan in blood plasma on a group of women, some of whom were in menopause and some who were not. As a result, a number is obtained that describes the probability of whether the examined female person has entered the menopause phase or not. The diagnostic procedure according to the invention is useful in determining whether the examined female person has entered the perimenopause or menopause phase. The subject invention discloses a procedure for diagnosing menopause and passing through perimenopause by quantitative analysis of N-glycans bound to immunoglobulin G from the blood plasma of women. The amounts of quantitative parts of one or more IgG glycans are inserted in one or more numerical models previously obtained by statistical processing of the results of a study of variations in the quantitative part of IgG glycan in blood plasma on a group of women, some of who were in menopause and some who were not. As a result, a number is obtained that describes the probability of whether the examined female person has entered the menopause phase or not. The diagnostic procedure according to the invention is useful in determining whether the examined female person has entered the perimenopause or menopause phase.

Description

1.1 Field of the invention

The subject invention relates to the procedure for diagnosing menopause and determining the period of perimenopause through the analysis of N-glycans bound to immunoglobulin G (IgG) from the blood plasma of a female subject.

1.2 Technical problem

The subject invention solves the technical problem of reliably determining whether the examined female person has entered the perimenopause or menopause phase. It is known that perimenopause or the early stages of menopause are difficult to diagnose due to the frequent variations of sex hormones such as estradiol or through the analysis of other known biochemical markers such as follicle-stimulating hormone (FSH), anti-Müllerian hormone (AMH) or inhibitin A or B. The present invention solves the mentioned technical problem based on the quantitative analysis of IgG N-glycans from a single blood sample of an examined female person aged 45 to 55 years.

1.3 Prior art

Glycans are complex carbohydrates mainly based on N-acetyl-glucosamine (■), fucose (▼), mannose (●), galactose (○) and N-acetyl-neuraminic acid (♦), which are often attached to proteins, typically N -glycosidic bond and involved in numerous physiological and pathological processes. Due to their involvement in a large number of biological processes, they are of great importance as biochemical indicators of the general state of health and various physiological and pathological conditions of the human organism; see literature reference 1:

1) G. Opdenakker, P. M. Rudd, C. P. Ponting, R. A. Dwek: Concepts and principles of glycobiology, FASEB J. 7 (1993) 1330-1337.

Immunoglobulin G (IgG) is the most abundant antibody in human blood plasma and plays an important role in the body's defense against various pathogens. IgG is a glycoprotein whose stability and function are key to the glycans attached to its heavy chains. Glycosylation of IgG also depends on different physiological (age, sex, pregnancy) and pathological conditions (tumors, infections, autoimmune diseases). It is known in the state of the art that the glycosylation pattern of IgG changes with the age of an individual, and for example, by analyzing IgG N-glycans, it is possible to monitor the aging process itself and draw conclusions about the biological age of the subject; see literature references 2-5:

2) R. Parekh, I. Roitt, D. Isenberg, R. Dwek, T. Rademacher: Age-related galactosylation of the N-linked oligosaccharides of human serum IgG, J. Exp. Honey. 167 (1988) 1731-1736.

3) M. Pučić, A. Knežević, J. Vidič, B. Adamczyk, M. Novokmet, O. Polašek, O. Gornik, S. Supraha-Goreta, M. R. Wormald, I. Redžić, H. Campbell, A. Wright , N. D. Hastie, J. F. Wilson, I. Rudan, M. Wuhrer, P. M. Rudd, D. Josić, G. Lauc: High Throughput Isolation and Glycosylation Analysis of IgG-Variability and Heritability of the IgG Glycome in Three Isolated Human Populations, Molecular & Cellular Proteomics 10.10; doi:10.1074/mcp.M111.010090.

4) EP3011335B1; G. Lauc, M. Pučić-Baković, F. Vučković: Method for the analysis of N-glycans attached to immunoglobulin G from human blood plasma and its use; holder: Genos d.o.o. (HR); priority date: 20.06.2013.

5) J. Krištić, F. Vučković, C. Menni, L. Klarić, T. Keser, I. Beceheli, M. Pučić-Baković, M. Novokmet, M. Mangino, K. Thaqi, P. Rudan, N. Novokmet, J. Sarac, S. Missoni, I. Kolčić, O. Polašek, I. Rudan, H. Campbell, C. Hayward, Y. Aulchenko, A. Valdes, J. F. Wilson, O. Gornik, D. Primorac, V Zoldoš, T. Spector, G. Lauc: Glycans are a novel biomarker of chronological and biological ages, J. Gerontol. And Biol. Sci. Honey. Sci. 69 (2014) 779-789. doi: 10.1093/gerona/glt190.

Menopause is defined as the period in a woman's life that occurs 12 or more months after the last menstrual period. It is characterized by complete or almost complete exhaustion of the ovaries, which results in very low levels of the female sex hormone estradiol in the serum, and a significantly increased concentration of follicle-stimulating hormone (FSH). Common symptoms usually begin around age 47 or 4-6 years before menopause. The most common symptoms of menopause are hot flashes, irregular menstrual bleeding, insomnia, mood changes (anxiety, depression), mastodynia, headache and vaginal dryness.

The period of transition from a woman's normal fertile age to the onset of menopause is known as perimenopause. It is characterized by a decrease in the concentration of inhibin B, a variable or increased concentration of FSH, a decreased concentration of AMH and a slight decrease in the number of antral follicles. These changes are accompanied by a variation in the interval of menstruation, a decrease in fertility and the appearance of menopausal symptoms; see literature references 6 and 7:

6) T. Hillard: NICE guideline – Menopause: diagnosis and management, Post Reprod. Health. 22 (2016) 56-58;

7) J.L. Bacon: The Menopause Transition, Obstet. Gynecol. Clin. N. Am. 44 (2017) 285-296.

Although IgG N-glycans are known to change with age, their association with perimenopausal or menopausal status in women has generally not been examined. In particular, it is not known whether their analysis can draw conclusions about the onset of the perimenopause or menopause phase.

The subject invention solves the described technical problem in a new and inventive way with the application of the known analytical methodology of quantitative analysis of IgG N-glycans and the hitherto unknown connection of their variation with the onset of perimenopause and menopause.

1.4 The essence of the invention

The subject invention includes a procedure for diagnosing menopause and determining the period of perimenopause through the analysis of N-glycans bound to immunoglobulin G (IgG) from the blood plasma of women, denoted by the abbreviations GP1 to GP22, of the general chemical structure I:

[image]

AND

[image]

where said procedure includes:

(i) quantitative analysis of one or more:

(a) fluorescently derivatized glycans released from IgG by the enzyme peptide-N4-(N-acetyl-beta-glucosaminyl) asparagine amidase (PNGase F); or

(b) free glycans or corresponding glycopeptides or glycoforms;

(ii) inclusion of the obtained results of quantitative proportions of said glycans in one or more numerical models that were previously obtained by statistical processing of the results of the study of variations in the quantitative proportion of IgG glycans in blood plasma on a selected group of women, some of whom were in menopause and some who were not; you

(iii) calculating the result in the form of a number that describes the probability that the examined female person has passed through perimenopause and entered the menopause phase.

The selected group of women from step (ii) was between the ages of 45 and 55, and the aforementioned diagnostic procedure is applied to subjects who belong to that age range.

The result of the diagnostic procedure according to the invention from step (iii) is interpreted as follows:

(a) if the score is between 0.5 and 1.0, the female subject has passed through perimenopause and entered the menopause phase; And

(b) if the result is between 0 and 0.5, the examined woman has not gone through perimenopause and entered the perimenopause phase.

The diagnostic procedure according to the invention includes the calculation of the probability [image] that the examined female person has entered the menopause phase, which is calculated using a numerical model:

[image]

where: [image] , [image] , [image] and [image] Logit transformed values of the relative area under the peaks of the glycans of the same name GP2, GP4, GP12 and GP22 from the chromatogram of the selected quantitative analytical technique.

