Methods for diagnosis and monitoring NASH disease

Abstract

In the present invention, inventors used highly sensitive, classical and / or digital, OMICS technologies, to analyze combined markers profiles in serum of subjects with NAFLD and NASH, at the time of hospital admission or control subjects. Briefly, inventors through the PRECINASH study have access to an unprecedented number of liver and serum samples, from a cohort of > 1000 patients covering the full spectrum of disease progression, from normal liver to NASH. Thereafter, inventors used a supervised statistical approach to identify the biomarkers combinations that represent a reliable biomarker of NAFLD-related severity and elaborate a global Nash score in order to obtain an end value (global NASH Score or PreciNASH score), in order to predict liver-related morbi-mortality in NASH that will help to better define the population of risk, estimate the potential impact of therapeutic intervention on outcomes. This PreciNASH score set may be used as diagnostic or prognosis tool and thus set-up the basis for the development of a rapid functional specific test for NASH disease and also improved personalized patient management and monitoring.

Description

[0001] METHODS FOR DIAGNOSIS AND MONITORING NASH DISEASE FIELD OF THE INVENTION: The present invention relates to methods for diagnosis and monitoring of NASH disease (Non Alcoholic SteatoHepatitis). More specifically present invention relates to methods for diagnosis and prognosis of the severe form of Non-alcoholic fatty liver disease (NAFLD), through detection of specific serum proteins combination in a patient. BACKGROUND OF THE INVENTION: Non-alcoholic fatty liver disease (NAFLD) is one of the leading causes of chronic liver disease. Indeed, the prevalence of diabetes mellitus and obesity, the two main risk factors closely linked with NAFLD, raised worldwide since the nineties, provoking NAFLD expansion to reach a prevalence of 25% in the general population. Thus, nowadays, NAFLD constitutes an increasing cause of hepatocellular carcinoma and liver transplantation with growing public health repercussions. However, NAFLD's nosological spectrum varies significantly from mild histological damage as simple steatosis to severe progressive forms with a high risk of hepatic mortality such as NASH cirrhosis. Identifying patients with high related-liver morbidity and mortality risk or at risk to progress to such a state is crucial for care management. Screening fibrosis can be a good option because fibrosis is associated with liver-related complications or mortality. Identifying patients with established and advanced histological lesions does not consider the evolution kinetics of the disease. Indeed, NASH cannot be put aside as it constitutes NAFLD's evolutive form that will progress over time toward cirrhosis. Targeting fibrogenesis's leading cause appears as a potential medical strategy to reduce liver mortality at a population-based level. Furthermore, the evidence is piling up that NASH increases cardiovascular morbidity and mortality. Severe NAFLD and NASH, outside the liver disease context, worsen the cardiovascular event's risk after adjustment with classical metabolic syndrome features. The need to screen this subset of patients is therefore warranted. Besides, NASH is the target for future medical treatment. The F.D.A. (Food and Drug Administration) considers the reversal / resolution of NASH without worsening of fibrosis as a surrogate endpoint for phase IIb and III trials in patients with NASH and early fibrosis in the drug development strategy. Recently, a prospective cohort has shown that patients with resolution of NASH without worsening of fibrosis had a longer 15-year survival than patients with a persistence of NASH (after adjustment for fibrosis, age, gender, BMI, diabetes, hypertension and dyslipidemia). Thus, monitoring of NASH progression after treatment is essential for follow-up, therapeutic strategies and long-term outcomes. To date, only liver biopsy can diagnose NASH, its resolution, as well as its progression. Due to the high prevalence of NAFLD, the number of liver biopsies to perform appears colossal. The liver biopsy constraints make it unsuitable for screening at a population scale or the patients' follow-up. Indeed, liver biopsies are invasive, associated with a mortality risk of around 1 / 10000 (west ad card 2010), expensive, require an expert pathologist, and can also be associated with sample variability. Thus, Non-invasive biomarkers are warranted to manage the routine clinical practice. Research on this topic is growing, as shown by the recent development of a novel non-invasive algorithm score, the NIS4 using 4 blood biomarkers: miR-34-5p, alpha-2-macroglobulin, HbA1c, and YKL-40. However, the use of miRNA requires specific advanced technology that could limit the diffusion of the test. Thus, an ideal test involves, in addition to its optimal performances, to be reproducible and readily available everywhere. Besides, NASH is a dynamic disease with a progressive inflammatory status, which should be monitored. In terms of comparison, NAFLD necessitates a biomarker, as HbA1c for diabetes, permitting diagnosis, screening, and patient follow-up under therapy to reduce the need for liver biopsy and facilitate access to future medical treatment. For this purpose, an appropriate score or biomarker should be sensitive to change, which means that the biomarker as acceptable performance and still correlates with the results of the gold standard (liver biopsy) after treatment. Accordingly, there remains an unmet need in the art for specific and more rapid diagnostic test for NASH of NAFLD patients, reflecting directly the dysfunction of inflammatory process. This invention aimed to develop and validate a blood diagnosis test for NASH with serum biochemical markers associated with NAFLD's predictive value of the necro- inflammatory activity. The inventors therefore set up a diagnostic and monitoring method of the NASH disease and (severe form of NAFLD disease) that allows to directly reflect the necro- inflammatory status of the patient by using in a Global NASH score (or PreciNASH score) using specific combinations of serum biological biomarkers which could be combined with clinical marker. SUMMARY OF THE INVENTION: A first object of the present invention relates to an in vitro method for assessing a subject’s risk of having or developing NASH (non-alcoholic steatohepatitis) disease, comprising the steps of i) determining in a blood sample obtained from the subject the level of the 2 serum biological marker Galectin3 binding protein (Galectin3- BP) and cytokeratin-18 (CK-18) (protein markers), ii) combining said levels by implementing a statistical technique on said levels, said statistical technique involving a logistic regression in order to obtain an end value (global NASH Score) iii) comparing said score to a predetermined first reference value, and iv) when said score is higher than or equal to said first reference value is predictive of a high risk of having or developing NASH (non-alcoholic steatohepatitis) disease. In a particular embodiment, comprising comparing said score with a predetermined second reference value and concluding that when said score is lower than or equal to said second reference value is predictive of a low risk of having or developing a NASH (non- alcoholic steatohepatitis) disease In a particular embodiment wherein, in an additional step, determining in a blood sample obtained from the subject, the level of clinical marker gamma-glutamyltranspeptidase (gamma GT), the glycemia (Fasting Blood Glucose) higher than 7 mmol / L, and combining the levels of biological and clinical marker previously measured by implementing a statistical technique on said levels, said statistical technique involving a logistic regression in order to obtain an end value (global NASH Score). In a particular embodiment wherein, in an additional step, determining in a blood sample obtained from the subject, the level of 7-Ketocholesterol (7-KC) (lipid marker) and combining the levels of biological and clinical marker previously measured by implementing a statistical technique on said levels, said statistical technique involving a logistic regression in order to obtain an end value (global NASH Score or PreciNASH score). An additional object of the invention relates to an in vitro method for monitoring NASH (non-alcoholic steatohepatitis) disease comprising the steps of i) determining the level of the 2 serum biological marker Galectin3 binding protein (Galectin3- BP) and cytokeratin- 18 (CK-18) (protein markers) in a blood sample obtained from a subject at a first specific time of the disease, ii) combining said levels by implementing a statistical technique on said levels, said statistical technique involving a logistic regression in order to obtain an end value (global NASH Score) ,iii) determining the level of the same biological serum markers determined in step i) in a blood sample obtained from the subject at a second specific time of the disease, iv) combining said levels by implementing a statistical technique on said levels, said statistical technique involving logistic regression in order to obtain an end value, v) comparing the end value determined at step ii) with the end value determined at step iv) and vi) concluding that the disease has evolved in worse manner when the end value determined at step ii) is higher than the end value determined at step i). An additional object of the invention relates to an in vitro method for monitoring the treatment of a NASH (non-alcoholic steatohepatitis) disease comprising the steps of i) determining the level of the 2 serum biological marker Galectin3 binding protein (Galectin3- BP) and cytokeratin-18 (CK-18) (protein markers) in a blood sample obtained from a subject before the treatment, ii) combining said levels by implementing a statistical technique on said levels, said statistical technique involving a logistic regression in order to obtain an end value (global NASH Score) iii) determining the level of the same serum biological markers determined in step i) in a blood sample obtained from the subject after the treatment, iv) combining said levels by implementing a statistical technique on said levels, said statistical technique involving a logistic regression in order to obtain an end value v) comparing the end value determined at step iii) with the end value determined at step iv) and vi) concluding that the treatment is efficient when the end value determined at step iii) are lower than the end value determined at step iv). In a particular embodiment regarding the method for monitoring (the disease or the treatment) wherein in an additional step, determining in a blood sample obtained from the subject, the blood level of clinical marker gamma-glutamyltranspeptidase (gamma GT) and / or glycemia (Fasting Blood Glucose) higher than 7 mmol / L, and combining the blood levels of biological and clinical markers previously measured by implementing a statistical technique on said levels, said statistical technique involving a logistic regression in order to obtain an end value (global NASH Score). In a particular embodiment regarding the method for monitoring (the disease or the treatment) wherein, in an additional step, determining in a blood sample obtained from the subject, the level of 7-Ketocholesterol (7-KC) (lipid marker) and combining the levels of biological and clinical marker previously measured by implementing a statistical technique on said levels, said statistical technique involving a logistic regression in order to obtain an end value (global NASH Score). DETAILED DESCRIPTION OF THE INVENTION: In the present invention, inventors used highly sensitive, classical and / or digital, OMICS technologies, to analyze combined markers profiles in serum of subjects with NAFLD or control subjects. Briefly, inventors through the PRECINASH study have access to an unprecedented number of liver and serum samples, from a cohort of > 1000 patients covering the full spectrum of disease progression, from normal liver to NASH. Thereafter, inventors used a supervised statistical approach to identify the biomarkers combinations that represent a reliable biomarker of NAFLD-related severity and elaborate a global Nash score (by combining levels of said markers by implementing a statistical technique on said levels, said statistical technique involving a logistic regression) in order to obtain an end value (global NASH Score), in order to predict liver-related morbi-mortality in NASH that will help to better define the population of risk, estimate the potential impact of therapeutic intervention on outcomes. This non-invasive diagnostic test for NASH (global NASH Score or PreciNASH score) is validated in a second validation cohort. This global Nash score (or PreciNASH score) set may be used as diagnostic or prognosis tool. These results thus set-up the basis for the development of a rapid functional specific test for NASH disease and also improved personalized patient management through necro-inflammatory profiling. Diagnostic methods of NASH disease according to the invention The present invention relates to an in vitro method for assessing a subject’s risk of having or developing NASH (non-alcoholic steatohepatitis) disease, comprising the steps of i) determining in a blood sample obtained from the subject the level of the 2 protein serum marker Galectin3 binding protein (Galectin3-BP) and cytokeratin-18 (CK-18) (protein markers), ii) combining said levels by implementing a statistical technique on said levels, said statistical technique involving a logistic regression in order to obtain an end value (global NASH Score) iii) comparing said score to a predetermined first reference value, and iv) concluding that the subject has NASH (non-alcoholic steatohepatitis) when said score is higher than or equal to said first reference value is predictive of a high risk of having or developing NASH (non-alcoholic steatohepatitis) disease. The steps of the different methods of the invention which involves statistical technique involving a logistic regression are generally implemented by computer program. In a particular embodiment, comprising comparing said score with a predetermined second reference value and concluding that when said score is lower than or equal to said second reference value is predictive of a low risk of having or developing a NASH (non- alcoholic steatohepatitis) disease. In a particular embodiment wherein, in an additional step, determining in a blood sample obtained from the subject, the level of clinical marker gamma-glutamyltranspeptidase (gamma GT) and the glycemia (Fasting Blood Glucose) higher than 7 mmol / L (clinical marker) and combining the levels of biological and clinical markers previously measured by implementing a statistical technique on said levels, said statistical technique involving a logistic regression in order to obtain an end value (global NASH Score). In a particular embodiment wherein, in an additional step, determining in a blood sample obtained from the subject, the level of 7-Ketocholesterol (7-KC) (lipid marker) and combining the levels of biological and clinical marker previously measured by implementing a statistical technique on said levels, said statistical technique involving a logistic regression in order to obtain an end value (global NASH Score). In particular embodiments, a plurality of lipid biomarker and protein biomarkers (“Biomarker lipid”: 7-Ketocholesterol (7-KC) and / or “Biomarker protein”: Galectin3- BP (also named MAC2-binding protein (MAC2-BP) and cytokeratin-18) as well as clinical markers (Gamma GT and / or Glycemia) may be used in the methods of diagnostic / prognostic / monitoring / of the invention. In other words, the methods of the invention may comprise steps of: detecting in the biological sample the level serum marker (“Biomarker lipid”: 7- Ketocholesterol (7-KC) and / or “Biomarker protein”: Galectin3- BP and cytokeratin-18) and the level of 1 or 2 clinical biomarkers present in the biological sample; and detecting any biomarker of the invention. Accordingly, the methods of the present invention (diagnostic / prognosis / classification / monitoring) are performed using the level of 2, 3, 4 or 5 different biomarkers including the level of Galectin3 binding protein (Galectin3- BP), cytokeratin-18 (CK-18)) and optionally the level of gamma-glutamyltranspeptidase (gamma GT) and / or the level of glycemia (Fasting Blood Glucose) and / or the level of 7-Ketocholesterol (7-KC). In particular embodiments, the methods of the present invention (diagnostic / prognosis / classification / monitoring) are performed using the level of 4 or 5 different biomarkers including the level of Galectin3 binding protein (Galectin3- BP), cytokeratin-18 (CK-18) gamma-glutamyltranspeptidase (gamma GT) and glycemia (Fasting Blood Glucose) and optionally the level of 7-Ketocholesterol (7-KC). In preferred embodiments, the methods of the present invention (diagnostic / prognosis / classification / monitoring) are performed using the 2 different protein Biomarker including Galectin3- BP and cytokeratin-18, and the clinical biomarkers including Gamma GT and Fast Blood Glycemia. In the last step of the method of the invention, combining is obtained by implementing the statistical technique involving a mathematical function called binary (or ordinal) logistic regression or multiple linear regression with the following procedure: - first, the independent variables are tested by univariate analysis; - second, the independent variables that were significant in univariate analysis are tested in multivariable analysis by multivariable logistic regression model using a backward selection procedure with a removal criterion of 0.05. - the logistic regression produces the formula for each score (S) in the form: (formula 1 with 5 biomarkers) S = [β0 + βl log (Xl) + β2 X2 + β3 log2(X3) + β4 log2(X4)+ β5 log2(X5)] β6 or (formula 2 with 4 biomarkers, where β5=0) S = [β0 + βl log (Xl) + β2 X2 + β3 log2(X3) + β4 log2(X4)] β6 wherein the coefficients βi(β0, β1, β2, β3, β4, β5, β6)are constants coefficient, and the variables Xi are the independent variables; X1: Gamma GT level, X2: FBG (Free Blood Glucose or Glycemia) X3: Galectin3BP level; X4: CK18 level, X5: 7KC level. The FBG is the sole binary variable with a threshold at 7: if FBG higher than 7 mmol / L the FBG variable is 1, if FBG lower than 7 mmol / L, the FBG variable is null. This score corresponds to the logit of p where p is the probability of existence of the diagnostic target (Risk of Nash). This probability p (Risk of Nash or PreciNASH score) is calculated with the following formula: p = exp S / (1+exp S) The term constant coefficient "βi" means a "weighting factor" which gives the weight of the different biomarkers, in the classification condition. The term "Weighting factor" is a weight given to a variable to assign it a lighter, or heavier, importance in a function. The value of the associated coefficient βi (called [beta]) (β0, β1, β2, β3, β4, β5, β6) is given in the tables below and before the Example section), and the last two columns give the confidence interval, i.e. the confidence interval (called CI in the tables below) of the beta coefficients or corresponding odds-ratio (called Estimates (IC95%); see below). The coefficient β0 correspond to variable Intercept Table A: Example of Estimates (IC95%) for Formula 1 with 5 biomarkers. The coefficient β0correspond to variable Intercept. Variables P OR (IC 95%) Estimates (IC95%) Intercept <0.001 - -17.535 (-21.077 to -13.994) Gamma GT (log) 0.002 2.131 (1.329 to 3.417) 0.756 (0.284 to 1.229) Glycemia (> 7 mmol / L) <0.001 3.817 (2.086 to 6.983) 1.339 (0.735 to 1.943) MAC2B (log2) 0.025 1.701 (1.068 to 2.708) 0.531 (0.006 to 0.996) CK 18 (log2) <0.001 3.059 (2.235 to 4.186) 1.118 (0.804 to 1.432) 7 KC (log2) 0.003 1.763 (1.206 to 2.578) 0.567 (0.187 to 0.947) Table B: Example of Estimates (IC95%) for Formula 2 with 5 biomarkers. The coefficient β0 correspond to variable Intercept. Variables P OR (IC 95%) Estimates (IC95%) Intercept <0.001 - -15.722 (-18.884 to -12.561) Gamma GT (log) <0.001 2.402 (1.504 to 3.837) 0.876 (0.408 to 1.345) Glycemia (> 7 mmol / L) <0.001 4.054 (2.237 to 7.346) 1.4 (0.805 to 1.994) MAC2B (log2) 0.006 1.881 (1.198 to 2.952) 0.632 (0.181 to 1.083) CK 18 (log2) <0.001 2.947 (2.173 to 3.996) 1.081 (0.776 to 1.385) 7 KC (log2) 0.003 1.763 (1.206 to 2.578) 0.567 (0.187 to 0.947) Table C: combination of the models (depending on the number and type of biomarker used) modèletest HAUC & L corSeuil Spéci. VPP LR + Nb sub Sens. VPN LR - Nb inf % de BC2a gamma glycemie gal3 CK18 7KC 0,443 0,906 0,44 0,94 0,71 10,20 80 0,6 0,91 0,42 406 0,870,14 0,76 0,49 3,94 180 0,92632 0,98 0,10 306 0,802b gamma glycemie gal3 CK18 0,2985 0,902 0,49 0,95 0,71 10,19 73 0,55 0,90 0,48 413 0,870,08 0,65 0,40 2,71 224 0,94 0,98 0,10 262 0,712c gamma glycemie gal3 0,53 0,96 0,71 10,03 55 0,41 0,87 0,61 431 0,857KC0,832 0,850,08 0,53 0,33 2,02 273 0,95 0,98 0,10 213 0,612d gamma glycemie CK18 7KC 0,45 0,9 0,43 0,94 0,71 10,02 79 0,59 0,90 0,44 407 0,870,11 0,74 0,46 0,10 190 0,93 0,98 0,10 296 0,782e gamma gal3 CK18 7KC 0,633 0,892 0,48 0,95 0,71 10,08 69 0,52 0,89 0,51 417 0,860,09 0,64 0,39 2,58 231 0,94 0,98 0,10 255 0,702f glycemie gal3 CK18 7KC 0,346 0,899 0,50 0,95 0,71 10,29 70 0,53 0,89 0,50 416 0,870,13 0,76 0,49 3,89 181 0,93 0,98 0,10 305 0,792g gamma glycemie gal3 0,916 0,844 0,51 0,96 0,71 10,03 55 0,41 0,87 0,61 431 0,850,06 0,43 0,29 1,67 315 0,96 0,98 0,10 171 0,532h gamma glycemie CK18 0,455 0,895 0,41 0,94 0,71 10,10 76 0,57 0,90 0,46 410 0,870,08 0,67 0,41 2,84 218 0,94 0,98 0,09 268 0,722i gamma gal3 CK18 0,576 0,884 0,52 0,96 0,72 10,55 57 0,43 0,87 0,59 429 0,860,08 0,59 0,36 2,29 252 0,95 0,98 0,09 234 0,662j glycemie gal3 CK18 0,379 0,894 0,51 0,96 0,72 10,41 60 0,45 0,88 0,57 426 0,860,10 0,68 0,41 2,91 215 0,94 0,98 0,09 271 0,732k gamma glycemie 7KC 0,345 0,836 0,60 0,97 0,71 10,10 38 0,28 0,85 0,74 448 0,840,05 0,34 0,26 1,46 351 0,97 0,98 0,09 135 0,462l gamma gal3 7KC 0,561 0,817 0,83 1,00 0,80 16,45 5 0,04 0,81 0,96 481 0,800,08 0,43 0,29 1,70 315 0,97 0,98 0,07 171 0,532m glycemie gal3 7KC 0,797 0,818 0,60 0,98 0,71 10,29 28 0,21 0,84 0,81 458 0,830,05 0,25 0,24 1,30 387 0,98 0,98 0,08 99 0,392n gamma CK18 7KC 0,718 0,878 0,49 0,94 0,71 10,29 77 0,58 0,90 0,45 409 0,870,09 0,54 0,33 2,07 269 0,95 0,98 0,10 217 0,622o glycemie CK18 7KC 0,385 0,893 0,44 0,94 0,71 10,29 77 0,58 0,90 0,45 409 0,870,07 0,60 0,37 2,37 246 0,95 0,98 0,09 240 0,672p gal3 CK18 7KC 0,478 0,878 0,53 0,96 0,71 10,03 55 0,41 0,87 0,61 431 0,850,07 0,51 0,32 1,95 283 0,96 0,98 0,08 203 0,602q gamma glycemie 0,666 0,828 0,62 0,98 0,73 11,17 26 0,20 0,83 0,81 460 0,830,05 0,24 0,24 1,28 391 0,98 0,98 0,09 95 0,382r gamma gal3 0,519 0,805 0,72 0,99 0,71 10,28 7 0,05 0,81 0,95 479 0,810,04 0,13 0,22 1,13 435 0,99 0,98 0,08 51 0,302s gamma CK18 0,163 0,874 0,53 0,96 0,71 10,03 55 0,41 0,87 0,61 431 0,850,06 0,47 0,30 1,80 299 0,96 0,98 0,09 187 0,562t gamma 7KC 0,221 0,8 0,79 0,99 0,71 10,28 7 0,05 0,81 0,95 479 0,810,06 0,26 0,24 1,32 382 0,98 0,98 0,08 104 0,402u glycemie gal3 0,508 0,803 0,37 0,88 0,55 5,10 103 0,60 0,90 0,45 383 0,830,06 0,20 0,23 1,23 404 0,98 0,98 0,10 82 0,362v glycemie CK18 0,111 0,886 0,45 0,95 0,71 10,08 65 0,52 0,89 0,51 421 0,860,08 0,61 0,37 2,45 237 0,95 0,98 0,09 249 0,682w glycemie 7KC 0,711 0,792 0,61 0,99 0,71 10,29 14 0,11 0,82 0,90 472 0,820,07 0,28 0,24 1,32 371 0,95 0,96 0,19 115 0,412x gal3 CK18 0,053 0,866 0,62 0,97 0,71 10,13 45 0,34 0,86 0,69 441 0,840,07 0,42 0,29 1,66 317 0,96 0,98 0,10 169 0,532y gal3 7KC 0,164 0,748 0,36 0,92 0,45 3,43 55 0,26 0,84 0,80 431 0,790,08 0,22 0,23 1,25 398 0,98 0,98 0,10 88 0,372z CK18 7KC 0,396 0,868 0,43 0,95 0,71 10,08 69 0,52 0,89 0,51 417 0,860,06 0,43 0,29 1,68 314 0,96 0,98 0,10 172 0,532ac gal3 0,111 0,699 0,29 0,92 0,40 2,74 55 0,23 0,83 0,84 431 0,780,10 0,17 0,22 1,17 416 0,97 0,96 0,18 70 0,332ad CK18 0,086 0,852 0,50 0,96 0,72 10,55 57 0,43 0,87 0,59 429 0,860,08 0,47 0,31 1,82 297 0,96 0,98 0,09 189 0,572ae 7KC 0,73 0,684 0,32 0,91 0,36 2,30 53 0,20 0,82 0,88 433 0,770,12 0,32 0,24 1,32 350 0,89 0,93 0,33 136 0,43 Table D: Illustrations of b values depending on the different models detailed in table 3b GGT Fasting Glc MAC2-BP CK18 7-KC S.F. Models b0 b 1 b 2 b 3 b 4 b 5 b 6 2a -17,535 0,756 1,339 0,531 1,118 0,567 0,957 2b -15,722 0,876 1,4 0,632 1,081 0 0,968 2c -9,81 1,116 1,638 0,674 0 0,453 0,968 2d -16,822 0,787 1,381 0 1,143 0,636 0,969 2g -8,688 1,227 1,668 0,763 0 0 0,98 2h -14,537 0,908 1,461 0 1,109 0 0,98 2k -8,624 1,138 1,712 0 0 0,548 0,981 2q -6,985 1,267 1,764 0 0 0 0,993 For all models, the βi values included in the score (S) or (PreciNASH score) are in ranges detailed bellow: • -18 ≤ β0 ≤ -5 • 0 ≤ β1 ≤ 1,4 • 0 ≤ β2 ≤ 1,85 • 0 ≤ β3 ≤ 0,9 • 0 ≤ β4 ≤ 1,3 • 0 ≤ β5 ≤ 0,8 • 0,95 ≤ β6 ≤ 1 The coefficient β6 correspond to the shrinkage factor. In statistics, shrinkage is the reduction in the effects of sampling variation. The value of the associated coefficient β6 (called [beta]) may be determined by the man in the art. Example of shrinkage factor (mean) are 0.957 (5 variables) and 0.968 (4 variables) Most accurate models’ values for βiare given in the table C. Table C which gives the cutoff values and the LR+ and LR- value for each models