Prognostic method for hematological malignancies

Abstract

The invention relates to a method for determining the risk of developing a hematological malignancy, comprising the following steps: - exposing a biological sample to radiation to obtain a characteristic spectrum of said sample, said spectrum being processed to obtain a spectral signature; - comparing said spectral signature obtained in the previous step with one or more reference spectral signatures; and - concluding that * if the spectral signature of said individual is significantly different from control spectral signatures, then the individual is likely to develop a hematological malignancy, * otherwise the individual is not likely to develop a hematological malignancy. Figure for abstract: no figure

Description

Title of the invention: Prognostic method for blood diseases

[0001] The invention relates to a prognostic method for blood diseases.

[0002] The field of blood diseases generally requires early screening, diagnosis and monitoring of the progression of the disease to a more serious condition, and involves analysis that can be performed on a specific biological sample, which is generally in addition to numerous blood tests.

[0003] Myelodysplastic syndromes (MDS) are pre-leukemic conditions whose frequency increases with the aging of the population and which present a progressive evolution with transformation into secondary acute leukemia in 30% of cases. Their diagnosis requires a bone marrow sample to carry out a cytological analysis by a biologist specialized in hematology (myelogram).

[0004] Such analyses are invasive, time-consuming and expensive, and it is important to be able to simplify the analyses and obtain faster results.

[0005] Fourier Transform Infrared (FTIR) spectroscopy is known and used to identify organic compounds and examine the biochemical composition of a biological sample (tissue or fluid).

[0006] Processes such as leukemogenesis can induce global changes in cellular biochemistry, leading to differences in absorption spectra when analyzed by FTIR spectroscopy techniques. Therefore, FTIR spectroscopy is commonly used to distinguish between normal and abnormal tissue by analyzing changes in absorption bands of macromolecules such as lipids, proteins, carbohydrates, and nucleic acids.

[0007] The teaching of application WO2011121588 is also known from the prior art, where a method and system for detecting and monitoring hematological cancer is described. More particularly, the inventors designated in this application have identified that mononuclear cell samples obtained from leukemia patients produce FTIR spectra that differ from those of healthy controls and non-cancer patients suffering from clinical symptoms that are similar to leukemia, for example, subjects suffering from a fever, thus allowing a differential diagnosis of leukemia patients. By distinguishing leukemia patients, patients with clinical symptoms similar to leukemia, and healthy controls, IR spectroscopy provides an effective diagnostic tool for the diagnosis of leukemia and / or other types of hematological malignancies.

[0008] However, although this document illustrates the usefulness of FTIR spectroscopy in the diagnosis of leukemia, this document does not provide tools allowing to refine the diagnosis according to subtypes of pathologies, or even to predict the progression of the disease.

[0009] Also, there is a need to provide a method for refining the detection of leukemia subtypes, and in particular myeloid leukemias.

[0010] The invention aims to overcome these shortcomings of the prior art.

[0011] One of the aims of the invention is to provide a prognostic method making it possible to simply determine the prognosis for the occurrence of a blood disease in an individual.

[0012] The invention therefore relates to a method for determining, in particular in vitro, the risk for an individual of developing a haemopathy from a biological sample of this individual, said method comprising the following steps:

[0013] - exposing said biological sample to mid-infrared radiation (MIR) of wavelength varying from 4000 cm 1 to 400 cm 1 to obtain a spectrum characteristic of said sample, said spectrum being processed in order to obtain a spectral signature consisting of absorbance peaks characteristic by their position, or wave number, and their intensity, or absorbance, of the nature and relative concentrations of the different molecules present in said sample;

[0014] - compare said spectral signature obtained in the previous step with one or several reference spectral signatures, said one or more reference spectral signatures being obtained from a population of reference individuals; and

[0015] - conclude that

[0016] * if the intensities of a first group of peaks of the spectral signature of said individual are significantly different from the intensities of these same peaks obtained in the reference spectral signature(s), then the individual is likely to develop a blood disease,

[0017] the first group of peaks corresponding to the wave numbers of the following first group: 1330 cm1, 1445 cm1, 1478 cm ', 1493 cm ', 1505 cm ', 1507 cm ', 1520 cm ', 1526 cm ', 1544 cm1, 1571 cm ', 1602 cm ', 1668 cm ', 1674 cm1, 1676 cm ', 1697 cm ', and 2852 cm1,

[0018] * otherwise the individual is not likely to develop a blood disease.

[0019] The invention is based on the surprising observation made by the inventors that the determination of an infrared spectrum of simple blood samples obtained from individuals makes it possible to obtain information concerning the risk of developing a haemopathy for said individual.