Alternatively, the diagnostic procedure according to the invention is based on the calculation of the probability[image] that the examined female person has entered the menopause phase, calculated using the formula:

[image]

where are they:

-[image] Logit transformed value of the relative area under the GP12 glycan peak from the chromatogram of the appropriate quantitative analytical technique,

-[image] ,[image] ,[image] and [image] average annual changes in Logit transformed values of the relative area under the peaks of glycans of the same name G11, G12, G13 and G16 from the chromatogram of the corresponding quantitative analytical technique.

1.5 Brief description of the images

Figure 1. Typical chromatogram of RapiFluor (RF) derivatized IgG N-glycans obtained by ultra-high performance liquid chromatography (HILIC-UPLC-FLR) method described in the subject invention with 22 separate peaks which are hereinafter denoted by the abbreviations GP1-GP22.

Figure 2A. Model-estimated average levels of IgG N-glycan GP1-GP11 in women before and after the onset of menopause. Error bars indicate the 95% confidence interval of the mean level of the indicated IgG N-glycans.

Figure 2B. Model-estimated average levels of IgG N-glycan GP12-GP22 in women before and after the onset of menopause. Error bars indicate the 95% confidence interval of the mean level of the indicated IgG N-glycans.

Figure 3A. Model estimated average annual changes in levels of IgG N-glycan GP1-GP11 in women before and after the perimenopause period, i.e. the onset of menopause. Error bars indicate the 95% confidence interval of the mean annual change in the level of the indicated IgG N-glycans.

Figure 3B. Model estimated average annual changes in levels of IgG N-glycan GP12-GP22 in women before and after the perimenopause period, i.e. the onset of menopause. Error bars indicate the 95% confidence interval of the mean annual change in the level of the indicated IgG N-glycans.

Figure 4. ROC curves A-C obtained from analysis of specificity and sensitivity of menopausal probabilities calculated by equations for classifying women into menopausal and non-menopausal women on the testing subset. The shaded area around the curve indicates the 95% confidence interval. The ROC curve shown in panel A corresponds to equation (3), in panel B to equation (4), and in panel C to equation (5).

Figure 5. ROC curves A-C obtained from the specificity and sensitivity analysis of the probabilities calculated by the equations for classifying women into menopausal and non-menopausal women on the testing subset. The shaded area around the curve indicates the 95% confidence interval. The ROC curve shown in panel A corresponds to equation (6), in panel B to equation (7), and in panel C to equation (8).

Figure 6. ROC curves A-C obtained from the specificity and sensitivity analysis of the probabilities calculated by the equations for classifying women into menopausal and non-menopausal women on the testing subset. The shaded area around the curve indicates the 95% confidence interval. The ROC curve shown in panel A corresponds to equation (9), in panel B to equation (10), and in panel C to equation (11).

1.6 Detailed description of the invention

The subject invention includes a procedure for diagnosing menopause and determining the period of perimenopause through the analysis of N-glycans bound to immunoglobulin G (IgG) from the blood plasma of women, denoted by the abbreviations GP1 to GP22, of the general chemical structure I and the specific structures shown in Table 1:

[image]

AND

[image]

Table 1. Structures of N-glycans bound to immunoglobulin G (IgG) from the blood plasma of women.

[image] [image]

characterized by the fact that said procedure includes:

(i) quantitative analysis of one or more:

(a) fluorescently derivatized glycans released from IgG by the enzyme peptide-N4-(N-acetyl-beta-glucosaminyl) asparagine amidase (PNGase F); or

(b) free glycans or corresponding glycopeptides or glycoforms;

by some of the suitable analytical techniques, whereby the quantitative proportions of glycans or corresponding glycopeptides are obtained by normalization, by dividing the area under the peak of each of the requested glycans by the total area of all peaks of glycans or glycoforms from the sample;

(ii) inclusion of the obtained results of quantitative proportions of said glycans in one or more numerical models that were previously obtained by statistical processing of the results of the study of variations in the quantitative proportion of IgG glycans in blood plasma on a selected group of women, some of whom were in menopause and some who were not; you

(iii) calculating the result in the form of a number that describes the probability that the examined female person has passed through perimenopause and entered the menopause phase.

The selected group of women from step (ii) was between the ages of 45 and 55, and the aforementioned diagnostic procedure is applied to subjects who belong to that age range.

The result of the procedure according to the invention from step (iii) is interpreted as follows:

(a) if the score is between 0.5 and 1.0, the female subject has passed through perimenopause and entered the menopause phase; And

(b) if the result is between 0 and 0.5, the examined woman has not gone through perimenopause and entered the perimenopause phase.

The diagnostic procedure according to the invention includes the calculation of the probability [image] that the examined female person went through perimenopause and entered the menopause phase, which is calculated using a numerical model:

[image]

where: [image] , [image] , [image] and [image] Logit transformed values of the relative area under the peaks of the glycans of the same name GP2, GP4, GP12 and GP22 from the chromatogram of the selected quantitative analytical technique.

Alternatively, the diagnostic procedure according to the invention is based on the calculation of the probability[image] that the examined female person went through perimenopause and entered the menopause phase, calculated using the formula:

[image]

where are they:

-[image] Logit transformed value of the relative area under the GP12 glycan peak from the chromatogram of the appropriate quantitative analytical technique,

-[image] ,[image] ,[image] and [image] average annual changes in Logit transformed values of the relative area under the peaks of glycans of the same name G11, G12, G13 and G16 from the chromatogram of the corresponding quantitative analytical technique.

Since glycans themselves do not absorb UV, they are not suitable for detection by standard UV-VIS or fluorescence (FLR) detectors used in various quantitative analysis techniques such as ultra-high performance liquid chromatography (UPLC). Therefore, after their release from immunoglobulin G (IgG) from the blood plasma, they are derivatized with suitable reagents that covalently bind to glycans, after which the corresponding glycan derivatives have UV absorption and are suitable for detection during quantitative analysis. In this sense, the diagnostic procedure according to the invention is based on glycan derivatization with reagents selected from the group consisting of:

(i) combination of a suitable aromatic amine such as 2-aminobenzamide (2AB) or procainamide (PR) and a suitable reductant for reductive amination: 2-picoline borane complex (BH3•NC5H4-2-CH3) or sodium cyanoborohydride (NaBH3CN);

[image]

or

(ii) 2,5-dioxopyrrolidin-1-yl-[2N-[2-(N',N'-diethylamino)ethyl] carbamoyl]-quinolin-6-yl-carbamate (RF), known under the trade name RapiFluor- MS from Waters (USA):

[image] .

The application of the described derivatization reagents is described in the state of the art; see literature reference 8:

8) T. Keser, T. Pavić, G. Lauc, O. Gornik: Comparison of 2-Aminobenzamide, Procainamide and RapiFluor-MS as Derivatizing Agents for High-Throughput HILIC-UPLC-FLR-MS N-glycan Analysis, Front. Chem. 6 (2018) 321; doi: 10.3389/fchem.2018.00324.

Suitable analytical techniques for quantitative analysis are: ultra-high performance liquid chromatography (UPLC), MALDI-TOF mass spectrometry (matrix-assisted laser desorption/ionization time-of-flight), liquid chromatography coupled with mass spectrometry (LC-MS ) or capillary electrophoresis (CE), or other suitable quantitative analytical techniques. The description of the quantitative analysis of IgG N-glycans using different analytical techniques is described in the state of the art; see literature reference 4.