using, 5 or 4 or less biomarkers. For other models, the ranges for βi are detailed after table D (given before the example section). The term “log” used herein refers to the natural logarithm function The term “log2” (or binary logarithm) used herein refers to the logarithm function with the basis of 2 (Log2) and is the inverse function of the power of two function. As well as log2, an alternative notation for the binary logarithm is lb (the notation preferred by ISO 31-11 and ISO 80000-2). The binary logarithm (log2 n) is the power to which the number 2 must be raised to obtain the value n. That is, for any real number x, 1 is 0, the binary logarithm of 2 is 1, the binary logarithm of 4 is 2, and the binary logarithm of 32 is 5. The term “diagnosis” means the identification of the condition or the assessment of the severity of the disease. In the context of the present invention the “diagnosis” is associated with the levels of 2 serum biomarkers Galectin3 binding protein (Galectin3- BP) also named MAC2-binding protein (MAC2-BP), and cytokeratin-18 (CK-18) (protein markers), and the level of clinical marker gamma-glutamyltranspeptidase (gamma GT) and the glycemia (Fasting Blood Glucose) higher than 7 mmol / L (clinical marker) and optionally 7-Ketocholesterol (7-KC) (lipid marker) and combining levels of said markers in a mathematical function through a logistic function including said markers in order to obtain an end value (Global NASH score) which in turn may be a risk for developing severe form of NAFLD. In some embodiments, the methods of the present invention are performed in vitro or ex vivo. The term “subject” as used herein refers to a mammalian, such as a rodent (e.g. a mouse or a rat), a feline, a canine or a primate. In a preferred embodiment, said subject is a human subject. The subject according to the invention can be a healthy subject or a subject suffering from a given disease such as NAFLD disease. As used herein, the term “NAFLD” (non-alcoholic fatty liver disease) has its general meaning in the art and refers to a spectrum of hepatic lipid disorders characterized by hepatic fat accumulation (steatosis) in people who drink little or no alcohol. NAFLD is also defined as a progressive liver disease that ranges from hepatic fat accumulation (simple steatosis) to non-alcoholic steatohepatitis (NASH) (Rinella et al, J Hepatol 2023). As used herein, the term “NASH” (non-alcoholic steatohepatitis) has its general meaning in the art and refers to a progressive disease of the liver characterized histologically by hepatic lipid accumulation, hepatocyte damage and inflammation resembling alcoholic hepatitis. NASH is a critical stage in the process that can lead to advanced fibrosis (also called “NASH associated fibrosis”), cirrhosis, liver failure and / or HCC (Hepatocellular Carninoma). A careful history of a lack of significant alcohol intake is essential to establish this diagnostic. NASH is one of the most common causes of elevated aminotransferases in patients referred for evaluation to hepatologists. NASH is generally associated with energy metabolism pathologies, including obesity, dyslipidemia, diabetes and metabolic syndrome. NASH is the hepatic expression of the metabolic syndrome. Extensive dysregulation of hepatic cholesterol homeostasis drives progressive hepatic inflammation and fibrosis and has been documented in NASH. This dysregulation occurs at multiple levels including decreased cholesterol excretion in bile, either as cholesterol or as bile acids (Ioannou G.N. Trends in Endocrinology & Metabolism, February 2016, Vol.27, No.2). NASH disease (Non Alcoholic SteatoHepatitis), was recently renamed MASH (Metabolic Dysfunction-associated Steatohepatitis) (Rinella et al, J Hepatol 2023). The present invention aimed to develop and validate a blood diagnosis test for NASH with serum biochemical markers associated with NAFLD's predictive value of the necro- inflammatory activity. In the context of present invention, the NAS score (NAFLD Activity Score), which allows to classifies disease severity for patients, and the Brunt score to grade the severity of necroinflammatory activity in NASH patient (see Brunt EM, et al Am J Gastroenterol. 1999 Sep; 94(9):2467-74). were used to validate the severity classification of the present invention to NAFLD disease severity in general. The N.A.S. score, defined as the unweighted sum of scores for steatosis (0-3), lobular inflammation (0-3), and ballooning (0- 2), ranging from 0 to 8, was also evaluated for the histological follow-up of NASH after bariatric surgery. The Brunt score is defined by a 3-step scale: mild (grade 1), moderate (grade 2) and severe (grade 3). As used herein, the term “blood sample” means a whole blood sample obtained from a subject (e.g. an individual for which it is interesting to determine whether a population of serum biomarkers can be identified). As used herein, the term "7-Ketocholesterol " or “7-KC” also known as 7- Oxocholesterol, 566-28-9, Cholesterol, 7-oxo- or 7-oxo-cholesterol (PubChem CID 91474) has its general meaning in the art refers to a cholestanoid that consists of cholesterol bearing an oxo substituent at position 7. 7KC is a toxic oxysterol which is the most common product of a reaction between cholesterol and oxygen radicals and is the most concentrated oxysterol found in the blood and arterial plaques of coronary artery disease patients as well as various other disease tissues and cell types (Anderson A et al Redox Biology 29 (2020) 101380). Unlike cholesterol, 7KC consistently shows cytotoxicity to cells and its physiological function in humans or other complex organisms is unknown. Oxysterols, particularly 7KC, have also been shown to diffuse through membranes where they affect receptor and enzymatic function. 7KC is associated with many diseases and disabilities of aging, as well as several orphan diseases (Anderson A et al Redox Biology 29 (2020) 101380). As used herein, the term " Galectin3 binding protein " (Galectin3 BP) also known as LGALS3BP or also named MAC2-binding protein (MAC2-BP), has its general meaning in the art and refers to protein that in humans is encoded by the LGALS3BP gene (gene ID 3959) The galectins are a family of beta-galactoside-binding proteins implicated in modulating cell–cell and cell–matrix interactions. Using fluorescence in–situ hybridization, the full length 90K cDNA has been localized to chromosome 17q25. The native protein binds specifically to a human macrophage-associated lectin known as Mac-2 (or galectine 3) and also binds to galectin 1 (Tinari N, et al. (2001). Int. J. Cancer.91 (2): 167–72). Galectin-3 BP has been found elevated in the serum of patients with cancer and in those infected by the human immunodeficiency virus (HIV). It appears to be implicated in immune response associated with natural killer (NK) and lymphokine-activated killer (LAK) cell cytotoxicity. As used herein, the term "cytokeratin-18" (CK-18) also known as KRT18, CYK18, K18, keratin 18, has its general meaning in the art and refers to a type I cytokeratin that constitute the Type I intermediate filaments (IFs) of the intracytoplasmatic cytoskeleton, which is present in all mammalian epithelial cells. In humans, the cytokeratin-18 protein is encoded by the KRT18 gene (gene ID 3875). CK18 together with its filament partner keratin 8, is perhaps the most commonly found products of the intermediate filament gene family. They are expressed in single layer epithelial tissues of the body. Mutations in this gene have been linked to cryptogenic cirrhosis. Two transcript variants encoding the same protein have been found for this gene (see KRT18 gene NCBI website). CK18 is often used together with keratin 8 and keratin 19 to differentiate cells of epithelial origin from hematopoietic cells in tests that enumerate circulating tumor cells in blood (Jeffrey Allard W. et al. (2004). Clinical Cancer Research. 10 (20): 6897–6904). Several studies shown that cytokeratin-18 (CK-18) serum fragments, a marker of apoptosis, could be a relevant pathogenic mechanism involved in NASH (Wieckowska A, et al. Hepatology. 2006;44:27–33). CK-18 fragments have been shown to be significantly higher in patients with NASH compared to those with simple steatosis, with an AUC (Area Under Curve) ranging from 0.78 (Yilmaz Y, et al World J Gastroenterol. 2007;13:837–44) to 0.93 (Wieckowska A, et al. Hepatology. 2006;44:27–33). A meta-analysis of nine studies including 852 patients confirmed these data, showing a pooled AUC of 0.82 for the prediction of NASH (Musso G, et al. Ann Med.2011;43:617–49). As used herein, the term "gamma-glutamyltranspeptidase" (gamma GT) also known as γ-glutamyltransferase, GGT, gamma-glutamyl transpeptidase; (ENZYME entry: EC 2.3.2.2), has its general meaning in the art and refers to a mammalian enzyme (a transferase) which is part of the cell antioxidant defense mechanism. In humans, mains proteins that belong to gamma GT family include GGT1 encoded by the GGT1 gene (Gene ID 2678), GGT2 encoded by the GGT2 gene (Gene ID 2679). Gamma GT catalyzes the transfer of gamma- glutamyl functional groups from molecules such as glutathione to an acceptor that may be an amino acid, a peptide or water (forming glutamate) (Whitfield JB (2001). Critical Reviews in Clinical Laboratory Sciences. 38 (4): 263–355). Gamma GT plays a key role in the gamma- glutamyl cycle, a pathway for the synthesis and degradation of glutathione as well as drug and xenobiotic detoxification (Courtay C et al. (1992). Biochemical Pharmacology. 43 (12): 2527–33). Other lines of evidence indicate that gamma GT can also exert a pro-oxidant role, with regulatory effects at various levels in cellular signal transduction and cellular pathophysiology (Dominici S, et al (2005). Methods in Enzymology. 401: 484–501). This transferase is found in many tissues, the most notable one being the liver, and has significance in medicine as a diagnostic marker for liver disease. Latent elevations in gamma GT are typically seen in patients with chronic viral hepatitis infections often taking 12 months or more to present. As use herein, the term “glycemia” ((Free Blood Glucose) has its general meaning in the art and refers to the presence, or the level, of glucose in one's blood subject and Hyperglycemia, means an unusually high concentration of glucose in the blood and Hypoglycemia, means an unusually low concentration of glucose in the blood. The level of the markers of the invention may be determined by using standard electrophoretic and immunodiagnostic techniques, including immunoassays such as competition, direct reaction such as immunohistochemistry, or sandwich type assays. Such assays include, but are not limited to, Western blots; agglutination tests; enzyme-labelled and mediated immunoassays, such as ELISAs; biotin / avidin type assays; radioimmunoassays; immuneelectrophoresis; immunoprecipitation, etc. The reactions generally include revealing labels such as fluorescent, chemiluminescent, radioactive, enzymatic labels or dye molecules, or other methods for detecting the formation of a complex between the antigen and the antibody or antibodies reacted therewith. Standard methods for detecting the level of specific biomarker such as 7-Ketocholesterol (lipid marker) or Galectin3 binding protein (Galectin3- BP) also named MAC2-binding protein (MAC2-BP), cytokeratin-18 (CK-18) (protein marker) are well known in the art. For example, the lipid marker level can be determined using mass spectrometric-based metabolomic approach (to determine the concentration of several analyte classes, including lipids, free fatty acids, and oxysterols) Typically, the step consisting of detecting the protein marker may consist in using at least one differential binding partner directed against the marker. As used herein, the term “binding partner directed against the marker” refers to any molecule (natural or not) that is able to bind the surface marker with high affinity. The binding partners may be antibodies that may be polyclonal or monoclonal, preferably monoclonal antibodies. In another embodiment, the binding partners may be a set of aptamers. Polyclonal antibodies of the invention or a fragment thereof can be raised according to known methods by administering the appropriate antigen or epitope to a host animal selected, e.g., from pigs, cows, horses, rabbits, goats, sheep, and mice, among others. Various adjuvants known in the art can be used to enhance antibody production. Although antibodies useful in practicing the invention can be polyclonal, monoclonal antibodies are preferred. Monoclonal antibodies of the invention or a fragment thereof can be prepared and isolated using any technique that provides for the production of antibody molecules by continuous cell lines in culture. Techniques for production and isolation include but are not limited to the hybridoma technique originally; the human B-cell hybridoma technique; and the EBV-hybridoma technique. The binding partners of the invention such as antibodies or aptamers may be labelled with a detectable molecule or substance, such as preferentially a fluorescent molecule, or a radioactive molecule or any others labels known in the art. Labels are known in the art that generally provide (either directly or indirectly) a signal. As used herein, the term "labelled", with regard to the antibody or aptamer, is intended to encompass direct labelling of the antibody or aptamer by coupling (i.e., physically linking) a detectable substance, such as a fluorophore [e.g. fluorescein isothiocyanate (FITC) or phycoerythrin (PE) or Indocyanine (Cy5)]) or radioactive molecule or a non-radioactive heavy metals isotopes to the antibody or aptamer, as well as indirect labelling of the probe or antibody by reactivity with a detectable substance. An antibody or aptamer of the invention may be labelled with a radioactive molecule by any method known in the art. More particularly, the antibodies are already conjugated to a fluorophore (e.g. FITC-conjugated and / or PE-conjugated). The aforementioned assays may involve the binding of the binding partners (ie. antibodies or aptamers) to a solid support. The solid surface could a microtitration plate coated with the binding partner for the surface marker. Alternatively, the solid surfaces may be beads, such as activated beads, magnetically responsive beads. Beads may be made of different materials, including but not limited to glass, plastic, polystyrene, and acrylic. In addition, the beads are preferably fluorescently labelled. Such methods comprise contacting a biological sample obtained from the subject to be tested under conditions allowing detection of 7-Ketocholesterol (7-KC) (lipid marker) or Galectin3 binding protein (Galectin3- BP) also named MAC2-binding protein (MAC2-BP) cytokeratin-18 (CK-18) (protein markers). Once the sample from the subject is prepared, the level of NASH biomarkers (“Biomarker protein”: Galectin3- BP and CK-18 and optionally “Biomarker Lipid”: 7-KC) may be measured by any known method in the art. Typically, regarding the reference value using “Biomarker protein” (Galectin3- BP and CK-18) also named MAC2-binding protein (MAC2-BP) and / or clinical marker (Gamma GT and Glycemia) and optionally «Biomarker lipid” (7-KC), as indicated in the Experimental section using spectrometric based metabolic approach and immunoassay approach to identify and quantify lipid marker (7-KC) and protein markers (Galectin3- BP and CK-18) also named MAC2-binding protein (MAC2-BP) and clinical marker (Gamma GT and Glycemia), and wherein the NASH score value is higher than 0.48 is predictive of a high risk of having or developing NASH (non-alcoholic steatohepatitis) disease and wherein the NASH score value is lower than 0.08 is predictive of a high risk of not having or not developing NASH (non-alcoholic steatohepatitis) disease and wherein the NASH score value is in between 0.08 and 0.48, the threshold of 0.16 can be used. A score higher than 0.16 is at risk of NASH and score lower than 0.16 a at low risk of NASH. A “reference value” can be a “threshold value” or a “cut-off value”. Typically, a "threshold value" or a "constant factor” are constant numbers that can be determined experimentally, empirically, or theoretically. A threshold value can also be arbitrarily selected based upon the existing experimental and / or clinical conditions, as would be recognized by a person of ordinary skilled in the art. The threshold value has to be determined in order to obtain the optimal sensitivity and specificity according to the function of the test and the benefit / risk balance (clinical consequences of false positive and false negative). Typically, the optimal sensitivity and specificity (and so the threshold value) can be determined using a Receiver Operating Characteristic (ROC) curve based on experimental data. Preferably, the person skilled in the art may compare the levels of “Biomarker Lipid” (7-KC) and / or “Biomarker protein” (Galectin3- BP also named MAC2-binding protein (MAC2-BP) and CK- 18) with a defined threshold value. In one embodiment of the present invention, the threshold value is derived from the lipid levels and / or protein levels (or ratio, or score) determined in a blood sample derived from one or more subjects who are responders (to the method according to the invention). In one embodiment of the present invention, the threshold value may also be derived from lipid level and / or protein levels (or ratio, or score) determined in a blood sample derived from one or more subjects or who are non-responders (ie asymptomic subject). Furthermore, retrospective measurement of the lipid levels and / or protein levels (or ratio, or scores) in properly banked historical subject samples may be used in establishing these threshold values. Reference values are easily determinable by the one skilled in the art, by using the same techniques as for determining the classification of patients’ severity by N.A.S from samples previously collected from the patient under testing. "Risk" in the context of the present invention, relates to the probability that an event will occur over a specific time period, as in the conversion to critical form of coronavirus disease, and can mean a subject's "absolute" risk or "relative" risk. Absolute risk can be measured with reference to either actual observation post-measurement for the relevant time cohort, or with reference to index values developed from statistically valid historical cohorts that have been followed for the relevant time period. Relative risk refers to the ratio of absolute risks of a subject compared either to the absolute risks of low-risk cohorts or an average population risk, which can vary by how clinical risk factors are assessed. Odds ratios, the proportion of positive events to negative events for a given test result, are also commonly used (odds are according to the formula p / (l-p) where p is the probability of event and (1- p) is the probability of no event) to no conversion. Alternative continuous measures, which may be assessed in the context of the present invention, include time to critical form of coronavirus disease conversion risk reduction ratios. "Risk evaluation," or "evaluation of risk" in the context of the present invention encompasses making a prediction of the probability, odds, or likelihood that an event or disease state may occur, the rate of occurrence of the event or conversion from one disease state to another, i.e., from a normal condition or NAFLD to a NASH condition or to one at risk of developing a NASH disease. Risk evaluation can also comprise prediction of future clinical parameters, traditional laboratory risk factor values, or other indices of critical form of NASH disease, such as cellular population determination in peripheral tissues, in serum or other fluid, either in absolute or relative terms in reference to a previously measured population. The methods of the present invention may be used to make continuous or categorical measurements of the risk of conversion to NASH disease, thus diagnosing and defining the risk spectrum of a category of subjects defined as being at risk for a NASH disease. In the categorical scenario, the invention can be used to discriminate between normal and other subject cohorts at higher risk for NASH disease. In other embodiments, the present invention may be used so as to help to discriminate those having NAFLD from NASH disease. Accordingly, the method of detection of the invention is consequently useful for the in vitro diagnosis of NASH from a blood sample. In particular, the method of detection of the invention is consequently useful for: - Exclusion of the risk of NASH in patient at risk - Identify patient a high risk of NASH and should be screened for liver biopsy or treatment. Monitoring methods and Management of the invention After the identification of lipid and protein subsets that harbour an necro inflammatory phenotype, inventors highlighted, that both “Biomarker lipid” (7-KC) and / or “Biomarker protein” (Galectin3- BP also named MAC2-binding protein (MAC2-BP) and CK-18) strongly correlated with NAS severity scores, commonly used in clinical practice for NASH disease severity for patients. Accordingly, inventors provided evidence that these biomarkers subset may serve as a severity biomarker in NASH for prognosis and monitoring purpose (pathology or treatment). Indeed, the simplified acute physiology score (NAS), was used to validated the severity classification of the present invention, Furthermore, the inventors showed that depending on the patient an necro inflammatory profile, different biomarkers combinations needs to be used for monitoring disease progression. • Monitoring the disease Accordingly the invention also describe an in vitro method for monitoring NASH (non-alcoholic steatohepatitis) disease comprising the steps of i) determining the level of the 2 serum markers Galectin3 binding protein (Galectin3- BP), cytokeratin-18 (CK-18) (protein markers) in a blood sample obtained from a subject at a first specific time of the disease, ii) combining said levels by implementing a statistical technique on said levels, said statistical technique involving a logistic regression in order to obtain an end value (global NASH Score), iii) determining the level of the same serum markers determined in step i) in a blood sample obtained from the subject at a second specific time of the disease, iv) combining said levels by implementing a statistical technique on said levels, said statistical technique involving a logistic regression in order to obtain an end value, v) comparing the end value determined at step ii) with the end value determined at step iv) and vi) concluding that the disease has evolved in worse manner when the end value determined at step ii) is higher than the end value determined at step i). In a particular embodiment wherein in an additional step determining in a blood sample obtained from the subject, the blood level of the gamma-glutamyltranspeptidase (gamma GT) and / or the glycemia is higher than 7 mmol / L and combining the blood levels of biological and clinical markers previously measured by implementing a statistical technique on said levels, said statistical technique involving a logistic regression in order to obtain an end value (global NASH Score). In a particular embodiment wherein in an additional step determining in a blood sample obtained from the subject, the level 7-Ketocholesterol (7-KC) (lipid marker) and combining the blood levels of biological and optionally clinical markers previously measured by implementing a statistical technique on said levels, said statistical technique involving a logistic regression in order to obtain an end value (global NASH Score). In a particular embodiment, said method is performed using the level of 2, 3, 4 or 5 different biomarkers including with the level of Galectin3 binding protein (Galectin3- BP) also named MAC2-binding protein (MAC2-BP), cytokeratin-18 (CK-18) and optionally the level of gamma-glutamyltranspeptidase (gamma GT) and / or the level of glycemia ((Free Blood Glucose) and / or the level of 7-Ketocholesterol (7-KC). In a particular embodiment, said method is performed using the level of 4 or 5 different biomarkers including the level of Galectin3 binding protein (Galectin3- BP) also named MAC2-binding protein (MAC2-BP), cytokeratin-18 (CK-18) gamma- glutamyltranspeptidase (gamma GT) and glycemia (Fasting Blood Glucose) and optionally the level of 7-Ketocholesterol (7-KC). In a preferred embodiment said method is performed using the level the level of Galectin3 binding protein (Galectin3- BP) also named MAC2-binding protein (MAC2-BP), cytokeratin-18 (CK-18) gamma-glutamyltranspeptidase (gamma GT) and glycemia (Fasting Blood Glucose) • Monitoring the treatment of the disease Accordingly the invention also describe an in vitro method for monitoring the treatment of a NASH (non-alcoholic steatohepatitis) disease comprising the steps of i) determining the level of the 2 serum biological marker Galectin3 binding protein (Galectin3- BP) also named MAC2-binding protein (MAC2-BP) and cytokeratin-18 (CK-18) (protein markers) in a blood sample obtained from a subject before the treatment, ii) combining said levels by implementing a statistical technique on said levels, said statistical technique involving a logistic regression in order to obtain an end value (global NASH Score) iii) determining the