[0020] Spectroscopy is a simple and rapid process, which does not require any reagents to be implemented (apart from appropriate equipment), which makes it possible to obtain information concerning the macromolecular structure of the compounds contained in a biological sample. Typically, infrared (FTIR) spectra are composed of many absorption bands, each corresponding to specific functional groups associated with cellular components such as lipids, proteins, carbohydrates, and nucleic acids. Any physiological changes occurring in an individual, including carcinogenesis, can lead to global changes in metabolism, which will alter the absorption spectra when the sample is analyzed by FTIR techniques. Therefore, FTIR is commonly used to distinguish between normal and abnormal tissue by analyzing changes in the absorption bands of molecules.

[0021] The infrared portion of the electromagnetic spectrum is divided into three regions: near, mid, and far infrared, named in relation to the visible spectrum. The far infrared, ranging from approximately 400 to 10 cm-1 (1000-25 qm, in practice range 1000-30 qm), adjacent to the microwave region, has low energy and can be used for rotational spectroscopy. Mid-infrared radiation, ranging from approximately 4000 to 400 cm-1 (25-2.5 qm, in practice range 30-1.4 qm) can be used to study fundamental vibrations and the associated vibrational structure. The more energetic near infrared, ranging from approximately 14000 to 4000 cm-1 (2.5-0.7 qm, in practice range 1.4-0.8 qm) can excite harmonic vibrations. The names and classifications of these sub-regions are essentially conventions. In the invention, reference will be made to mid-infrared spectroscopy according to the definition above.

[0022] The infrared spectrum of a sample is established by passing a beam of infrared light through this sample. Examination of the transmitted light indicates the amount of energy absorbed at each wavelength. It can be carried out with a monochromatic beam, with a change in wavelength over time, or by using a Fourier transform instrument to measure all the absorbances simultaneously by interferometry. Spectra can then be produced in absorbance or transmittance, and the absorption wavelengths analyzed. Analysis of these characteristics reflects the molecular structures of the sample.

[0023] This technique works almost exclusively on samples with covalent bonds. Simple spectra are obtained from samples with few active bonds in the infrared and with high degrees of purity. More complex molecular structures lead to more absorption bands and therefore to more complex spectra but is still used for the characterization of very complex mixtures.

[0024] The method described in the present invention lies in the simplicity of the steps implemented:

[0025] - a first step consists of obtaining an infrared spectrum from a sample biological of an individual, and

[0026] - the second step consists of comparing the spectrum obtained in the previous step with one or more reference spectra in order to conclude on the risk or not of developing a blood disease.

[0027] In the invention, the term "hemopathy" means any pathology affecting the components of the blood, and in particular malignant hematopathies such as leukemias, lymphomas, myelomas as well as myelodysplastic and myeloproliferative syndromes.

[0028] In the invention the result of the exposure to infrared radiation will be processed, in particular by Fourier transform to obtain a spectral signature characteristic of a given sample.

[0029] More particularly, interest will be focused on the following specific wave numbers (inverse of the wavelength): 1330 cm1, 1445 cm1, 1478 cm ', 1493 cm ', 1505 cm ', 1507 cm ', 1520 cm ', 1526 cm ', 1544 cm ', 1571 cm ', 1602 cm ', 1668 cm *, 1674 cm ', 1676 cm ', 1697 cm ', and 2852 cm ', as well as the relative intensity of each of the peaks corresponding to these wave numbers following the Fourier transformation.

[0030] Also, to summarize, a given sample is subjected to mid-infrared radiation, to obtain a spectrum which will be processed by Fourier transform to obtain a spectral signature for at least the wave numbers 1330 cm ', 1445 cm ', 1478 cm ', 1493 cm ', 1505 cm1, 1507 cm ', 1520 cm ', 1526 cm ', 1544 cm ', 1571 cm ', 1602 cm ', 1668 cm ', 1674 cm ', 1676 cm ', 1697 cm ', and 2852 cm '.

[0031] Once this spectral signature is obtained, it is compared to a reference spectral signature, or several reference spectral signatures.

[0032] These reference spectral signatures are obtained from reference samples subjected to the same infrared treatment (and Fourier transform) as the sample analyzed. For the comparison to be most effective, it is essential that the reference spectral signatures be obtained from biological samples of the same nature (for example blood, serum, plasma, etc.) as the biological sample tested. Thus, for example, if a blood sample is tested for an individual according to the method of the invention, the reference sample(s) will be those obtained from other blood samples.