Analysis of N-glycans from blood plasma

Isolation of a blood plasma sample from women for the purpose of diagnosing menopause and determining the passage through perimenopause was carried out according to the methodology known in the state of the art; see literature reference 4. Also, isolation of IgG from blood plasma was performed according to a procedure known in the state of the art; see literature references 3 and 9:

9) I. Trbojević Akmačić, I. Ugrina, G. Lauc: Comparative Analysis and Validation of Different Steps in Glycomics Studies, Methods Enzymol. 586 (2017) 37-55. doi: 10.1016/bs.mie.2016.09.027.

For a demonstration of the implementation of the subject invention, in the derivatization phase of the released N-glycans with IgG, they were derivatized with the RF reagent according to the procedure described in literature reference 8 and the manufacturer's instructions, see literature reference 10:

10) GlycoWorks RapiFluor-MS N-glycan Kit – 96 Samples; Waters Corporation, 34 Maple Street, Milford, MA 01757 (USA); www.waters.com; see at the link: https://www.waters.com/waters/library.htm?cid=511436&lid=134834845&lcid=134834844&locale=en_US.

A typical chromatogram of RapiFluor (RF) derivatized IgG N-glycans obtained using the described HILIC-UPLC-FLR method with 22 separate peaks, which are further denoted by the abbreviations GP1-GP22, is shown in Figure 1.

The order of appearance of glycan peaks GP1-GP22 in the described HILIC-UPLC-FLR method and the corresponding retention times (tR) are shown in Table 2.

Table 2. Retention times (tR) of RapiFluor (RF)-labeled IgG glycans marked with abbreviations GP1-GP22 obtained by the described UPLC-HILIC method, whose typical chromatogram is shown in Figure 1.

[image]

A study of monitoring the variability of N-glycans bound to immunoglobulin G in women aged 45 to 55 years, some of whom have entered menopause and some who have not

TwinsUK, the world's largest registry of adult twins and one of the most clinically described cohorts, founded in 1992, was used in the study of the variation of N-glycans GP1-GP22 bound to IgG from the blood plasma of women. The aim of the TwinsUK registry is to investigate the genetic and environmental background of various complex pathophysiological conditions. TwinsUK is today one of the most genotyped and phenotyped cohorts in the world and currently includes around 14,000 identical and fraternal twins.

Whole blood samples were collected by the TwinsUK registry at multiple time points (minimum 1, maximum 3 per person) over 20 years. Whole blood was collected in an EDTA tube and mixed well. The test tube was then left at room temperature and then centrifuged to separate the plasma. The plasma was transferred to a clean test tube and stored in a freezer at -80 or -20 ̊C.

A total of 6032 samples were analyzed:

(i) 1865 individuals at three time points (5595 samples in total);

(ii) 156 individuals at two time points (312 samples in total); you

(iii) 125 individuals at one time point (125 samples).

Isolation of IgG from blood plasma was performed according to the procedure described in literature references 3 and 9. Deglycosylation of the isolated IgG was performed using the enzyme peptide-N4-(N-acetyl-beta-glucosaminyl) asparagine amidase (PNGase F). of glycans was carried out by treatment with the RapiFluor-MS reagent from Waters (USA), which is based on 2,5-dioxopyrrolidin-1-yl-[2N-[2-(N',N'-diethylamino)ethyl]carbamoyl]-quinoline -6-yl-carbamate (RF). Subsequently, the derivatized N-glycans were purified by solid phase extraction based on hydrophilic interactions. The purified samples were analyzed by high-performance ultrahigh-performance liquid chromatography based on hydrophilic interactions with a fluorescent detector (HILIC-UPLC-FLR).

The chromatograms obtained as a result of the described analysis were manually integrated according to the associated separated glycan groups; see for example Figure 1. The amount of glycans in each glycan group (GP) is expressed as a percentage of the total integrated area (% area) to allow relative quantification of immunoglobulin G N-glycans. Chromatographic peaks containing glycan groups correspond to glycan structures described in the literature reference 8.

The suitability of the UHPLC system for N-glycan analysis was controlled during each analytical run using an in-house prepared 2-aminobenzamide (2AB)-labeled immunoglobulin G N-glycan standard as described in literature reference 9.

A detailed description of the experimental part of the study of the variability of IgG N-glycans GP1-GP22 in women aged 45-55 years is described in Example 1.

Statistical processing of the study results and the formation of numerical models for the diagnosis of perimenopause and menopause

In order to be able to compare the areas under the chromatogram peaks of different samples, the relative areas under the chromatogram peaks were calculated in such a way that the area under the chromatogram peak was divided by the total chromatogram area. The obtained relative areas were logit transformed according to equation (1), which enabled the approximation of the distribution of relative areas by a normal distribution. The batch effect on the measurements was removed using the ComBat method (R package "all"); see literature reference 11:

[image]

11) J. T. Leek, W. E. Johnson, H. S. Parker, A. E. Jaffe, J. D. Storey: The sva package for removing batch effects and other unwanted variation in high-throughput experiments, Bioinformatics 28 (2012) 882-883; doi:10.1093/bioinformatics/bts034.

The data thus obtained were used in all further analyses. A linear mixed model (R package "lme4") was used to determine the relationship between menopause and immunoglobulin G N-glycome; see literature reference 12:

12) D. Bates, M. Mächler, B. Bolker, S. Walker: Fitting Linear Mixed-Effects Models Using lme4, J. Stat. Software. 67 (2015) doi:10.18637/jss.v067.i01.

For each glycan structure of N-glycomes, a model was adapted in which the dependent variable was the logit transformed relative glycan area, and the fixed factors were menopause status, and age nested in the menopause status factor so that for each group of women (depending on menopause status) evaluated the effect of age on glycan change. The dependence of individual measurements, as a consequence of the design of the study in which individual women were sampled from one to three times and belong to the same family (twins), was controlled by random effects, among which were the unique identifier of the subject - nested in the unique identifier of the family as random sections, and age as a random slope. Using the fitted model, the age-corrected mean relative area (and corresponding 95% confidence interval) was estimated for samples sampled from pre- and postmenopausal women, and the estimated means were compared by post-hoc t-test with adjustment for Benjamini multiple testing. -Hochberg methods (R package "emmeans"); see literature references 13 and 14:

13) R. Lenth: Emmeans: Estimated Marginal means, aka Least-Squares Means. R Package Version 1.5.4.; see link:

https://cran.r-project.org/web/packages/emmeans/index.html;

14) Y. Benjamini, Y. Hochberg: Controlling the False Discovery Rate: A Practical and Powerful Approach to Multiple Testing, J.R. Stat. Soc. Sir. B 57 (1995) 289-300.

doi:10.1111/j.2517-6161.1995.tb02031.x.

Mean annual changes in relative areas (and corresponding 95% confidence intervals) were also estimated for samples collected from pre- and postmenopausal women. Comparisons were also made by post-hoc t-test with adjustment for multiple testing according to the Benjamini-Hochberg method described in references 13 and 14.

The obtained results are presented graphically in which the estimated mean value is shown as a point, and the corresponding 95% confidence interval is shown with error bars. The statistical significance of the difference between the values obtained by sampling before and after the onset of menopause is indicated on the graphic display as a p value (corrected for multiple testing) above the displayed mean values (R package "ggplot2"); see literature reference 15:

15) H. Wickham: ggplot2: Elegant Graphics for Data Analysis (2016) Springer-Verlag, New York, USA.

Development of a numerical model for the diagnosis of menopause and determination of the perimenopause period

Model A: model based on the N-glycome of a single sample. Considering the age of women at the time of onset of menopause, only samples obtained from women aged between 45 and 55 were used for the development of the model, which enables the diagnosis of menopause based on the N-glycome of immunoglobulin G. The data of the analyzed samples are divided into a subset for model training and a subset for model testing. The subset for model training is based on the random selection of one measurement from each family with the aim of removing mutual dependence between samples. The model testing subset contained all remaining data. An L1-regularized logistic model was adapted, which has menopause status as a dependent variable (dichotomous variable – "yes" or "no"), while as an independent variable it takes the logit transformed [equation (1)] of the relative areas under all peaks of the N-glycome chromatogram IgG; see literature reference 16:

16) J. Friedman, T. Hastie, R. Tibshirani: Regularization Paths for Generalized Linear Models via Coordinate Descent, J. Stat. Software. 33 (2010) 1-22; doi:10.18637/jss.v033.i01.