level of the same biological serum markers determined in step i) in a blood sample obtained from the subject after the treatment, iv) combining said levels by implementing a statistical technique on said levels, said statistical technique involving a logistic regression in order to obtain an end value v) comparing the end value determined at step iii) with the levels determined at step iv) and vi) concluding that the treatment is efficient when the end value determined at step iii) are lower than the end value determined at step iv). In a particular embodiment wherein in an additional step, determining in a blood sample obtained from the subject, the blood level of clinical marker gamma- glutamyltranspeptidase (gamma GT) and / or glycemia (Fasting Blood Glucose) higher than 7 mmol / L, and combining the blood levels of biological and clinical markers previously measured by implementing a statistical technique on said levels, said statistical technique involving a logistic regression in order to obtain an end value (global NASH Score). In a more particular embodiment wherein in an additional step determining in a blood sample obtained from the subject, the blood level of the gamma-glutamyltranspeptidase (gamma GT) and / or the glycemia is higher or lower than 7 mmol / L and combining the blood levels of biological and clinical markers previously measured by implementing a statistical technique on said levels, said statistical technique involving a logistic regression in order to obtain an end value (global NASH Score). In a particular embodiment wherein in an additional step determining in a blood sample obtained from the subject, the level 7-Ketocholesterol (7-KC) (lipid marker) and combining the blood levels of biological and optionally clinical markers previously measured by implementing a statistical technique on said levels, said statistical technique involving a logistic regression in order to obtain an end value (global NASH Score). In a particular embodiment, said method is performed using the level of 2, 3, 4 or 5 different biomarkers including with the level of Galectin3 binding protein (Galectin3- BP) also named MAC2-binding protein (MAC2-BP), cytokeratin-18 (CK-18) and optionally the level of gamma-glutamyltranspeptidase (gamma GT) and / or the level of glycemia (Free Blood Glucose) and / or the level 7-Ketocholesterol (7-KC). In a particular embodiment, said method is performed using the level of 4 or 5 different biomarkers including the level of Galectin3 binding protein (Galectin3- BP) also named MAC2-binding protein (MAC2-BP), cytokeratin-18 (CK-18) gamma- glutamyltranspeptidase (gamma GT) and glycemia (Fasting Blood Glucose) and optionally the level of 7-Ketocholesterol (7-KC). In a preferred embodiment said method is performed using the level the level of Galectin3 binding protein (Galectin3- BP) also named MAC2-binding protein (MAC2-BP), cytokeratin-18 (CK-18) gamma-glutamyltranspeptidase (gamma GT) and glycemia (Fasting Blood Glucose). The invention will be further illustrated by the following figures and examples. However, these examples and figures should not be interpreted in any way as limiting the scope of the present invention. FIGURES: Figure 1: (A) Flow Chart of the ABOS / PreciNASH cohort (B) Methodological steps for score creation and validation Figure 2: A) Distribution if the PreciNASH risk in the derivation cohort. Boxplots of the preciNASH risk in the derivation cohort (in the imputed dataset corresponding to the median AUC) for NASH and NAFL patients separately. Dashed horizontal lines represent the three thresholds. Very few NASH patients present a risk below the 0.08 threshold whereas very few NAFL patients present a risk over the 0.48 threshold B): Thresholds interpretations for the use of the score (model 2b). Figure 3: Performances of the PreciNASH test in the ABOS validation cohort A) Receiver Operating (AUROC) characteristics of the validation cohort B) Calibration plot of the preciNASH test C) Distribution of the preciNASH test according to the Nafld Activity Score. Spearman’s correlation “r” = 0.61 (CI 95%; 0.52-0.68) (p = 3.6 x 10-25). Figure 4: Sensitivity to change of the PreciNASH test. A: Boxplots of the preciNASH risk at the time of surgery and one year after surgery on paired 89 patients. Boxplots of the PreciNASH risk at the surgery time and one year after surgery on the 89 patients. The risk of NASH decreases significantly in this period of time (p<0.001) 4B: Boxplots of the preciNASH risk at the surgery time and one year after for the 45 NASH patients before the surgery. On the panel (Part 1), boxplots of the PreciNASH risk at the surgery time and one year after surgery on the 45 patients. The risk of NASH decreases significantly in this period of time (p<0.001). On the panel (Part 2), boxplots of the PreciNASH risk at the surgery time and one year after surgery, first on the 6 patients who are still NASH after surgery and second on the 39 patients who are no longer NASH after surgery. Figure 5: Performances of the PreciNASH test in the ABOS validation cohort (5 variables) A) Receiver Operating (AUROC) characteristics of the validation cohort B) Distribution of the preciNASH test according to the Nafld Activity Score. EXAMPLES The thresholds values for diagnosis varies with the numbers of values included in the formula. When a value is not considered for the diagnosis, then “βi” has a value equalled to 0. The thresholds and the values of bi are detailed in the tables C and D. An example with 4 variables and with a 7-KC equal to 0 is given and detailed bellow. Example 1Case with 4 variables. Methods Study design and population The study was performed using the ABOS cohort (NCT01129297). The ongoing ABOS cohort started in June 2006 and included morbid obese patients with or without diabetes candidates for bariatric surgery. All patients from the ABOS cohort were included in the Centre Hospitalier de Lille (CHU de Lille) which is a tertiary care center. Patients were adults (18 years or older) and were referred for bariatric surgery from primary and secondary care centers and, at the time of evaluation, fulfilled the criteria for bariatric surgery according to French national guidelines (see Appendix A1). Following the Declaration of Helsinki, all patients were informed and consented to participate. The methodology of the present study was in accordance with the TRIPOD guidelines (Moons, et al., 2015). The study population was divided into a derivation cohort (including ABOS patients between 1st June 2006 and 31 December 2016) and a validation cohort (including ABOS patients between 1st January 2017 until 30 November 2020). The derivation cohort was initially composed of 1234 patients. After exclusion of 390 liver biopsies to fulfill the quality criteria recommendation, 844 NAFLD patients were left. To better select the biomarker candidates, it has been decided to exclude from this first step all patients with a borderline NASH. In the same manner, all patients without steatosis (No NAFL) at baseline were also excluded, has the test is designed to decipher NASH from NAFL in patients with steatosis. At the end, the derivation cohort had 496 patients, with 399 NAFL and 97 NASH (20% of NASH). The validation cohort included 318 patients. After exclusion of 83 liver biopsies to fulfill the quality criteria recommendation, 235 NAFLD patients were left. The flow chart is detailed in Figure 1. Procedures All patients had several samples: liver biopsy at the time of surgery, and 1 year later; blood samples for current biological tests (liver, metabolic, lipids) and for research (genetics, omics) as well as muscle, subcutaneous fat, and visceral fat in the great omentum samples during surgery. Blood samples were performed during the same hospitalization as tissue sampling. Liver biopsies were systematically planned during the surgical procedure. To avoid artifacts related to prolonged exposure of the liver, the surgeon performed a liver needle biopsy during the first part of the surgical procedure after trocar insertion and abdominal exploration, within 10 minutes after pneumo-peritoneum installation. The Hepafix needle biopsy system was used until 2010, and the MONOPTY needle biopsy system (16G, ref: 121620; C. R. Bard, Tempe AZ, U.S.A.) after that. Biopsies were routinely stained with H&E saffron, red Sirius, and Perl's staining. Steatosis was quantified by the percentage of hepatocytes containing fat droplets (amount of steatosis). The pathologists (E.L., V.G., B.B. and D.B.) followed the NASH CRN recommendation. They first diagnosed NASH, then determined the NAFLD Activity Score (NAS). The NAS is defined as the unweighted sum of scores for steatosis (0-3), lobular inflammation (0-3), and ballooning (0-2), ranging from 0 to 8. If NASH was diagnosed, the severity of necroinflammatory activity was graded using the Brunt score, which is defined by a 3-step scale: mild (grade 1), moderate (grade 2) and severe (grade 3). The NAS was also evaluated for the histological follow-up of NASH after bariatric surgery. All biopsies were graded prospectively with these two scores, and results were standardized in a report. Pathologists were blinded to clinical and biological data. Liver fibrosis was assessed semi-quantitatively using the Kleiner fibrosis score (defined as follows: F0, normal; F1 stage is divided into three subclasses: 1a, mild pericellular fibrosis in zone 3, 1b, moderate pericellular fibrosis in zone 3, and 1c, portal fibrosis; F2, perivenular, and pericellular fibrosis confined to zones 2 and 3, with or without portal or periportal fibrosis; F3, bridging or extensive fibrosis with architectural distortion and no clear-cut cirrhosis; and F4, cirrhosis). Outcomes and predictor variables The primary objective of the study was to develop a NASH diagnostic score using clinical data and miRNAs, genotypes, proteins, metabolites extracted from blood or serum among patients at risk of NAFLD. The primary outcome was the status of the patient regarding NASH diagnosis detected as described above. The predictor variables considered for the building of the score were collected at the time of the surgery: past medical history, demographic and clinical characteristics (age, body mass index, gender), routine biological, hematological and metabolic blood samples. Besides all these variables, complementary biomarkers were extracted from the serum data bank: miRNAs, genotypes, proteins, metabolites analysis (complete details on the data are giving in Appendix A2). Statistical analysis Statistical testing was conducted at the two-tailed α-level of 0.05. Quantitative variables were expressed as mean (standard deviation) in normal distribution or median (interquartile range, IQR) otherwise. Normality of distributions was assessed using histograms and the Shapiro-Wilk test. Categorical variables were expressed as count (percentage). Statistical analyses were performed using the S.A.S. software package, release 9.4 (S.A.S. Institute) and using the R Software (Version 3.6). Screening of biomarkers candidates The number of NASH patients in the ABOS cohort was 97. The number of variables included in the model of constructing the score was fixed at approximately 10 to respect the Concato rule (Peduzzi, Concato, Kemper, Holford, & Feinstein, 1996). This number of variables was compatible with parsimony principle, which facilitates routine clinical use of the score. A comprehensive screening of biomarkers candidates (clinical and standard biological variables, proteins, metabolites, miRNAs, and genotypes) including a preprocessing step for omics data (see Appendix A3), and using univariate (with correction for multiple testing) as well as multivariate statistical analyses (with penalized logistic regression), was carried out in two matched groups of ABOS patients (NAFL and NASH patients) in order to identify a subset of variables which could be used to build the predictive score of NASH. To reduce investigator bias, we randomly selected the patients with NASH in the cohort and, each NASH patient was matched to a NAFL patient based on age (+ / - 5 years), gender, and body mass index (+ / -5 Kg / m2), using the overall optimal algorithm (Rosenbaum, 1989) . Derivation in the ABOS cohort The statistical analyses for the derivation of the predictive score were performed on the preselected variables during the screening phase. For each continuous predictor, the log- linearity assumption was assessed using the restricted cubic spline functions, and the absence of collinearity between variables was checked by calculating the variance inflation factors. When the log-linearity was rejected, variables were log-transformed or split into classes (with thresholds determined according to clinical opinion). We developed a multivariable predictive model by considering all candidate predictors identified during the screening phase irrespective of their univariate associations with the diagnosis status (NASH / NAFL). Missing data on predictors were handled by multiple imputations. We derived m = 10 imputed datasets from the original sample using regression switching approach by chained equations under the missing at random assumption using the predictors selected by the screening procedure (including the diagnosis status and the matching variables) (Van Buuren & Groothuis-Oudshoorn, 2011). The selection of predictors was performed using multivariable logistic regression performed on each imputed dataset combined with backward selection procedure with a removal criterion of 0.05 using the Rubin rules to pool the regression coefficients estimates at each step of the selection (see appendix A5 for details). The predictors selected by the backward procedure were introduced in multivariable logistic regression performed in each imputed dataset. In each imputed dataset, we examined the performance of the selected model in terms of calibration using the Hosmer-Lemeshow (HL) goodness of fit-test and a visual inspection of the calibration plot, and in terms of discrimination by calculating the AUC (Area Under the Roc Curve). Finally, Rubin’s rules (Marshall, Altman, Holder, & al., 2009) were used to combine the estimates of these 10 multivariable logistic regressions and build the final model. Median of AUCs and median of p-values of Hosmer-Lemeshow test were considered as a summary of performances of the final model in term of discrimination and calibration. We considered the ROC curve corresponding to the median of the AUCs in the 10 imputed datasets to identify three thresholds: the first threshold was based on the Youden index for optimizing the discrimination. The other two thresholds were determined using the likelihood ratios (positive LR+ and negative LR-) (Deeks & Altman, 2004): one threshold allowed to identify with a good accuracy the NASH patients (threshold from which the LR+ exceeds 10 with a minimum LR-) and the other threshold allowed to exclude with good accuracy the patients without NASH (among all the LR- less than to 0.1 we take the threshold which maximizes the LR+). Internal validation was done by using bootstrap resampling (200 repetitions) combined with handling of missing data by multiple imputations to estimate the AUC corrected for the over-optimism and the shrinkage factor. The shrinkage factor is the estimated shrinking in the regression coefficients to improve the prediction in future patients (Dunkler, Sauerbrei, & Heinze, 2016). Sensitivity to change Among the patients from the derivation cohort, 89 patients (including 45 NASH at baseline) had an evaluation after bariatric surgery (including all the clinical and omics data used for the score computation). Bariatric surgery could modify the NAFLD and NASH status as reported in Mathurin 2009 et al and Lassailly et al 2020. We then took advantage using these data to assess the sensitivity to change which is an important property of a diagnostic score. We computed the preciNASH score and the associated risk of NASH before surgery and at the visit post-surgery (which occurred at least 12 months after surgery). A description of change in the risks (before and after surgery) was performed using box plots. Comparisons of the risks before and after surgery were performed using the Wilcoxon signed rank test. Cohen's d (standardized differences) was computed as effect size with the 95% confidence interval. The association between the changes in the score and in the NAS score (which is the existing NAFLD activity score) was assessed using the Spearman’s correlation coefficient. Validation of the score The validation cohort included ABOS patients recruited between 1st January 2017 until 30 November 2020. The performance of the score, corrected by shrinkage factor, in terms of discrimination and calibration was assessed by calculating the AUC with the 95% confidence interval, and Hosmer and Lemeshow test with a calibration plot. The validity of the three thresholds identified in the derivation cohort was assessed by computing sensitivity, specificity and the likelihood ratios (LR+ and LR-). Results Population Selection of patients for the derivation cohort is shown in Figure 1. This cohort included 496 patients: 97 NASH patients, 399 NAFL patients. Patients were women in 70% of cases, 45% had diabetes mellitus, median age was 44 [34 - 52] years and BMI was 46.0 kg / m² [41.8 - 51.6]. Patients with NASH were older 47 [39 - 54] vs.43 [33 - 51] than NAFL patients. BMI was similar between NAFL (46 [41.8 - 51.7]) and NASH patients (45.9 [41.5 - 50.7]). As expected, NASH patients were more frequently diabetic (80.4% vs. 36.2%), hypertensive (85.6% vs.55.6%), and had worse liver blood tests as shown with ALT 46 IU / L [34 - 62] vs 27 IU / L [20 - 36] or with γGT 59 IU / L [42 - 95] vs 31 IU / L [22 - 48] than NAFL patients. Biological metabolic tests were more pathologic in NASH patients than in NAFL patients with fasting glucose 83 mg / dL [60 – 113] vs 56 mg / dL [51 – 66] and A1c glycated hemoglobin 7.7% [6.1 - 9.5] vs. 5.8% [5.5 - 6.4]. In terms of histological features, 37% of patients had a fibrosis higher than F2 in the NASH group vs 2.6% in the NAFL group, median of liver fat amount (steatosis evaluated on liver biopsy) was 60% [35 - 70] vs 20% [10 - 35]. Selection of patients for the validation cohort is shown in Figure 1. This cohort included 235 patients: 37 NASH patients, 198 no NASH patients. The characteristics of the cohort were as follow: median age 43 years [33-54], BMI 43.9 [40.6-48.0], diabetes 34.9%, NASH 15.7%. The characteristics of patients with NASH were more severe with 28.6% of patients with a fibrosis ≥ F3 compared to 2.3% in others. In terms of biological values were significantly higher in NASH patients as shown with: ALT 44.5 IU / L [29 – 60] vs 25 IU / L [18 – 36], GGT 55 IU / L [35 - 103] vs 28 IU / L [20 – 43], fasting glucose 67 mg / dL [58 – 87] vs. 54 mg / dL [50 – 68]. All characteristics of the derivation and validation cohorts are summarized in Table 1. Screening biomarkers For the screening of biomarkers, 40 NASH patients from the derivation cohort were matched to 40 NAFL patients from the derivation cohort based on age (+ / - 5 years), gender, and body mass index (+ / -5 Kg / m2) (see Appendix A4). The characteristics of these patients are shown in Appendix A7. At first, a pre-processing step was performed on omics data (normalization, limit of detection, quantification, see Appendix A3). After the preprocessing of data (764 biomarkers), there were 126 miRNAs, 153 metabolites, 165 proteins, 4 genotypes. Besides omics data, 43 clinical variables (demographics, anthropometric and laboratory results) were studied: 41 variables are quantitative while 2 are qualitative. Very few values were missing (2 values for 1 qualitative variable); hence no data imputation has been done. The screening of candidate predictors allowed to select 4 subsets of candidates: Subset 1.1 (selection by family without miRNAs : ALT, gamma G.T, fasting glucose, MAC2BP, CK18 protein, CDCP1 protein, 7KC metabolite, NC170Oh metabolite, MBOAT7 genotype) ; subset 1.2 (selection by family with miRNAs : same variables as in 1.1 but with miR.34a.5p) ; subset 2.1 (all families combined without miRNAs : CK18 protein, galectin3 protein, CDCP1 protein, 7KC metabolite, betaine metabolite, NC170Oh metabolite, MBOAT7 genotype); subset 2.2 (all families combined with miRNAs : same variables as 2.1). Consequently, we obtained three different subsets of candidate predictors. Since our goal is to develop a score for use in clinical practice, it should contain clinical variables. Since there are no clinical variable in subset 2.1, no score have been built on this subset. . Score building After the screening step, 6 omics data were candidates. For building the score, one metabolite (NC170Oh, because less than 10% of the values were above the detection threshold) and two proteins (CDCP1 and 7KC because of technological problems related to detection tools) were removed. Characteristics of omics data selected as predictors as shown in Table 2. All quantitative variables meet the log linear assumption after applying a log transformation except for the fasting glucose, which were split into two classes (≤7 mmol / L vs > 7 mmol / L) according to clinical expertise. Description of the full models identified by multivariable logistic regressions performed on the 2 subsets of selected predictors are given in Appendix A9. In these two models, ALT, MBOAT7 and Mir.34a.5p are not significant. Multivariate logistic regressions with backward selection combined with multiple imputations, performed on these 2 subsets of variables separately, selected the same 4 following variables: γGT, fasting glucose (dichotomous condition ≥ 7 mmol / L or 126 mg / dL), MAC2BP and CK18. After running logistic regression using these variables in each imputed dataset and combining the corresponding coefficients using the Rubin’s rules, we obtained the final predictive preciNASH score as well as the preciNASH risk (risk of NASH). Adjusted odds ratio (95% IC) and p-values of the final model are given in Table 3. The preciNASH score was considered discriminative with an AUC of 0.906 (95% CI, 0.906 -0.907) (median of the AUC computed on the 10 imputed datasets). The 10 ROC curves are shown in Appendix A11. Calibration plots are shown in Appendix A12. After bootstrap with 200 replacements, the AUC corrected for over-optimism was 0.90 and the shrinkage factor was 0.97 leading to a corrected score of 0.97*S. The preciNASH score and its associated risk equation are defined by: S = [β0 + 0.88 ^log(Gamma GT) + 1.4 ^ Fasting Glucose + 1.08 log2(CK18) + 0.63 log2(MAC2BP)] ^0.97 In the equation of the score S, β0is the intercept of the model and 0.97 is the shrinkage factor. The distribution of the risk of NASH according to the histological status of the patients is shown in Figure 2. The risk of NASH ranges from 0.0008 to 0.996 with a mean of 0.53 in NASH patients and 0.11 in NAFL patients. The ROC curve associated with the median of AUC over the 10 imputed datasets is presented in Figure 3. The optimal cut-off, determined by maximizing the Youden index, was 0.156 leading to a LR+ of 4.32 and a LR- of 0.15. The sensitivity was 0.88 and the specificity was 0.80. Two other thresholds were determined: one, equals to 0.48, to identify with a good accuracy the NASH patients was associated to LR+ of 10.19, leading to a specificity of 0.95 and 87% of well-classified patients; The other threshold, equals to 0.08, to exclude with good accuracy the patients without NASH was associated with an LR-of 0.1, leading to a sensitivity of 0.94. Validation cohort In the validation cohort, the AUC of the preciNASH score was 0.84 (95% CI; 0.77 to 0.90) (Figure 4). The optimal cut-off (0.156), identified on the derivation cohort, leads to a LR+ of 2.59, a LR- of 0.31, a sensitivity of 0.78 and a specificity of 0.70. For the 0.08 threshold, LR+ and LR- were respectively 1.65 and 0.18, sensitivity and specificity were respectively 0.41 and 0.94 whereas for the 0.48 threshold LR+ and LR- were 6.69 and 0.63 respectively, sensitivity and specificity were 0.92 and 0.44 respectively. These diagnostic performances are presented in Table 4. The distribution of the risk of NASH according to NAS classes is presented in Figure 4. PreciNASH score is correlated to the NAS [r = 0.61 (CI 95%; 0.52-0.68) (p < 0.001)] (Figure 4). Sensitivity to change of the PreciNASH score Sensitivity to change was analyzed among 89 patients having an evaluation before and after bariatric surgery. The 44 no NASH patients before surgery did not change status after surgery. Among the 45 NASH patients before surgery, only 6 remained NASH after surgery. The evolution of the risk of NASH among the 89 is shown in Figure 4a and among the 45 NASH patients is shown in Appendix A12. The preciNASH risk decreased significantly at one year after surgery (0.40 [0.13; 0.70] vs 0.04 [0.02; 0.10], sign rank -test, p < 0.001). The corresponding effect size is equal to 1.55 (95% CI [1.26; 1.86]). The Spearman’s correlation coefficient between the change in preciNASH risk and the change in NAS was 0.49 (0.31; 0.63), p < 0.001 on the 89 patients. For the 45 NASH patients before surgery, the preciNASH risk decreased also significantly at one year after surgery (0.65 [0.45; 0.85] vs 0.05 [0.02; 0.13], sign rank -test, p < 0.001). On these 45 patients, the corresponding effect size is equal to 2.31 (95% CI [1.85; 2.88]).