[0033] The reference spectral signatures are obtained from reference individuals who can be either healthy individuals, i.e. individuals not presenting any pathology, or individuals presenting a disease whose symptoms are different from those of a hemopathy, as defined in the invention.

[0034] A reference individual may also correspond to the individual tested according to the method of the invention, the reference samples having been taken before said individual has, or is likely to develop, a blood disease.

[0035] When comparing the spectral signature of the individual with the reference spectral signature(s), the intensity (absorbance) of the different peaks corresponding to the aforementioned wave numbers is compared.

[0036] From this comparison, it follows that if all the intensities of the peaks corresponding to said wave numbers are significantly different (increase or decrease) compared to the intensity of these same peaks in the reference spectral signatures, then said individual whose sample was tested according to the method of the invention will be likely to develop a hemopathy. The signature of a given individual is therefore constituted by a pattern-type structure which is distributed over a set of spectral variables (16 in the present case). A reference signature (healthy or pathological) is therefore constituted by a “profile”, the pattern. For each type of patient, a profile is identified which is specific to the physiological state of the individual. The identification of the physiological state of any individual is therefore based on the comparison of this profile (or pattern) with those of references.A distance calculation between the individual's pattern and the reference pattern(s) allows this individual to be assigned to a particular class / category (healthy or sick for example). The individual will be "classified" according to the reference pattern which is at the shortest distance / closest in a space with (here for example) 16 dimensions.

[0037] Also, on the basis of an FTIR spectrum obtained from a biological sample of blood, or a blood by-product, or bone marrow, it is possible to determine the risk of developing a malignant hematological disease in an individual.

[0038] Advantageously, the invention relates to the aforementioned method, where when the individual is likely to develop a hemopathy, it is further concluded that:

[0039] * if the intensities of a second group of peaks of the spectral signature of said individual are significantly different from the intensities of these same peaks obtained in the reference spectral signature(s), then the individual is likely to develop leukemia,

[0040] the second group of peaks corresponding to the wave numbers of the following first group: 3316 cm1, 3283 cm1, 3281 cm1, 3256 cm ', 3118 cm ', 3116 cm ', 1345 cm1, 1343 cm ', 1340 cm1 and 1338 cm ',

[0041] * otherwise the individual is likely to develop myelodysplastic syndrome.

[0042] The inventors have demonstrated that if the first group of peaks, or wave number, makes it possible to determine whether an individual is likely to develop a haemopathy, it is possible by studying a second group of peaks of the spectral signature and according to the differences obtained to determine whether the individual in question, tested according to the method of the invention, is likely to develop leukaemia or a myelodysplastic syndrome.

[0043] Myelodysplastic syndromes (MDS) are acquired clonal hematopathies of the bone marrow hematopoietic stem cell, with excessive proliferation of myeloid progenitors that differentiate abnormally (= dysmyelopoiesis). Excessive apoptosis of the precursors leads to defective production and peripheral cytopenias (= ineffective hematopoiesis).

[0044] There are several classes, defined by the WHO (2016) according to the nature and number of cytopenias, signs of myelodysplasia (morphological abnormalities of bone marrow cells), and the presence or absence of an excess of blasts.

[0045] The evolution is prolonged and relatively indolent in 70% of cases, with progressive worsening of cytopenias (bone marrow failure). In 30% of cases the evolution is faster and more aggressive towards acute myeloid leukemia by accumulation of blasts, explaining why MDS are also called "preleukemic states".

[0046] According to this embodiment, the study of the first two groups of wave numbers of the spectral signature does not make it possible to distinguish so-called “low risk” MDS from so-called “high risk” MDS of transition to secondary leukemia.

[0047] In the invention, two types of leukemia are distinguished in particular, in particular two types of acute myeloid leukemia: de novo acute myeloid leukemia, and secondary myeloid leukemia.

[0048] De novo leukemias occur in patients spontaneously, and directly without the patient having previously detected a myeloproliferative syndrome. Such leukemias may appear due to the simultaneous accumulation of abnormalities affecting the proliferation and cellular differentiation of myeloid progenitors.

[0049] Secondary acute myeloid leukemias occur following an aggravation of a myeloproliferative syndrome, in particular by accumulating genetic anomalies inhibiting the differentiation of progenitors.

[0050] In an advantageous embodiment, the invention relates to the aforementioned method, where when the individual is likely to develop a myelodysplastic syndrome, it is concluded that

[0051] * if the intensities of a third group of peaks of the spectral signature of said individual are significantly different from the intensities of these same peaks obtained in the reference spectral signature(s) the individual is likely to develop a low-risk myelodysplastic syndrome

[0052] the third group of peaks corresponding to the wave numbers of the following first group: 3060cm ', 3062cm ', 3396cm ', 3384cm1 and 3052cm1,

[0053] *otherwise the individual is likely to develop high-risk myelodysplastic syndrome.