L1-regularization, also known as Lasso regularization, was applied with the aim of preventing overlearning and reducing the complexity of the final model. The tenfold cross-validation method was applied to calculate the coefficients of the model's independent variables on the training subset of the model. Hyperparameter[image] was used to reduce the number of predictors to 4 or less (R package "caret"); see literature reference 17:

17) M. Kuhn: caret: Classification and Regression Training (2020). R package version 6.0-86. https://CRAN.R-project.org/package=caret.

The final model formula calculated the probability that N-glycome IgG came from the population of perimenopausal and menopausal women for each sample from the model testing subset. The obtained predictions of menopause and the actual statuses of menopause were analyzed with ROC (Receiver Operating Characteristic) analysis (R package "pROC"), and the obtained results were presented with a ROC curve; see literature references 15 and 18:

18) X. Robin, N. Turck, A. Hainard, N. Tiberti, F. Lisacek, J.-C. Sanchez, M. Müller: pROC: an open-source package for R and S+ to analyze and compare ROC curves. BMC Bioinform. 12 (2011) 77. doi:10.1186/1471-2105-12-77.

The confidence interval (with a level of 95%) of the area under the ROC curve was determined by the bootstrap method with the number of samples of 2000. With the aim of demonstrating that the overall N-glycome IgG, and not the levels of individual structures, is responsible for the predictability of menopause. the complete procedure of this chapter was repeated two more times with the elimination of glycan structures as predictors that were selected in the final model of earlier iterations.

Model B: model based on the average annual change in the N-glycome of immunoglobulin G. The average annual change is calculated by dividing the difference of the logit-transformed relative areas under the peaks by the difference in age between sampling points expressed in years according to equation (2):

[image]

Considering the age of women at the time of onset of menopause, for the development of the model, which enables the diagnosis of menopause based on the average annual change in N-glycome IgG, only changes were used in which the second time point was obtained from women between the ages of 45 and 55, and age difference between the first and second time points that is less than 10 years. The data of the analyzed samples are divided into a subset for model training and a subset for model testing. The subset for model training is based on the random selection of one measurement from each family with the aim of removing mutual dependence between samples. The model testing subset contained all remaining data. An L1-regularized logistic model was adapted, which has menopause status as a dependent variable (dichotomous variable - "yes" or "no"), while as an independent variable it takes the average annual changes, according to equation (2), of the relative area under all peaks of the chromatogram N - glycoma IgG; see literature reference 16. The ten-fold cross-validation method was applied to calculate the coefficients of the model's independent variables on the model's training subset. Hyperparameter[image] was used to reduce the number of predictors to 4 or less (R package "caret"); see literature reference 17.

The final model formula calculated the probability that the average annual change in N-glycome IgG came from a population of perimenopausal and postmenopausal women for each measurement from the model testing subset. The obtained predictions of menopause and the actual statuses of menopause were analyzed by ROC (Receiver Operating Characteristic) analysis (R package "pROC"), and the obtained results were presented with a ROC curve; see literature references 15 and 18. The confidence interval (at the 95% level) of the area under the ROC curve was determined by the bootstrap method with the number of samples of 2000. With the aim of demonstrating that the entire N-glycome IgG is responsible for the predictability of menopause. , and not the levels of individual structures, the complete procedure of this chapter was repeated two more times with the elimination of glycan structures as predictors that were selected in the final model of earlier iterations.

Combination of Model A and Model B. The development of a model combining information on the average annual change in immunoglobulin G N-glycome and the level of IgG N-glycome structures sampled at a different time point was fitted on the same subset of data as the model based only on the average annual change in N-glycome IgG. The subset for testing the model was also equal to the subset for testing the model based only on the mean annual change in N-glycome IgG. An L1-regularized logistic model was adapted, which has menopause status as a dependent variable (dichotomous variable - "yes" or "no"), while as an independent variable it takes the average annual changes, according to equation (2), of the relative area under all peaks of the chromatogram N -glycoma IgG and logit transformed, according to equation (1), the relative area under all peaks of the chromatogram of N-glycoma IgG sampled at the second time point; see literature reference 16. The ten-fold cross-validation method was applied to calculate the coefficients of the model's independent variables on the model's training subset. Hyperparameter[image] was used to reduce the number of predictors to 5 or less (R package "caret"); see literature reference 17.

The final model formula calculated the probability that the measured N-glycome was from the population of perimenopausal and menopausal women for each measurement from the model testing subset. The obtained predictions of menopause and the actual statuses of menopause were analyzed by ROC (Receiver Operating Characteristic) analysis (R package "pROC"), and the results were presented with a ROC curve; see literature references 15 and 18. The confidence interval (at the 95% level) of the area under the ROC curve was determined by the bootstrap method with the number of samples of 2000. With the aim of demonstrating that the entire N-glycome IgG is responsible for the predictability of menopause. , and not the levels of individual structures, the complete procedure of this chapter was repeated two more times with the elimination of glycan structures as predictors that were selected in the final model of earlier iterations.

Results of statistical analysis

Of a total of 6032 samples, 5080 samples came from women with known menopausal status at the time of sampling: 185 women sampled at one time point, 370 women sampled at two time points, and 1385 women sampled at three time points. The remaining samples came from either men or women with unknown menopausal status. The average age of onset of menopause, for women whose menopause was included in the sampling, is 48.6 years, with a standard deviation (SD) of 6.0 years. Causes from non-menopausal women were sampled at an average age of 43.1 ± 7.0 years, while samples from postmenopausal women were sampled at an average age of 62.8 ± 8.2 years.

Relationship between N-glycome IgG and menopause. The number of successfully determined IgG N-glycan profiles in samples sampled from women with known menopausal status is described in Table 3.

Table 3. Samples with which the N-glycome of immunoglobulin G was determined.

[image]

* The total number of test subjects/families may be less than or equal to the sum of the number of test subjects/families in the group of women who are and in the group of women who are not in menopause because an individual test subject/family may have a sample before and after the onset of menopause.

Linear mixed model analysis showed that for most glycan structures of N-glycome IgG there is a difference in the average relative area under the peaks of the chromatogram, depending on whether the sample was sampled before or after the onset of menopause; see Figure 2A (GP1-GP11) and Figure 2B (GP12-GP22). The results also indicate a difference in the average annual change in the relative area under the chromatogram peaks depending on whether the second sample was sampled before or after the onset of menopause; see Figure 3A (GP1-GP11) and Figure 3B (GP12-GP22).

Diagnosis of menopause

Menopause can be diagnosed using the IgG N-glycan profile. L1-regularized logistic model based on one-sample N-glycome. The total number of samples used to develop the N-glycome IgG-based diagnostic test is shown in Table 4. Descriptions of the subsets used for model training and testing are shown in Tables 5 and 6.

Table 4. All samples sampled from women aged between 45 and 55 years.

[image]

Table 5. A randomly selected subset of data (one sample per family) for training a model based on the amount of glycan structures of the immunoglobulin G N-glycome.

[image]

Table 6. Data subset for testing the model based on the amount of glycan structures of the immunoglobulin G N-glycome and adapted to the training subset.

[image]

The obtained results show that using N-glycome IgG it is possible to determine the probability that the period of perimenopause has passed and that menopause has occurred. Using equation (3), it is possible to calculate the probability that menopause has occurred in the tested woman based on the 4 peaks of the IgG N-glycome chromatogram:

[image]

By ROC analysis (Receiver Operating Characteristic analysis), the probabilities obtained by equation (3) on the testing subset give a ROC curve that closes the middle surface of [image] ([image] ), and is shown in Figure 4(A).