[0002] Table 1: Characteristics of the 496 patients from the derivation cohort and of the 235 patients from the validation cohort Derivation cohort Validation cohort Overall NAFL NASH Overall No NASH NASH C Values are expressed as the number / total number (%), mean ± standard-deviation, or median [interquartile range]. Abbreviations: BMI, body mass index ; ALT, alanine aminotransferase ; AST, aspartate aminotransferase ; Gamma GT, γ-glutamyltransferase ; HbA1c, glycated hemoglobin ; HDL, high-density lipoprotein ; LDL, low-density lipoprotein Missing values:^9 missing values in the derivation cohort;*8 missing values in the validation cohort;**18 and 43 missing values in the derivation cohort and in the validation cohort respectively;#117 and 47 missing values in the derivation cohort and in the validation cohort respectively;†2 and 9 missing values in the derivation cohort and in the validation cohort respectively;††8 and 9 missing values in the derivation cohort and in the validation cohort respectively;˅2 and 8 missing values in the derivation cohort and in the validation cohort respectively; 3 and 8 missing values in the derivation cohort and in the validation cohort respectively;^^1 and 5 missing value in the derivation cohort and in the validation cohort respectively;##4 missing values in the validation cohort;˅˅1 and 2 missing values in the derivation cohort and in the validation cohort respectively;~11 missing values in the validation cohort;§1 missing value in the validation cohort;&14 and 30 missing values in the derivation cohort and in the validation cohort respectively;Δ6 and 2 missing values in the derivation cohort and in the validation cohort respectively;π4 and 1 missing values in the derivation cohort and in the validation cohort respectively;^1 missing value in the derivation cohort