[0054] Using the first, second and third groups of peaks of the spectral signature, it is possible to discriminate the occurrence of a low-risk or high-risk myelodysplastic syndrome.

[0055] Advantageously, the invention relates to the method described above, where when the individual is likely to develop leukemia, it is concluded that:

[0056] * if the intensities of a fourth group of peaks of the spectral signature of said individual are significantly different from the intensities of these same peaks obtained in the reference spectral signature(s) the individual is likely to develop secondary leukemia,

[0057] the fourth group of peaks corresponding to the wave numbers of the following first group: 3270cm1, 3268cm1, 3266cm ', 3264cm1, 3192cm1, 3190cm ', 2850cm1, 2840cm1, 1707cm1, 1705cm1, 1664cm1, 1662cm1, 1633cm1, 1631cm1, 1493cm1, 1491cm1, 1489cm1, 1458cm1, 1456cmy and 1256 cm1,

[0058] * otherwise the individual will be likely to develop leukemia new.

[0059] Using the first, second and fourth groups of peaks of the spectral signature, it is possible to discriminate between the occurrence of de novo leukemia or secondary leukemia.

[0060] Even more advantageously, the invention relates to the aforementioned method, where said biological sample is a blood plasma sample.

[0061] The advantageous biological sample for implementing the invention is blood plasma which can be obtained during a routine blood test.

[0062] Blood plasma is the liquid fraction of blood. It constitutes approximately 55% of the blood volume and serves to transport blood cells, platelets and hormones and other soluble components (proteins, metabolites, hormones, salts, etc.) throughout the body.

[0063] Advantageously, the invention relates to the aforementioned method, where the spectral signature and the reference spectral signature(s) are obtained by the second derivative of the respective infrared spectroscopy data.

[0064] The calculation of the second derivative of each of the spectra is advantageously carried out. This second derivative improves the resolution of the infrared bands as well as the discrimination of the peaks obtained. The second derivation of the infrared spectra provides a clear improvement compared to the use of raw (non-derived) spectra for the characterization and identification of compounds contained in a sample.

[0065] These treatments are carried out with software generally integrated into the spectrometer.

[0066] The invention also relates to the use of a sample of blood plasma from an individual for determining, in particular in vitro, the risk of developing a hemopathy in said individual, where

[0067] * if the peak intensities corresponding to a first group of wave numbers of a spectral signature obtained, for said sample, by infrared spectroscopy are significantly different from the intensities of the same peaks obtained from the spectral signature of one or more control individuals, then the individual is likely to develop a hemopathy,

[0068] said first wave number group corresponding to the following wave numbers: 1330 cm1, 1445 cm1, 1478 cm1, 1493 cm1, 1505 cm1, 1507 cm1, 1520 cm1, 1526 cm ', 1544 cm1, 1571 cm1, 1602 cm1, 1668 cm1, 1674 cm1, 1676 cm1, 1697 cm1, and 2852 cm1,

[0069] * otherwise the individual is not likely to develop a blood disease.

[0070] Advantageously, the invention relates to the aforementioned use, where if the individual is likely to develop a hemopathy, and

[0071] * if the peak intensities corresponding to a second group of wave numbers of a spectral signature obtained, for said sample, by infrared spectroscopy are significantly different from the intensities of the same peaks obtained from the spectral signature of one or more control individuals, then the individual is likely to develop leukemia,

[0072] said second wave number group corresponding to the following wave numbers: 3316 cm1, 3283 cm1, 3281 cm1, 3256 cm1, 3118 cm1, 3116 cm1, 1345 cm1, 1343 cm-', 1340 cm1 and 1338 cm1,

[0073] * otherwise the individual is likely to develop myelodysplastic syndrome.

[0074] Even more advantageously, the invention relates to the aforementioned use, where

[0075] - if the individual is likely to develop myelodysplastic syndrome,

[0076] * if the peak intensities corresponding to a third group of wave numbers of a spectral signature obtained, for said sample, by infrared spectroscopy are significantly different from the intensities of the same peaks obtained from the spectral signature of one or more control individuals, the individual will be likely to develop a low-risk myelodysplastic syndrome,

[0077] said third wave number group corresponding to the following wave numbers: 3060 cm1, 3062 cm1, 3396 cm1, 3384 cm1 and 3052 cm1,

[0078] * otherwise the individual will be likely to develop myelodysplastic syndrome at high risk, and