The selected peaks in equation (3) are the result of L1-regularization with the aim of simplifying the model for calculating the results. Simplifying the model by dropping predictors is possible due to the significant correlation between the relative areas under the chromatogram peaks that correspond to the relative amounts of individual glycan structures in the IgG N-glycome. This fact is confirmed by the possibility of defining other models, by selecting some other structures from the N-glycome of IgG, which by ROC analysis give ROC curves with a slightly smaller area. For example, the probability obtained by equation (4) results in the mean area under the ROC curve[image] ([image] ), and by equation (5) results in the mean area under the ROC curve[image] ([image] ). ROC curves for the given examples are shown in Figure 4 (B,C).

[image]

[image]

L1-regularized logistic model based on change in N-glycome between two time points. The total number of measurements used to develop the N-glycome IgG-based diagnostic test is shown in Table 7. Descriptions of the subsets used for model training and testing are also given in Tables 8 and 9.

Table 7. Distribution of all measurements obtained from samples sampled at two time points, where the age of women at the second time point is between 45 and 55 years, the time gap between these two time points is less than 10 years, and none of the subjects was in menopause in at the moment of the first sampling.

[image]

Table 8. Randomly selected subset of data (one measurement per family) for training a model based on the change in the amount of glycan structures of the N-glycome of IgG.

[image]

Table 9. Data subset for testing the model based on the change in the amount of glycan structures of the N-glycome of IgG and adapted to the training subset.

[image]

The obtained results show that using the average annual change in N-glycome IgG, it is possible to determine the probability that the perimenopause period has passed and that menopause has occurred. Using equation (6), it is possible to calculate the probability that menopause has occurred in the tested woman based on the 4 peaks of the IgG N-glycome chromatogram.

[image]

By ROC analysis (Receiver Operating Characteristic analysis), the probabilities obtained by equation (6) on the testing subset give a ROC curve that closes the middle surface of [image] ([image] ), and is shown in Figure 5 (A).

The selected peaks in equation (6) are the result of L1-regularization with the aim of simplifying the model for calculating the results. Simplification of the model by dropping the predictor is possible due to the significant correlation between the relative mean annual changes of the areas under the chromatogram peaks. This fact is confirmed by the possibility of defining other models, by selecting some other structures from the N-glycome of IgG, which, by ROC analysis, give ROC curves with a slightly smaller area. For example, probabilities obtained by equation (7) result in the mean area under the ROC curve[image] ([image] ), and equation (8) results in the mean area under the ROC curve[image] ([image] ). ROC curves for the given examples are shown in Figure 5 (B,C).

[image]

[image]

L1-regularized logistic model based on change in N-glycome between two time points and N-glycome of another time point. The total number of measurements used to develop the N-glycome IgG-based diagnostic test is shown in Table 7. Descriptions of the subsets used for model training and testing are also given in Tables 8 and 9.

The obtained results show that by combined application of the average annual change of N-glycome IgG and the N-glycome profile of the second sample, it is possible to determine the probability that the perimenopause period has passed and that menopause has occurred. Using equation (9), it is possible to calculate the probability that menopause has occurred in the tested woman based on the 4 peaks of the IgG N-glycome chromatogram.

[image]

By ROC analysis (Receiver Operating Characteristic analysis), the probabilities obtained by equation (9) on the testing subset give a ROC curve that closes the middle surface of [image] ([image] ), and is shown in Figure 6 (A).

The selected peaks in equation (9) are the result of L1-regularization with the aim of simplifying the model for calculating the results. Simplification of the model by dropping the predictor is possible due to the significant correlation between the relative mean annual changes of the areas under the chromatogram peaks. This fact is confirmed by the possibility of defining other models, by selecting some other structures from the N-glycome of IgG, which, by ROC analysis, give ROC curves with a slightly smaller area. For example, the probability obtained by equation (10) results in the mean area under the ROC curve[image] ([image] ), and by equation (11) it results in the mean area under the ROC curve[image] ([image] ). ROC curves for the given examples are shown in Figure 6 (B,C).

[image]

[image]

Interpretation of model results

With the proposed models, based on N-glycome and/or change of N-glycome IgG as a result, we get a number between 0 and 1, which corresponds to the estimated probability that the woman whose sample is in menopause at the time of taking the sample. A value of 0.5 was determined as the limit value. Women with an estimated probability of less than 0.5 are declared not to be in menopause, while women with an estimated probability of more than 0.5 are declared to be in menopause.

A detailed description of the statistical processing of the results of the described study of the variability of IgG N-glycans GP1-GP22 in women aged 45-55 and the procedures for generating numerical models used for the diagnosis of menopause according to the invention are described in Example 2.

The use of the diagnostic procedure according to the invention and associated numerical models in clinical practice

The diagnostic procedure according to the invention is used to determine whether the examined female person has gone through perimenopause and entered the menopause phase. Alternatively, the diagnostic procedure according to the invention is used to determine whether the examined female person has entered the perimenopause phase.

1.7 Examples of the implementation of the invention

General notes

Names of IgG N-glycans that are key to the present invention FA1, A2, A2B, FA2G0, M5, FA2BG0, A2G1[6], A2G1[3], FA2G1[6], A2G1[3], FA2BG1[6], FA2BG1[ 3].

The meanings of the abbreviations used are as follows:

PNGase F = peptide-N4-(N-acetyl-beta-glucosaminyl)asparagine amidase enzyme;

2AB = 2-aminobenzamide;

DMF = N,N-dimethylformamide, solvent;

EDTA = N,N,N',N'-ethylenediamine-tetraacetic acid, disodium salt dihydrate;

PR = procainamide;

RF = 2,5-dioxopyrrolidin-1-yl-[2N-[2-(N',N'-diethylamino)ethyl] carbamoyl]-quinolin-6-yl-carbamate (RapiFluor-MS);

SD = standard deviation;

e = Euler's number;

HILIC = liquid chromatography with hydrophilic interaction, comes from the English term "hydrophilic interaction liquid chromatography";

IgG = immunoglobulin G;

UPLC = ultra-high performance liquid chromatography, comes from the English term "ultra-high performance liquid chromatography."

Example 1. A study of monitoring the variability of N-glycans bound to immunoglobulin G in women aged 45 to 55 years, some of whom have entered menopause and some who have not

The TwinsUK registry was used in the study of the variation of N-glycans GP1-GP22 bound to IgG from the blood plasma of women. Whole blood samples were collected by the TwinsUK registry at multiple time points (minimum 1, maximum 3 per person) over 20 years. Whole blood was collected in an EDTA tube and mixed well. The test tube was then left at room temperature and then centrifuged to separate the plasma. The plasma was transferred to a clean test tube and stored in a freezer at -80 or -20 ̊C.

A total of 6032 samples were analyzed:

(i) 1865 individuals at three time points (5595 samples in total);

(ii) 156 individuals at two time points (312 samples in total); you

(iii) 125 individuals at one time point (125 samples).

Blood samples were taken close to the time of collection of phenotype information. The samples arrived in about 60 10x10 boxes and were not randomized. Computer and then manual randomization was performed on 67 tiles, in three series (approximately 20 tiles per series) with intervals of two weeks in order to minimize the probability of error due to long-term manual randomization. Each of the 67 plates contains 96 wells that are filled according to the following schedule: 90 samples from individuals from the TwinsUK cohort, five standard samples (so-called standard plasma samples, internal standards) and one sample of ultrapure water (so-called blank sample). During randomization, the type of sample (patient or control), gender and age were taken into account, since each of the 67 tiles must describe the analyzed population equally well. Five standard samples per plate were used to correct for batch differences (the so-called batch effect).