[0003] Table 2: Description of OMICS candidate predictors for the derivation step Overall NAFL patients NASH patients (N = 496) (N = 399) (N = 97) CK18, IU / L 191.7 [142.9 ; 299.2] 173.6 [136.3 ; 234.6] 413.3 [261.1 ; 878.7] MAC2BP**, µg / mL 5.0 [3.8 ; 6.7] 4.7 [3.6 ; 6.5] 6.3 [5.0 ; 8.1] Genotype rs641738 CC 145 / 440 (33) 125 / 353 (35.4) 20 / 87 (23.0) CT / TT 295 / 440 (67) 228 / 353 (64.6) 67 / 87 (77.0) miR 34a 5p*, Relative Units -4.3 ± 1.7 -4.7 ± 1.6 -2.9 ± 1.6 * 2 missing values ** 1 missing value Table 3: Final model at the derivation step Variables OR (95% CI) p-values Intercept - < 0.001 Gamma GT, IU / L (log) 2.40 (1.50 to 3.84) < 0.001 Fasting glucose >7mmol / L 4.05 (2.24 to 7.35) < 0.001 MAC2BP, mg / mL (log2) 1.88 (1.20 to 2.95) 0.006 CK18, U / L (log2) 2.95 (2.17 to 4.00) < 0.001 Results are expressed as multivariate odds-ratio (95% confidence interval) and adjusted p-values Table 4: Diagnostic performance of the preciNASH score in the validation cohort according to thresholds Threshold for LR Youden’s Threshold for LR + > 10 threshold - < 0.10 Threshold 0.48 0.16 0.08 Sensitivity 0.41 0.78 0.92 Specificity 0.94 0.70 0.44 Positive 0.56 0.33 0.24 predictive value Negative 0.89 0.95 0.97 predictive value LR + 6.69 2.59 1.65 LR - 0.63 0.31 0.18 Abbreviations: LR, likelihood ratio Additional Information List of the different variables screened and analysed in the study. Clinical Gender Height, age, weight, Body Mass Index, fasting glucose, fasting insulinemia, A1c glycated heamogloblin, total cholesterol, LDL cholesterol, HDL cholesterol, Total bilirubin, uric acid, uremia, LDH, blood creatin, AST, ALT, GGT, alpha 2 macroglobulin, haptoglobin, Apo-A1-lipoprotein, platelets, hematocrit, leucocyte count, monocytes count, prothromin time, transferrin, ferrtin, Albumin, pre-albumin,vit D, CDD, Protein Luminex Technology: Human CKCL-10; Human FAS-Ligand; FAS; Galectin-3- Binding Protein or MAC2BP; VAP-1; Serpin; Fetuin A; Human C3; Human FGF-21. Olink Technology: IL8; VEGFA; MCP-3; CDCP1; CD244; IL7; OPG; LAP TGF- beta-1; uPA; IL6; MCP-1; CXCL11; AXIN1; TRAIL; CXCL9; CST5; OSM; CXCL1; CCL4; CD6; SCF; IL18; SLAMF1; TGF-alpha; MCP-4; CCL11; TNFSF14; FGF-23; MMP-1; LIF- R; FGF-21; CCL19; IL-10RB; IL-18R1; PD-L1; Beta-NGF; CXCL5; TRANCE; HGF; IL- 12B; MMP-10; IL10; CCL23; CD5; CCL3; Flt3L; CXCL6; CXCL10; 4E-BP1; SIRT2; DNER; EN-RAGE; CD40; FGF-19; MCP-2; CASP-8; CCL25; CX3CL1; TNFRSF9; NT-3; TWEAK; CCL20; ST1A1; STAMPB; ADA; TNFB; CSF-1;TNFRSF14; LDL receptor; ITGB2; IL-17RA; TNF-R2; MMP-9; EPHB4; IL2-RA; OPG; ALCAM; TFF3; SELP; CSTB; MCP-1; CD163; Gal-3; GRN; NT-proBNP; BLM hydrolase; PLC; LTBR; Notch 3; TIMP4; CNTN1; CDH5; TLT-2; FABP4; TFPI; PAI; CCL24; TR; TNFRSF10C; GDF-15; SELE; AZU1; DLK-1; MPO; CXCL16; IL-6RA; RETN; IGFBP-1; CHIT1; TR-AP; GP6; PSP-D; PI3; Ep-CAM; AP-N; AXL; IL-1RT1; MMP-2; FAS; MB; TNFSF13B; PRTN3; PCSK9; U- PAR; OPN; CTSD; PGLYRP1; CPA1; JAM-A; Gal-4; IL-1RT2; SHPS-1; CCL15; CASP-3; uPA; CPB1; CHI3L1; ST2; t-PA; SCGB3A2; EGFR; IGFBP-7; CD93; IL-18BP; COL1A1; PON3; CTSZ; MMP-3; RARRES2; ICAM-2; KLK6; PDGF subunit A; TNF-R1; IGFBP-2; vWF; PECAM-1; MEPE; CCL16" Genetics rs641738; rs738409; rs12979860; rs58542926 Metabolomic and lipidomic Lysophosphatidylcholines (acyl or ether bond) (8 species); Phosphatidylcholines (diacyl bonds) (30 species); Lysophosphatidylglycerols (alkyl bond) (4 species); Phosphatidylglycerols (diacyl bonds) (16 species); Phosphatidylglycerols (acyl / ether bonds) (6 species); Lysophosphatidylethanolamines (acyl or ether bond) (12 species); Phosphatidylethanolamines (diacyl bonds) (32 species); Phosphatidylcholines (acyl / ether bonds) (22 species); Phosphatidylserines (diacyl bonds) (26 species); Phosphatidylserines (acyl / ether bonds) (6 species); Phosphatidylethanolamines (acyl / ethyl bonds) (26 species); Sphingomyelins (66 species) ; Ceramides ( 46 species); N-C17:1-Cer (2 species); 2- Hydroxyacyl-ceramides (28 species); 2-Hydroxyacyl-dihydroceramides (24 species); Dihydroceramides (46 species) ; Free Fatty Acids (64 species); Free Oxysterols (36 species) Transcriptomic (mi-RNA) hsa-let-7a-5p; hsa-let-7b-3p; hsa-let-7b-5p; hsa-let-7c-5p; hsa-let-7d-3p; hsa-let-7d-5p; hsa-let-7e-5p; hsa-let-7f-5p; hsa-let-7g-5p; hsa-let-7i-5p; hsa-miR-1; hsa-miR-100-5p; hsa- miR-101-3p; hsa-miR-103a-3p; hsa-miR-106a-5p; hsa-miR-106b-3p; hsa-miR-106b-5p; hsa- miR-107; hsa-miR-10b-5p; hsa-miR-122-5p; hsa-miR-125a-5p; hsa-miR-125b-5p; hsa-miR- 1260a; hsa-miR-126-3p; hsa-miR-126-5p; hsa-miR-127-3p; hsa-miR-128-3p; hsa-miR-1290; hsa-miR-130a-3p; hsa-miR-130b-3p; hsa-miR-132-3p; hsa-miR-133a-3p; hsa-miR-133b; hsa- miR-136-3p; hsa-miR-136-5p; hsa-miR-139-5p; hsa-miR-140-3p; hsa-miR-140-5p; hsa-miR- 141-3p; hsa-miR-142-3p; hsa-miR-142-5p; hsa-miR-143-3p; hsa-miR-144-3p; hsa-miR-144- 5p; hsa-miR-145-5p; hsa-miR-146a-5p; hsa-miR-146b-5p; hsa-miR-148a-3p; hsa-miR-148b- 3p; hsa-miR-150-5p; hsa-miR-151a-3p; hsa-miR-151a-5p; hsa-miR-152-3p; hsa-miR-154-5p; hsa-miR-155-5p; hsa-miR-15a-5p; hsa-miR-15b-3p; hsa-miR-15b-5p; hsa-miR-16-2-3p; hsa- miR-16-5p; hsa-miR-17-5p; hsa-miR-181a-5p; hsa-miR-181b-5p; hsa-miR-181d-5p; hsa- miR-185-5p; hsa-miR-186-5p; hsa-miR-18a-5p; hsa-miR-18b-5p; hsa-miR-191-5p; hsa-miR- 192-5p; hsa-miR-193a-5p; hsa-miR-193b-3p; hsa-miR-194-5p; hsa-miR-195-5p; hsa-miR- 197-3p; hsa-miR-199a-3p; hsa-miR-199a-5p; hsa-miR-199b-5p; hsa-miR-19a-3p; hsa-miR- 19b-3p; hsa-miR-200a-3p; hsa-miR-200b-3p; hsa-miR-200c-3p; hsa-miR-205-5p; hsa-miR- 208a-3p; hsa-miR-20a-5p; hsa-miR-20b-5p; hsa-miR-210-3p; hsa-miR-2110; hsa-miR-215- 5p; hsa-miR-21-5p; hsa-miR-221-3p; hsa-miR-222-3p; hsa-miR-223-3p; hsa-miR-223-5p; hsa-miR-22-3p; hsa-miR-22-5p; hsa-miR-23a-3p; hsa-miR-23b-3p; hsa-miR-24-3p; hsa-miR- 25-3p; hsa-miR-26a-5p; hsa-miR-26b-5p; hsa-miR-27a-3p; hsa-miR-27b-3p; hsa-miR-28-3p; hsa-miR-28-5p; hsa-miR-29a-3p; hsa-miR-29b-3p; hsa-miR-29c-3p; hsa-miR-301a-3p; hsa- miR-30a-5p; hsa-miR-30b-5p; hsa-miR-30c-5p; hsa-miR-30d-5p; hsa-miR-30e-3p; hsa-miR- 30e-5p; hsa-miR-320a; hsa-miR-320b; hsa-miR-320c; hsa-miR-320d; hsa-miR-324-3p; hsa- miR-324-5p; hsa-miR-32-5p; hsa-miR-326; hsa-miR-328-3p; hsa-miR-331-3p; hsa-miR-335- 3p; hsa-miR-335-5p; hsa-miR-338-3p; hsa-miR-339-3p; hsa-miR-339-5p; hsa-miR-33a-5p; hsa-miR-33b-5p; hsa-miR-342-3p; hsa-miR-34a-5p; hsa-miR-361-5p; hsa-miR-362-3p; hsa- miR-363-3p; hsa-miR-365a-3p; hsa-miR-374a-5p; hsa-miR-374b-5p; hsa-miR-375; hsa-miR- 376a-3p; hsa-miR-376c-3p; hsa-miR-382-5p; hsa-miR-409-3p; hsa-miR-421; hsa-miR-423- 3p; hsa-miR-423-5p; hsa-miR-424-5p; hsa-miR-425-3p; hsa-miR-425-5p; hsa-miR-451a; hsa- miR-454-3p; hsa-miR-4668-3p; hsa-miR-483-5p; hsa-miR-484; hsa-miR-485-3p; hsa-miR- 486-5p; hsa-miR-495-3p; hsa-miR-497-5p; hsa-miR-501-3p; hsa-miR-502-3p; hsa-miR-505- 3p; hsa-miR-532-3p; hsa-miR-532-5p; hsa-miR-543; hsa-miR-574-3p; hsa-miR-584-5p; hsa- miR-590-5p; hsa-miR-629-5p; hsa-miR-652-3p; hsa-miR-660-5p; hsa-miR-7-1-3p; hsa-miR- 7-5p; hsa-miR-766-3p; hsa-miR-874-3p; hsa-miR-877-5p; hsa-miR-885-5p; hsa-miR-92a-3p; hsa-miR-92b-3p; hsa-miR-93-3p; hsa-miR-93-5p; hsa-miR-99a-5p; hsa-miR-99b-5p; mmu- miR-378a-3p; UniSp6 CP Pre-processing of data miRNAs The Exiqon lab performed real-time PCR panel analysis of miRNAs on the serum / plasma samples. 187 miRNAs were analyzed. An assay detected 5 Cqs (Cq = quantification cycles) lower than the negative control was included in the data analysis. For assays that did not yield any signal on the negative control, the upper limit of detection was set at Cq = 35. We retained miRNAs with 75% of quantified data in at least one of the two groups, 126 miRNAs. A random value imputed missing data (unquantified miRNAs) from a uniform distribution between 37 and 40. Then, data are normalized. For the normalization, 5 miRNAs have been selected (most stable and most expressed miRNAs). For a sample and a miRNA , the corresponding normalized value is equal to the difference between the mean of the 5 stable miRNAs of sample and the value observed for sample and miRNA . Thus, a higher normalized value indicates that the miRNA is more abundant in this particular sample . Variables from this family are considered as quantitative variables. Metabolites The Biocrates lab applied a mass spectrometric-based metabolomic approach to determine the concentration of several analyte classes, including lipids, free fatty acids, and oxysterols. In short, 374 metabolites have been analyzed by Biocrates from serum samples. Furthermore, the Evotec lab analyzed 3 metabolites with a XLC-MS / MS method for the absolute quantification in serum samples. Firstly, a transformation was applied to thedata set. As some values could be ( ), these values were replaced by. For each metabolite, the limit of detection was known. We have retained metabolites having 75% of values higher than the limit of detection in at least one of the two groups, that is, 163. Also (for the 163), we identified metabolites for which one group has 25% or more data lower than the limit of detection. We controlled with an exact Fisher's test from these metabolites if the rate of data lower than the limit of detection is significantly different between the 2 groups. If the difference was significant, the active metabolite was kept in the database; else it was removed. In the end, 153 metabolites have been retained for the analysis. Note that, Desmos metabolite was removed from the database because of its interaction with Amiodarone drug. Variables from this family are considered as quantitative variables. Proteins The Firalis lab analyzed 9 proteins with Luminex or Elisa kits from plasma samples. The Evotec lab analyzed 5 proteins with the Elisa method. Sanofi analyzed 92 proteins from an Inflammatory panel and 92 proteins from a cardiovascular disease panel by P.E.A. technology on plasma samples. No data were missing; hence no data imputation has been done. We have retained proteins with 75% of values higher than the limit of detection in at least one of the two groups; 165 proteins were retained. Note that some proteins were analyzed several times due to the use of kits / panels. Duplicates have been removed. Variables from this family are considered as quantitative variables. Genotypes The CNRS UMR 8199 lab analyzed 4 SNPs by Taqman probes from D.N.A. samples. Very few values were missing (2 or 3 per SNP). Hence no data imputation has been done. Variables from this family are considered qualitative variables. Screening of biomarkers candidates Given the absence of data regarding the analyzed biomarkers, the sample size estimation for the screening phase was based on the minimal effect size. According to Cohen (Cohen, 1988), an effect size of 0.5 is considered medium while 0.8 is considered significant. Considering a type I error of 5%, a power of 80%, 40 patients per group were required to detect an effect size greater or equal to 0.65. To reduce investigator bias, we randomly selected the patients with NASH in the cohort and, each NASH patient was matched to a NAFL patient based on age (+ / - 5 years), gender, and body mass index (+ / -5 Kg / m2), using the overall optimal algorithm (Rosenbaum, 1989). Forty-three clinical variables (demographics, anthropometric and laboratory results) and 764 biomarkers (196 proteins, 377 metabolites / lipids, 187 miRNAs, and 4 genotypes) from serum, plasma, or D.N.A. samples have been analyzed for the two matched groups. As the variables analyzed were biomarkers from Omics technology, it was necessary to perform a preprocessing (cleaning, normalization, transformation) for this data type. This step is described in detail in the section Pre-processing of data. Statistical analyses were performed using the R Software. The analyses were carried out by family of variables. We performed univariate analyses (Mann-Whitney tests and Chi- square tests or Fisher's exact tests) followed by multiple testing procedures. For the proteins, metabolites, and miRNAs, the Benjamini-Hochberg false discovery rate method (Benjamini & Hochberg, 1995) was used to control multiple hypothesis testing. For the genotypes, the Bonferroni family-wise error rate was used to control for multiple hypotheses testing. No control for multiple hypotheses testing was done for the clinical variables. P-values (or corrected p-values for omics data) were considered significant if lower than 0.05. Volcano plots, which show p-values and fold-changes between two groups, were plotted. We also performed multivariate analysis using penalized logistic regression framework based on LASSO (with the glmnet package) ( (Tibshirani, 1996), (Friedman, Hastie, & Tibshirani, 2010)). We implemented stability selection to select variables that were often selected, whatever the subsampling was (Meinshausen & Bühlmann, 2010). Correlations between quantitative variables were described. Based on the results of the univariate statistical analyses (groups' comparison, volcano plot), the multivariable analysis (stability selection implemented to lasso), the correlation results (in order to avoid multicollinearity), and a discussion with the clinicians (clinical relevance, facility of dosage, reproducibility), we selected the 10 most relevant variables for the discrimination between NAFL and NASH patients. Because miRNA determination is not yet routine practice in clinical laboratories, we also considered a set of variables without miRNAs. In another strategy, the selection of variables was carried out on all the variables (all families combined) and based on multivariate analysis (with and without miRNAs). This procedure will allow us to build four different sets of candidate predictors, which will be assessed on the derivation cohort. Derivation of the preciNASH score We developed a multivariable predictive model by considering all candidate predictors identified during the screening phase irrespective of their univariate associations with the diagnosis status (NASH / NAFL) and by handling missing data using multiple imputations. (a) The first stage of the procedure consisted in creating imputed datasets ( )from the original dataset using a regression switching approach (chained equation) under missing at random hypothesis and including the predictors selected by the screening procedure (including the diagnosis status and the matching variables) in the imputation model. Imputations were done with predictive mean matching method for quantitative variables and logistic regression models (binary, ordinal, or polynomial) for categorical variables (Van Buuren & Groothuis-Oudshoorn, 2011). (b) A multivariable logistic regression with backward selection at the 0.05 level was performed on the 10 imputed datasets according to the following procedure: firstly, all variables were entered in the model and we applied the Rubin’s rules to combine the regression coefficients estimates (Marshall, Altman, Holder, & al., 2009). The variable with greatest p-value was removed if its p-value was greater than 0.05. This process was repeated until no variable could be removed (at the 0.05 level). (c) Multivariable logistic regression was performed in each imputed dataset with the predictors selected using the backward procedure. We examined the performance of the selected model in terms of calibration using the Hosmer-Lemeshow (HL) goodness of fit-test and a visual inspection of the calibration plot, and in terms of discrimination by calculating the AUC (Area Under the ROC Curve). Finally, Rubin’s rules were used to combine the estimates derived from these multivariable logistic regressions and build the final model. Median of AUCs and median of p-values of Hosmer-Lemeshow test were considered as a summary of performances of the final model in term of discrimination and calibration. (d) We considered the ROC curve corresponding to the median of the AUCs in the 10 imputed datasets to identify three thresholds. The first threshold was based on the Youden index for optimizing the discrimination. The other two thresholds were determined using the likelihood ratios (positive LR+ and negative LR-) (Deeks & Altman, 2004): one threshold allowed to identify with a good accuracy the NASH patients (threshold from which the LR+ exceeds 10 with a minimum LR-) and the other threshold allowed to exclude with good accuracy the patients without NASH (among all the LR- less than to 0.1 we take the threshold which maximizes the LR+). (e) Internal validation was done by using bootstrap resampling combined with handling of missing data to estimate the AUC corrected for the over-optimism and the shrinkage factor. The shrinkage factor is the estimated shrinking in the regression coefficients to improve the prediction in future patients (Dunkler, Sauerbrei, & Heinze, 2016). Denote by the AUC calculated in the imputed dataset with the procedure described in (a). We proceeded according to the following steps: a. Bootstrap resampling with 200 repetitions of the imputed dataset ().b. In each bootstrapped sample ( ), perform a multivariablelogistic regression with backward selection at 0.05 level. Compute the AUC of this model denoted . c. Apply this model to compute the logit for each individual on the imputed data and compute the AUC associated to this logit, denoted by . For the shrinkage factor, perform the logistic regression on the imputed data with the status (NASH / NAFL) as dependent variable and the logit as independent variable to compute the slope, namely the shrinkage factor, denoted by . d. Compute the bias and the mean of on the bootstrapped samples, denoted by . the means of the bootstrapped samples, denoted . e. Then is the AUC of the model developed on base corrected for over-optimism bias and is the shrinkage factor of the model. f. Finally, we compute the median of the 10 and the mean of 10 shrinkage factors .