Isolation of immunoglobulin G from human plasma. Isolation of N-glycans bound to immunoglobulin G in human plasma, which precedes glycan analysis with the HILIC-UPLC-FLR system, consists of two parts: isolation of immunoglobulin G from human plasma with a protein G plate and deglycosylation, labeling and purification of N-glycans using GlycoWorks RapiFluor-MS N-glycan kit. Immunoglobulin G was isolated from human plasma samples using a protein G monolithic 96-well plate (Bia separations, Slovenia) according to the protocol described in literature references 3 and 9. After isolation of immunoglobulin G using a protein G plate, an appropriate volume of immunoglobulin G eluate (average mass 15 µg) was aliquoted into a PCR plate (Starlab, UK) and dried in a vacuum centrifuge.

Rapid deglycosylation of immunoglobulin G. 5% (w/v) RapiGest SF surfactant solution was prepared by dissolving the contents of two vials of RapiGest SF (each vial has a mass of 10 mg). Each vial was dissolved in 200 µL of 5-fold concentrated GlycoWorks Rapid buffer. Both prepared RapiGest SF solutions were combined in one vial, mixed and aliquoted into PCR tubes. The dried eluate of immunoglobulin G was resuspended in 10.8 µL of ultrapure water, and 3 µL of 5% RapiGest SF solution was added to each sample and mixed with a pipette. The PCR sample plate was sealed with 8 connected caps and incubated for 3 minutes at 99 °C to denature immunoglobulin G, after which the plate was cooled for 3 minutes at room temperature. A volume of 1.2 µL of GlycoWorks Rapid PNGase F was added to each sample and mixed with a pipette. The PCR plate was closed with the aforementioned caps and incubated for 5 minutes at 50 °C to carry out deglycosylation. After that, the plate was left at room temperature for 3 minutes to cool down.

Rapid RapiFluor-MS labeling of N-glycans. The glycan labeling dye was prepared during the deglycosylation step by dissolving four vials containing 23 mg of RapiFluor-MS Reagent powder in 168 µL of anhydrous DMF. All four vials were combined into one, vortexed, and aliquoted into PCR tubes. From the mentioned tubes, 6 µL of the RapiFluor-MS reagent solution was immediately added to each sample and mixed by resuspending the pipette. The PCR plate with the samples was covered and left for 5 minutes at room temperature, in order to carry out the labeling reaction. After the past 5 minutes, 179 µL of acetonitrile (Honeywell, USA) was added to the samples, and the resulting mixture was immediately transferred to a one-milliliter microtiter plate with 96 wells.

Purification of N-glycans by solid phase extraction based on hydrophilic interactions. In order to precondition the GlycoWorks HILIC µElution plate, 200 µL of ultrapure water and 200 µL of a mixture of ultrapure water and acetonitrile (15:85, v/v) were added three times to each well. Then the excess liquid was removed using a vacuum manifold (Pall, USA). Acetonitrile-diluted samples were applied to a µElution plate and vacuum removed. Each plate was washed twice with 600 µL microliters of a mixture of methanoic acid (Merck, USA), ultrapure water and acetonitrile (1:9:90, v/v/v). The waste rack was then replaced with a clean 0.8 ml round-bottom 96-well collection plate (Waters, USA). The samples were eluted in three steps with three times 30 µL of SPE Elution buffer (200 mmol/L mixture of ammonium acetate and acetonitrile, 95:5, v/v, pH 7), and all three eluted fractions were collected in the same 0.8 military plate. For sample dilution, 310 µL of sample dilution solution (Sample Diluent; mixture of DMF and acetonitrile, 32:68, v/v) was added to each sample and resuspended with a pipette. The final volume per well was 400 µL. A volume of 40 µL of each sample was transferred to vials for ultra high performance liquid chromatography based on hydrophilic interactions with fluorescence detection (HILIC-UHPLC-FLR) analysis, while the remaining volume was stored at -20 °C.

HILIC-UHPLC-FLR analysis of immunoglobulin G N-glycans RapiFluor-MS labeled N-glycans were analyzed on a Waters Acquity UPLC H-class instrument using Waters UPLC Glycan BEH amide chromatography columns (130 Å, 1.7 µm BEH particles, 2.1x100 mm). 50 mmol/L ammonium formate, pH 4.4 was used as solvent A, and 100% acetonitrile of LC-MS class was used as solvent B according to the method described in literature reference 8. The method was adapted to the analysis of these samples in such a way that it was used linear gradient of 75-61.5% acetonitrile (v/v) at a flow rate of 0.4 mL/min over 30 minutes. The entire analytical run of one sample takes 42 minutes. The chromatograms obtained as a result of the described analysis were manually integrated according to the associated separated glycan groups; see Figure 1. The amount of glycans in each glycan group is expressed as a percentage of the total integrated area (%Area) to allow relative quantification of immunoglobulin G N-glycans. Chromatographic peaks containing glycan groups correspond to the glycan structures described in literature reference 8.

The suitability of the UHPLC system for N-glycan analysis was controlled during each analytical run using an in-house prepared 2-aminobenzamide (2AB)-labeled immunoglobulin G N-glycan standard as described in literature reference 9.

Example 2. Statistical processing of the results obtained from the study of the variation of IgG N-glycan GP1-GP22 in women aged 45-55 years, some of whom have entered the menopause phase and some who have not, and the formation of numerical models for the diagnosis of perimenopause and menopause according to the invention

In order to be able to compare the areas under the chromatogram peaks of different samples, the relative areas under the chromatogram peaks were calculated in such a way that the area under the chromatogram peak was divided by the total chromatogram area. The obtained relative areas were logit transformed according to equation (1), which enabled the approximation of the distribution of relative areas by a normal distribution. The batch effect on the measurements was removed using the ComBat method (R package "all"); see literature reference 11. The data thus obtained were used in all further analyses. When determining the relationship between menopause and N-glycoma IgG, a linear mixed model was used (R package "lme4"); see literature reference 12.

For each glycan structure of N-glycomes, a model was adapted in which the dependent variable was the logit transformed relative glycan area, and the fixed factors were menopause status, and age nested in the menopause status factor so that for each group of women (depending on menopause status) evaluated the effect of age on glycan change. The dependence of individual measurements, as a consequence of the design of the study in which individual women were sampled from one to three times, and belong to the same family (twins), is controlled by random effects, among which are the unique identifier of the subject - nested in the unique identifier of the family as random sections, and age as a random slope. Using the fitted model, the age-corrected mean relative area (and corresponding 95% confidence interval) was estimated for samples sampled from pre- and postmenopausal women, and the estimated means were compared by post-hoc t-test with adjustment for Benjamini multiple testing. -Hochberg methods (R package "emmeans"); see references 13 and 14.

Mean annual changes in relative areas (and corresponding 95% confidence intervals) were also assessed for samples taken from pre- and postmenopausal women, comparisons were also made by post-hoc t-test with adjustment for multiple testing according to the Benjamini-Hochberg method described in literary references 13 and 14.

The obtained results are presented graphically in which the estimated mean value is shown as a point, and the corresponding 95% confidence interval is shown with error bars. The statistical significance of the difference between the values obtained by sampling before and after the onset of menopause is indicated on the graphic display as a p value (corrected for multiple testing) above the displayed mean values (R package "ggplot2"); see literature reference 15.

Development of a numerical model for the diagnosis of menopause

Model A: model based on the N-glycome of a single sample. Considering the age of women at the time of onset of menopause, only samples obtained from women aged between 45 and 55 were used for the development of the model, which enables the diagnosis of menopause based on the N-glycome of immunoglobulin G. The data of the analyzed samples are divided into a subset for training the model and a subset for testing the model. The subset for model training is based on the random selection of one measurement from each family with the aim of removing mutual dependence between samples. The model testing subset contained all remaining data. An L1-regularized logistic model was adapted, which has menopause status as a dependent variable (dichotomous variable – "yes" or "no"), while as an independent variable it takes the logit transformed [equation (1)] of the relative areas under all peaks of the N-glycome chromatogram IgG; see literature reference 16.