[0004] Table 5 - Characteristics of 80 patients from the ABOS cohort included in the screening step of biomarkers Overall NAFL patients NASH patients Characteristics (N = 80) (N = 40) (N = 40) Female gender 56 / 80 (70.0) 29 (72.50) 29 (72.50) Age (years) 2 49.0 [37.5 ; 54.5] 49.0 [38.0 ; 54.5] 48.5 [37.5 ; 55.0] BMI (kg / m ) 45.5 ± 5.7 45.1 ± 5.4 45.9 ± 6.0 Diabetes status No diabetes 6 / 80 (7.5) 6 / 40 (15.0) 0 / 40 (0.0 Insulin resistant 28 / 80 (35.0) 16 / 40 (40.0) 12 / 40 (30.0) Diabetes mellitus 46 / 80 (57.5) 18 / 40 (45.0) 28 / 40 (70.0) Arterial hypertension 59 / 80 (73.8) 24 / 40 (60.0) 35 / 40 (87.5) Dyslipidaemia 58 / 80 (72.5) 28 / 40 (70.0) 30 / 40 (75.0) Biological features HDL cholesterol13, 1.1 [0.9 ; 1.2] 1.1 [1.0 ; 1.2] 1.1 [0.9 ; 1.2] LDL chol 1esterol , 2.9 ± 0.9 3.0 ± 0.9 2.8 ± 0.9 Triglycerides , 1.8 [1.2 ; 2.6] 1.6 [1.2 ; 2.4] 1.9 [1.3 ; 2.8] Total cholesterol1, 4.9 ± 1.0 4.9 ± 1.0 4.8 ± 1.1 AST, IU / L 31.0 [23.0 ; 40.0] 24.0 [20.0 ; 35.0] 37.5 [28.5 ; 46.0] ALT, IU / L 40.0 [27.0 ; 51.0] 27.5 [20.0 ; 43.0] 46.0 [36.5 ; 59.0] Gamma GT, IU / L 49.0 [29.5 ; 78.5] 33.0 [24.0 ; 57.0] 58.0 [42.0 ; 96.5] Bilirubin, mg / dL 4 4.0 [3.0 ; 5.0] 4.0 [3.0 ; 5.0] 4.0 [3.0 ; 5.0] Fasting glucose , 6.4 [5.4 ; 11.1] 5.8 [5.2 ; 7.2] 8.1 [5.9 ; 12.9] Hemoglo 1bin A1c, % 6.3 [5.8 ; 8.7] 6.2 [5.7 ; 7.6] 7.0 [5.9 ; 10.0] Albumin , g / L 43.7 ± 2.6 44.0 ± 2.5 43.5 ± 2.6 Platelets, g / L 1 265 ± 61.7 268.3 ± 64.8 261.6 ± 59.1 Prothrombin **time , 100.0 [96.0 ; 100.0] 100.0 [98.0 ; 100.0] 100.0 [96.0 ; 100.0] HOMA IR 3.1 [2.3 ; 4.9] 2.7 [1.8 ; 4.2] 3.7 [2.7 ; 5.4] Histologic features Biopsy lengt 1h2, mm 13.0 [10.0 ; 16.0] 13.5 [11.0 ; 17.0] 13.0 [10.0 ; 16.0] Portal triads 9.0 [6.0 ; 11.0] 10.0 [8.0 ; 11.5] 7.0 [5.0 ; 10.0] NAS score 3.0 [1.0 ; 5.0] 1.0 [1.0 ; 2.0] 5.0 [4.0 ; 6.0] 1 – 2 37 / 80 (46.3) 37 (92.5) 0 (0.0) 3 – 4 18 / 80 (22.5) 3 (7.5) 15 (37.5) 5 – 8 25 / 80 (31.3) 0 (0.0) 25 (62.5) Steatosis, % 40 [25 ; 70] 25.0 [12.5 ; 40.0] 65.0 [37.5 ; 72.5] Ballonning 0 40 / 80 (50.0) 40 / 40 (100.0) 0 / 40 (0.0) 1 30 / 80 (37.5) 0 / 40 (0.0) 30 / 40 (75.0) 2 10 / 80 (12.5) 0 / 40 (0.0) 10 / 40 (25.0) Lobular 0 40 / 80 (50.0) 40 / 40 (100.0) 0 / 40 (0.0) 1 25 / 80 (31.3) 0 / 40 (0.0) 25 / 40 (62.5) 2 14 / 80 (17.5) 0 / 40 (0.0) 14 / 40 (35.0) 3 1 / 80 (1.3) 0 / 40 (0.0) 1 / 40 (2.5) Steatosis 1 35 / 80 (43.8) 27 / 40 (67.5) 8 / 40 (20.0) 2 22 / 80 (27.5) 10 / 40 (25.0) 12 / 40 (30.0) 3 23 / 80 (28.8) 3 / 40 (7.5) 20 / 40 (50.0) Fibrosis (Kleiner’s 0 35 / 75 (46.7) 30 / 39 (76.9) 5 / 36 (13.9) 1a 6 / 75 (8.0) 1 / 39 (2.6) 5 / 36 (13.9) 1b 5 / 75 (6.7) 1 / 39 (2.6) 4 / 36 (11.1) 1c 5 / 75 (6.7) 3 / 39 (7.7) 2 / 36 (5.6) 2 10 / 75 (13.3) 2 / 39 (5.1) 8 / 36 (22.2) 3q 8 / 75 (10.7) 2 / 39 (5.1) 6 / 36 (16.7) 3s 5 / 75 (6.7) 0 / 39 (0.0) 5 / 36 (13.9) 4 1 / 75 (1.3) 0 / 39 (0.0) 1 / 36 (2.8) Values are expressed as the number / total number (%), mean ± standard-deviation or median [interquartile range]. Abbreviations: BMI, body mass index; ALT, alanine aminotransferase ; AST, aspartate aminotransferase ; GGT, γ-glutamyltransferase ; HbA1c, glycated hemoglobin ; HDL, high- density lipoprotein ; LDL, low-density lipoprotein11 missing value ;22 missing values ;33 missing values ;44 missing values ;**6 missing values ;##20 missing values Table 6 - Description of pre-selected variables at the screening step on the 80 matched patients Overall NAFL patients NASH patients Effect size (N = 80) (N = 40) (N = 40) Fasting blood glucose >7 34 / 76 (44.7) 11 / 37 (29.7) 23 / 39 (59.0) - Fasting blood glucose, 6.31 [5.39 ; 10.41] 5.83 [5.27 ; 7.23] 8.11 [5.89 ; 12.38] 0.60 mmol / L Gamma G. T. 49.0 [29.5 ; 78.5] 33.0 [24.0 ; 57.0] 58.0 [42.0 ; 96.5] 0.80 ALT, IU / L 40.0 [27.0 ; 51.0] 27.5 [20.0 ; 42.0] 46.0 [37.75 ; 58.5] 0.86 Proteins MAC2BP, pg / mL 205124 [159628; 167089 [135645; 258078 [205305 ; 291 0.52 279314] 204084] 056] CK18, IU / L 403.1 [298.2 ; 667.8] 301.8 [221.8 ; 401.5] 572.5 [403.5 ; 1368.4] 1.28 CDCP1 7.51 [5.33 ; 11.03] 5.56 [4.24 ; 7.48] 9.72 [7.66 ; 13.48] 0.84 Metabolites 7 KC, µM 0.016 [0.012 ; 0.019] 0.013 [0.010 ; 0.016] 0.017 [0.015 ; 0.023] 0.51 Betaine, ng / mL 3222 [2561 ; 3908] 3501 [2955 ; 4030] 2892 [2015 ; 3449] 0.67 NC17 0.010 [0.008 ; 0.013] 0.009 [0.006 ; 0.011] 0.011 [0.009 ; 0.014] 0.71 Genotype rs641738 CC 26 / 79 (32.9) 19 / 39 (48.7) 7 / 40 (17.5) CT / TT 53 / 79 (67.1) 20 / 39 (51.3) 33 / 40 (82.5) 4.48 miR 34a 5p, Relative - 4.3 ± 1.9 - 5.1 ± 1.5 - 3.5 ± 2.0 1.62 Units Values are expressed as the number / total number (%), mean ± standard-deviation, or median [interquartile range]. Abbreviations: ALT, alanine aminotransferase ; Gamma GT, γ-glutamyltransferase Effect sizes are expressed as log fold change for proteins, metabolites and miRNA, as OR for the genotype and as Edge’s G for clinical variables. Table 7 – Complete models on the imputed datasets on the derivation cohort Subset 1.1 Subset 1.2 OR (CI 95%) p-values OR (CI 95%) p-values Intercept -16.69 (-20.10 to -13.29) <0.001 -15.21 (-19.47 to -10.95) < 0.001 ALT (log) 0.43 (-0.32 to 1.19) 0.26 0.39 (-0.37 to 1.14) 0.31 Gamma GT (log) 0.79 (0.28 to 1.31) 0.003 0.79 (0.27 to 1.31) 0.003 Fasting glucose (<7 mmol / L) 1.43 (0.83 to 2.03) < 0.001 1.38 (0.77 to 1.99) < 0.001 MAC2BP (log2) 0.68 (0.21 to 1.14) 0.004 0.66 (0.20 to 1.12) 0.005 CK18 (log2) 0.97 (0.62 to 1.33) <0.001 0.88 (0.49 to 1.28) < 0.001 MBOAT7 0.67 (-0.06 to 1.41) 0.072 0.67 (-0.06 to 1.40) 0.073 Mir.34a.5p - - 0.12 (-0.10 to 0.35) 0.274 Example 2 Example with 5 variables Multivariate logistic regressions with backward selection, performed on the three subsets of candidate variables in parallel, selected the same 5 following variables: gamma G.T., fasting blood glucose (F.B.G.), galectin 3 protein, CK18 protein, and the 7KC metabolite. Therefore, we developed a single score that exhibits a corrected for over-optimism A.U.C. of 0.91 (median on the multiple imputations), which can be considered as promising an excellent diagnostic ability. The ROC curve associated with the median of AUC over the 10 imputed datasets is presented in Figure 5a . The optimal cut-off, determined by maximizing the Youden index, was 0.156 leading to a LR+ of 4.71 and a LR- of 0.10. The sensitivity was 0.92 and the specificity was 0.76. Two other thresholds were determined: one, equals to 0.45, to identify with a good accuracy the NASH patients was associated to LR+ of 10.2, leading to a specificity of 0.94.1% and 87.5% of well-classified patients; The other threshold, equals to 0.08, to exclude with good accuracy the patients without NASH was associated with an LR-of 0.1, leading to a sensitivity of 0.94.