L1-regularization, also known as Lasso regularization, was applied with the aim of preventing overlearning and reducing the complexity of the final model. The tenfold cross-validation method was applied to calculate the coefficients of the model's independent variables on the training subset of the model. Hyperparameter[image] was used to reduce the number of predictors to 4 or less (R package "caret"); see literature reference 17. The final model formula calculated the probability that N-glycome IgG came from the population of menopausal women for each sample from the model testing subset. The obtained predictions of menopause and the actual statuses of menopause were analyzed by ROC (Receiver Operating Characteristic) analysis (R package "pROC"), and the obtained results were presented with a ROC curve [5,8]; see references 15 and 18.

The confidence interval (with a level of 95%) of the area under the ROC curve was determined by the bootstrap method with the number of samples of 2000. With the aim of demonstrating that the overall N-glycome IgG, and not the levels of individual structures, is responsible for the predictability of menopause. the complete procedure of this chapter was repeated two more times with the elimination of glycan structures as predictors that were selected in the final model of earlier iterations.

Model B: model based on average annual change in immunoglobulin G N-glycome. Average annual change was calculated by dividing the difference of the logit-transformed relative areas under the peaks by the difference in age between sampling points expressed in years according to equation (2). Considering the age of women at the time of onset of menopause, for the development of the model, which allows determining the passage through perimenopause and the diagnosis of menopause based on the average annual change in N-glycome of immunoglobulin G, only changes were used in which the second time point was obtained from women aged between 45 and 55 years old, and the age difference between the first and second time points is less than 10 years. The data of the analyzed samples are divided into a subset for training the model and a subset for testing the model. The subset for model training is based on the random selection of one measurement from each family with the aim of removing mutual dependence between samples. The model testing subset contained all remaining data. An L1-regularized logistic model was adapted, which has menopause status as a dependent variable (dichotomous variable - "yes" or "no"), while as an independent variable it takes the average annual changes, according to equation (2), of the relative area under all peaks of the chromatogram N - glycoma IgG; see literature reference 16. The ten-fold cross-validation method was applied to calculate the coefficients of the model's independent variables on the model's training subset. Hyperparameter[image] was used to reduce the number of predictors to 4 or less (R package "caret"); see literature reference 17.

The final model formula calculated the probability that the average annual change in N-glycome IgG came from a population of menopausal women for each measurement from the model testing subset. The obtained predictions of menopause and the actual statuses of menopause were analyzed by ROC (Receiver Operating Characteristic) analysis (R package "pROC"), and the obtained results were presented with a ROC curve; see literature references 15 and 18. The confidence interval (at the 95% level) of the area under the ROC curve was determined by the bootstrap method with the number of samples of 2000. With the aim of demonstrating that the entire N-glycome IgG is responsible for the predictability of menopause. , and not the levels of individual structures, the complete procedure of this chapter was repeated two more times with the elimination of glycan structures as predictors that were selected in the final model of earlier iterations.

Combination of Model A and Model B. The development of a model combining information on the average annual change in immunoglobulin G N-glycome and the level of IgG N-glycome structures sampled at a different time point was fitted on the same subset of data as the model based only on the average annual change in N-glycome IgG. The subset for testing the model was also equal to the subset for testing the model based only on the mean annual change in N-glycome IgG. An L1-regularized logistic model was adapted, which has menopause status as a dependent variable (dichotomous variable - "yes" or "no"), while as an independent variable it takes the average annual changes, according to equation (2), of the relative area under all peaks of the chromatogram N -glycoma IgG and logit transformed, according to equation (1), the relative area under all peaks of the chromatogram of N-glycoma IgG sampled at the second time point; see literature reference 16. The ten-fold cross-validation method was applied to calculate the coefficients of the model's independent variables on the model's training subset. Hyperparameter[image] was used to reduce the number of predictors to 5 or less (R package "caret"); see literature reference 17.

The final model formula calculated the probability that the measured N-glycome was from the population of menopausal women for each measurement from the model testing subset. The obtained predictions of menopause and the actual statuses of menopause were analyzed by ROC (Receiver Operating Characteristic) analysis (R package "pROC"), and the results were presented with a ROC curve; see literature references 15 and 18. The confidence interval (at the 95% level) of the area under the ROC curve was determined by the bootstrap method with the number of samples of 2000. With the aim of demonstrating that the entire N-glycome IgG is responsible for the predictability of menopause. , and not the levels of individual structures, the complete procedure of this chapter was repeated two more times with the elimination of glycan structures as predictors that were selected in the final model of earlier iterations.

Results of statistical analysis

Of a total of 6032 samples, 5080 samples came from women with known menopausal status at the time of sampling: 185 women sampled at one time point, 370 women sampled at two time points, and 1385 women sampled at three time points. The remaining samples came from either men or women with unknown menopausal status. The average age of onset of menopause, for women whose menopause was included in the sampling, is 48.6 years, with a standard deviation (SD) of 6.0 years. Causes from non-menopausal women were sampled at an average age of 43.1 ± 7.0 years, while samples from postmenopausal women were sampled at an average age of 62.8 ± 8.2 years.

Relationship between N-glycome IgG and menopause. The number of successfully determined IgG N-glycan profiles in samples sampled from women with known menopause status is described in Table 3. Linear mixed model analysis showed that for most glycan structures of IgG N-glycomes there is a difference in the average relative area under the chromatogram peaks, depending on whether the sample was taken before or after menopause; see Figure 2A (GP1-GP11) and Figure 2B (GP12-GP22). The results also indicate a difference in the average annual change in the relative area under the chromatogram peaks depending on whether the second sample was sampled before or after the onset of menopause; see Figure 3A (GP1-GP11) and Figure 3B (GP12-GP22).

Diagnosis of menopause. Menopause can be diagnosed using the IgG N-glycan profile.

L1-regularized logistic model based on one-sample N-glycome. The total number of samples used to develop the N-glycome IgG-based diagnostic test is shown in Table 4. Descriptions of the subsets used for model training and testing are also given in Tables 5 and 6.

The obtained results show that by using N-glycome IgG it is possible to determine the probability that menopause has occurred. Using equation (3), it is possible to calculate the probability that menopause has occurred in the tested woman based on the 4 peaks of the IgG N-glycome chromatogram. By ROC analysis (Receiver Operating Characteristic analysis), the probabilities obtained by equation (3) on the testing subset give a ROC curve that closes the middle surface of [image] ([image] ), and is shown in Figure 4(A). The selected peaks in equation (3) are the result of L1-regularization with the aim of simplifying the model for calculating the results. Simplifying the model by dropping predictors is possible due to the significant correlation between the relative areas under the chromatogram peaks that correspond to the relative amounts of individual glycan structures in the IgG N-glycome.

This fact is confirmed by the possibility of defining other models, by selecting some other structures from the N-glycome of IgG, which by ROC analysis give ROC curves with a slightly smaller area. For example, the probability obtained by equation (4) results in the mean area under the ROC curve[image] ([image] ), and by equation (5) results in the mean area under the ROC curve[image] ([image] ). ROC curves for the given examples are shown in Figure 4 (B,C).

L1-regularized logistic model based on change in N-glycome between two time points. The total number of measurements used to develop the immunoglobulin G N-glycome-based diagnostic test is shown in Table 7. Descriptions of the subsets used for model training and testing are given in Tables 8 and 9.