[0005] REFERENCES: Throughout this application, various references describe the state of the art to which this invention pertains. The disclosures of these references are hereby incorporated by reference into the present disclosure. 1. Rinella ME & (2023). NAFLD Nomenclature consensus group. A multisociety Delphi consensus statement on new fatty liver disease nomenclature. J Hepatol. 79(6):1542-1556. doi: 10.1016 / j.jhep.2023.06.003. 2. Benjamini, Y. &. (1995). Controlling the false discovery rate: a practical and powerful approach to multiple testing. Journal of the Royal statistical society: series B (Methodological), 57(1), 289-300. Doi : 10.1111 / j.2517-6161.1995.tb02031.x 3. Cohen, J. (1988). Statistical power analysis for the behavorial sciences.doi : 10.4324 / 9780203771587 4. Deeks, J. J., & Altman, D. G. (2004). Diagnostic tests 4: likelihood ratios. BMJ, 329(7458), 168-169. doi: 10.1136 / bmj.329.7458.168 5. Friedman, J., Hastie, T., & Tibshirani, R. (2010). Regularization paths for generalized linear models via coordinate descent. Journal of statistical software, 33(1), 1.doi : 10.18637 / jss.v033.i01 6. Meinshausen, N., & Bühlmann, P. (2010). Stability selection. . Journal of the Royal Statistical Society: Series B (Statistical Methodology), 72(4), 417-473.doi : 10.1111 / j.1467-9868.2010.00740.x 7. Peduzzi, P., Concato, J., Kemper, E., Holford, T. R., & Feinstein, A. R. (1996). A simulation study of the number of events per variable in logistic regression analysis. Journal of clinical epidemiology, 49(12), 1373-1379.DOI: 10.1016 / s0895- 4356(96)00236-3 8. Rosenbaum, P. R. (1989). Optimal matching for observational studies. Journal of the American Statistical Association, 84(408), 1024-1032. Doi : 10.2307 / 2290079 9. Rosenbaum, P., & Rubin, D. (1985). Constructing a control group using multivariate matched sampling methods that incorporate the propensity score. The American Statistician, 39(1), 33-38. Doi : 10.2307 / 2683903 10. Tibshirani, R. (1996). Regression shrinkage and selection via the lasso. Journal of the Royal Statistical Society: Series B (Methodological), 58(1), 267-288.doi : 10.1111 / j.2517-6161.1996.tb02080.x

Claims

CLAIMS:

1. An in vitro method for assessing a subject’s risk of having or developing NASH (non-alcoholic steatohepatitis) disease, comprising the steps of i) determining in a blood sample obtained from the subject the level of the 2 biological serum markers Galectin3 binding protein (Galectin3- BP) and cytokeratin-18 (CK-18), ii) combining said levels by implementing a statistical technique on said level, said statistical technique involving a logistic regression in order to obtain an end value (global NASH Score) iii) comparing said score to a predetermined first reference value, and iv) when is higher than or equal to said first reference value said score is predictive of a high risk of having or developing NASH (non- alcoholic steatohepatitis) disease.

2. The in vitro method according to claim 1, comprising comparing said score with a predetermined second reference value and concluding that when said score is lower than or equal to said second reference value said score is predictive of a low risk of having or developing a NASH (non-alcoholic steatohepatitis) disease.

3. The in vitro method according to claim 1 or 2 wherein in an additional step of determining in a blood sample obtained from the subject, the level of the gamma- glutamyltranspeptidase (gamma GT) and / or the glycemia (Fasting Blood Glucose) higher than 7 mmol / L and combining the levels of biological and clinical marker previously measured by implementing a statistical technique on said levels, said statistical technique involving a logistic regression in order to obtain an end value (global NASH Score).

4. The in vitro method according to claim 1 to 3, wherein in an additional step determining in a blood sample obtained from the subject, the level 7- Ketocholesterol (7-KC) (lipid marker) and combining the levels of biological and / or clinical marker previously measured by implementing a statistical technique on said levels, said statistical technique involving through a logistic regression function in order to obtain an end value (global NASH Score). 5 The in vitro method according to claim 1 to 4, wherein said method is performed using the level of 2, 3, 4 or 5 different biomarkers including the level of Galectin3binding protein (Galectin3- BP), cytokeratin-18 (CK-18) and optionally the level of 7-Ketocholesterol (7-KC) and / or the level of gamma-glutamyltranspeptidase (gamma GT) and / or the level of glycemia (Fasting Blood Glucose).

6. The in vitro method according to claim 5, wherein said method is performed using the level of 4 or 5 different biomarkers including the level of Galectin3 binding protein (Galectin3- BP), cytokeratin-18 (CK-18) gamma-glutamyltranspeptidase (gamma GT) and glycemia (Fasting Blood Glucose) and optionally the level of 7-Ketocholesterol (7-KC).

7. The in vitro method according to claim 6, wherein said method is performed using the level the level of Galectin3 binding protein (Galectin3- BP), cytokeratin-18 (CK-18) gamma-glutamyltranspeptidase (gamma GT) and glycemia (Fasting Blood Glucose).

8. An in vitro method for monitoring NASH (non-alcoholic steatohepatitis) disease comprising the steps of i) determining the level of the 2 serum marker Galectin3 binding protein (Galectin3- BP) and cytokeratin-18 (CK-18) (protein markers) in a blood sample obtained from a subject at a first specific time of the disease, ii) combining said levels by implementing a statistical technique on said levels, said statistical technique involving a logistic regression in order to obtain an end value (global NASH Score) ,iii) determining the level of the same biological serum markers determined in step i) in a blood sample obtained from the subject at a second specific time of the disease, iv) combining said levels by implementing a statistical technique on said levels, said statistical technique involving a logistic regression in order to obtain an end value, v) comparing the end value determined at step ii) with the end value determined at step iv) and vi) concluding that the disease has evolved in worse manner when the end value determined at step ii) is higher than the end value determined at step i).

9. An in vitro method for monitoring the treatment of a NASH (non-alcoholic steatohepatitis) disease comprising the steps of i) determining the level of the 2 serum marker Galectin3 binding protein (Galectin3- BP) and cytokeratin-18 (CK- 18) (protein markers) in a blood sample obtained from a subject before thetreatment, ii) combining said levels by implementing a statistical technique on said levels, said statistical technique involving a logistic regression in order to obtain an end value (global NASH Score) iii) determining the level of the same serum biological markers determined in step i) in a blood sample obtained from the subject after the treatment, iv) combining said levels in by implementing a statistical technique on said levels said statistical technique involving a logistic regression in order to obtain an end value v) comparing the end value determined at step iii) with the end value determined at step iv) and vi) concluding that the treatment is efficient when the end value determined at step iii) are lower than the end value determined at step iv).

10. The in vitro method according to any one of claim 8 or 9, wherein in an additional step determining in a blood sample obtained from the subject, the blood level of the gamma-glutamyltranspeptidase (gamma GT) and / or the glycemia higher than 7 mmol / L and combining the blood levels of biological and clinical markers previously measured by implementing a statistical technique on said levels, said statistical technique involving a logistic regression in order to obtain an end value (global NASH Score).

11. The in vitro method according to any one of claim 8 to 10 wherein in an additional step determining in a blood sample obtained from the subject, the level 7-Ketocholesterol (7-KC) (lipid marker).

12. The in vitro method according to any one of claim 8 to 11, wherein said method is performed using the level of 2, 3, 4 or 5 different biomarkers including the level of Galectin3 binding protein (Galectin3- BP), cytokeratin-18 (CK-18) and optionally the level of 7-Ketocholesterol (7 KC) and / or the level of gamma- glutamyltranspeptidase (gamma GT) and / or glycemia ((Free Blood Glucose).

13. The in vitro method according to claim 12, wherein said method is performed using the level of 4 or 5 different biomarkers including the level of Galectin3 binding protein (Galectin3- BP), cytokeratin-18 (CK-18) gamma- glutamyltranspeptidase (gamma GT) and glycemia (Fasting Blood Glucose) and optionally the level of 7-Ketocholesterol (7-KC).

14. The in vitro method according to claim 13, wherein said method is performed using the level the level of Galectin3 binding protein (Galectin3- BP), cytokeratin- 18 (CK-18) gamma-glutamyltranspeptidase (gamma GT) and glycemia (Fasting Blood Glucose).

Images