The obtained results show that using the average annual change of N-glycome IgG, it is possible to determine the probability that menopause has occurred. Using equation (6), it is possible to calculate the probability that menopause has occurred in the tested woman based on the 4 peaks of the IgG N-glycome chromatogram. By ROC analysis (Receiver Operating Characteristic analysis), the probabilities obtained by equation (6) on the testing subset give a ROC curve that closes the middle surface of [image] ([image] ), and is shown in Figure 5 (A). The selected peaks in equation (6) are the result of L1-regularization with the aim of simplifying the model for calculating the results. Simplifying the model by dropping predictors is possible due to the significant correlation between the relative mean annual changes of the areas under the chromatogram peaks.

This fact is confirmed by the possibility of defining other models, by selecting some other structures from the N-glycome of IgG, which, by ROC analysis, give ROC curves with a slightly smaller area. For example, probabilities obtained by equation (7) result in the mean area under the ROC curve[image] ([image] ), and equation (8) results in the mean area under the ROC curve[image] ([image] ). ROC curves for the given examples are shown in Figure 5 (B,C).

L1-regularized logistic model based on change in N-glycome between two time points and N-glycome of another time point. The total number of measurements used to develop the N-glycome IgG-based diagnostic test is shown in Table 7. Descriptions of the subsets used for model training and testing are also given in Tables 8 and 9.

The obtained results show that by combined application of the average annual change of N-glycome IgG and the N-glycome profile of the second sample, it is possible to determine the probability that menopause has occurred. Using equation (9), it is possible to calculate the probability that menopause has occurred in the tested woman based on the 4 peaks of the IgG N-glycome chromatogram. By ROC analysis (Receiver Operating Characteristic analysis), the probabilities obtained by equation (9) on the testing subset give a ROC curve that closes the middle surface of [image] ([image] ), and is shown in Figure 6 (A).

The selected peaks in equation (9) are the result of L1-regularization with the aim of simplifying the model for calculating the results. Simplification of the model by dropping the predictor is possible due to the significant correlation between the relative mean annual changes of the areas under the chromatogram peaks.

This fact is confirmed by the possibility of defining other models, by selecting some other structures from the N-glycome of IgG, which, by ROC analysis, give ROC curves with a slightly smaller area. For example, the probability obtained by equation (10) results in the mean area under the ROC curve[image] ([image] ), and by equation (11) it results in the mean area under the ROC curve[image] ([image] ). ROC curves for the given examples are shown in Figure 6 (B,C).

Interpretation of numerical model results.

With the proposed models, based on N-glycome and/or change of N-glycome IgG as a result, we get a number between 0 and 1, which corresponds to the estimated probability that the woman whose sample is in menopause at the time of taking the sample. A value of 0.5 was determined as the limit value. Women with an estimated probability of less than 0.5 are declared not to be in menopause, while women with an estimated probability of more than 0.5 are declared to be in menopause.

The state of perimenopause is generally characterized by a significantly milder disturbance of the normal concentrations of sex hormones that regulate the process of the menstrual cycle and, consequently, a milder spectrum of symptoms compared to the state of full menopause; see literature references 6 and 7. Although the numerical models according to the invention are not able to unambiguously distinguish the state of perimenopause from the state of full menopause, it is understandable to the average expert in the field that the degree of changes in the concentrations of key IgG N-glycans in the perimenopause phase is very likely milder compared to the state of full menopause . Because of this, the diagnostic method in question has a certain predictive value for determining the state of perimenopause.

Conclusion

The subject invention discloses a diagnostic method for determining whether a female subject has entered the perimenopause or menopause phase based on the quantitative analysis of N-glycans bound to immunoglobulin G (IgG) from her blood plasma, i.e., a blood sample.

The development of said diagnostic method was made possible by the results of a study conducted on a group of women aged 45-55. This study showed that IgG-bound N-glycans, labeled GP1-GP22, change significantly when women enter the perimenopause and menopause phases. Based on the statistical processing of the results, the possibility of forming several different numerical models that can be used for diagnostics to determine whether a woman has entered the perimenopause or menopause phase or not, based on a single blood sample, was established. All the mentioned numerical models have a high diagnostic potential for the detection of perimenopause or menopause based on the results of quantitative analysis of IgG N-glycan from the blood sample of the examined female person. The two relatively most predictive numerical models were preferentially selected.

1.8 Industrial Applicability

The subject invention discloses a diagnostic procedure for determining whether the examined female person has entered the perimenopause or menopause phase based on the quantitative analysis of N-glycans bound to IgG from blood plasma. Therefore, the industrial applicability of the subject invention is unquestionable.

Claims (9)

1. The procedure for diagnosing menopause and determining the period of perimenopause through the analysis of N-glycans bound to immunoglobulin G <IgG> from the blood plasma of women, denoted by the abbreviations GP1 to GP22 of the general chemical structure I: [image] [image] [image] [image] characterized by the fact that said procedure includes: (i) quantitative analysis of one or more: (a) fluorescently derivatized glycans released from IgG by the enzyme peptide-N4-<N-acetyl-beta-glucosaminyl>asparagine amidase <PNGase F>; or (b) free glycans or corresponding glycopeptides or glycoforms; (ii) inclusion of the obtained results of quantitative proportions of said glycans in one or more numerical models that were previously obtained by statistical processing of the results of the study of variations in the quantitative proportion of IgG glycans in blood plasma on a selected group of women, some of whom were in menopause and some who were not; you (iii) calculating the result in the form of a number that describes the probability that the examined female person has passed through perimenopause and entered the menopause phase. 2. The procedure according to claim 1, characterized by the fact that the selected group of women from step (ii) was between the ages of 45 and 55, and that the said procedure is applied to subjects belonging to that age range. 3. Procedure according to claim 1 or 2, characterized in that the result from step (iii) is interpreted in the following way: (a) if the score is between 0.5 and 1.0, the female subject has passed through perimenopause and entered the menopause phase; And (b) if the result is between 0 and 0.5, the examined female person has not gone through perimenopause and entered the menopause phase. 4. The procedure according to claims 1-3, characterized in that the probability [image] 4. that the examined female person has gone through perimenopause and entered the menopause phase is calculated using a numerical model: [image] where are they: [image] , [image] , [image] and [image] logit transformed value of the relative area under the peaks of glycans of the same name GP2, GP4, GP12 and GP22 from the chromatogram of the chosen quantitative analytical technique, where the logit function is defined as: [image] . 5. The procedure according to claims 1-3, characterized in that the probability [image] 5. that the examined female person has gone through perimenopause and entered the menopause phase is calculated using the formula: [image] where are they: - [image] logit transformed value of the relative area under the GP12 glycan peak from the chromatogram of the appropriate quantitative analytical technique, where the logit function is defined as: [image] And - [image] , [image] , [image] and [image] average annual changes in the Logit transformed values of the relative area under the peaks of glycans of the same name G11, G12, G13 and G16 from the chromatogram of the corresponding quantitative analytical technique defined as: [image] . 6. The method according to any of the preceding claims, characterized in that the glycans are derivatized with reagents selected from the group consisting of: (i) combination of a suitable aromatic amine such as 2-amino benzamide <2AB> or procainamide (PR) and some suitable reductant for reductive amination: 2-picoline borane complex <BH3•NC5H4-2-CH3> or sodium cyanoborohydride <NaBH3CN>; [image] or (ii) 2,5-dioxopyrrolidin-1-yl-<2N-<2-<N',N'-diethylamino>ethyl> carbamoyl>-quinolin-6-yl-carbamate <RF>: [image] . 7. The method according to any of the preceding claims, indicated by the fact that the analytical technique is selected from the group consisting of: ultra-high efficiency liquid chromatography <UPLC>; MALDI-TOF mass spectrometry; liquid chromatography coupled with mass spectrometry <LC-MS>; or capillary electrophoresis <CE>. 8. Use of a method according to any one of the preceding claims for determining whether a female subject has undergone perimenopause and entered the menopausal phase. 9. Use of a method according to any one of the preceding claims for determining whether a female subject has entered perimenopause.