[0079] - if the individual is likely to develop leukemia,

[0080] * if the peak intensities corresponding to a fourth group of wave numbers of a spectral signature obtained, for said sample, by infrared spectroscopy are significantly different from the intensities of the same peaks obtained from the spectral signature of one or more control individuals, the individual will be likely to develop secondary leukemia,

[0081] said fourth wave number group corresponding to the following wave numbers: 3270 cm1, 3268 cm1, 3266 cm1, 3264 cm1, 3192 cm1, 3190 cm1, 2850 cm1, 2840 cm1, 1707 cm1, 1705 cm1, 1664 cm1, 1662 cm1, 1633 cm1, 1631 cm1, 1493 cm *, 1491 cm1, 1489 cm1, 1458 cm1, 1456 cm1 and 1256 cm1,

[0082] * otherwise the individual will be likely to develop leukemia new.

[0083] The invention further relates to a computer program product, or software, comprising portions, means or instructions of program code for executing the steps of the method as defined above, when said program is executed on a computer.

[0084] Advantageously, said program is included in a computer-readable data recording medium. Such a medium is not limited to a portable recording medium such as a CD-ROM but may also be part of a device comprising an internal memory in a computer (for example RAM and / or ROM), or of an external memory device such as hard disks or USB keys, or a nearby or remote server.

[0085] Advantageously, the aforementioned computer program or software product is designed to allow,

[0086] - processing of infrared spectra in order to obtain the spectral signature,

[0087] - the comparison of the spectral signature obtained from the biological sample of the individual tested with the reference spectral signatures,

[0088] - or both.

[0089] The aforementioned computer program or software product can also advantageously be used to perform a second derivative of the spectra obtained after the Fourier transform.

[0090] The invention will be better understood in light of the following figure and examples. Brief description of the figures

[0091] [fig. 1] [fig.l] represents a table showing the synthetic results of the intergroup discrimination G0, Gl, G2, G3 and G4. G0: Healthy; G1: Low-risk myelodysplastic syndromes; G2: High-risk myelodysplastic syndromes; G3: Secondary acute myeloid leukemia and G4: De novo acute myeloid leukemia. Examples

[0092] Example 1: classification

[0093] Materials and Methods

[0094] A- Biological samples

[0095] Blood plasmas were isolated by double centrifugation (700g, 10 min) from whole blood from patients (n = 70) included in the MYLESYM study (ID—RCE 2011-A00271-40) who gave their consent. They were compared to the plasmas of 99 healthy donors recruited as part of the HEALTHOX study (ClinicalTrials.gov # NCT02789839).

[0096] The plasma samples (50 pL) are frozen at -80°C until use. Once thawed at room temperature and homogenized using a vortex mixer, 5 pL are deposited and spread on a 96-position multi-well plate made of silica or zinc selenide (ZnSe), materials transparent in the IR, left to dry for 15 min in an oven at 35°C and analyzed using an MIR spectrophotometer.

[0097] Alternatively, the samples (20 pL) can be placed on a microscope slide and left in the open air to dry for 24 hours.

[0098] B- Acquisition of blood plasma samples

[0099] Samples n= 169 (99 healthy controls / 70 patients)

[0100] G0: Healthy (60 Women, 39 Men),

[0101] G1: Low-risk myelodysplastic syndromes: LR-MDS (12 Women, 26 Men),

[0102] G2: High-risk myelodysplastic syndromes: MDS-HR (5 Women, 4 Men),

[0103] G3: Secondary acute myeloid leukemia AML Sec (3 Women, 8 Men),

[0104] G4: De novo acute myeloid leukemia: AML-Novo (9 Women, 7 Men).

[0105] C- LUMOS microscope (Bruker)

[0106] The LUMOS is a stand-alone IRTF microscope equipped with an integrated spectrometer. The innovation provided by a motorized crystal allows the system to switch from Transmission mode to Reflection and ATR mode without any operator intervention and to measure, in a fully automated manner, a sample or background noise, even when the ATR mode is activated. This type of device is suitable for measurements in attenuated total reflection (ATR) if the samples have been deposited on glass slides, a material which is not transparent in the mid-infrared.

[0107] An analogue of Bruker's IR Biotyper can also be used. The instrument is controlled using Bruker's proprietary software, OPUS. This type of spectrophotometer makes it easy to acquire around a hundred spectra per day, including plate preparation. The infrared spectra are therefore collected in "Transmission" mode, with the infrared beam passing through the sample and the multi-well plate, which is made of crystallized ZnSe, a material transparent in the mid-infrared range.

[0108] In all cases (reflection or transmission measurements) the spectral resolution is 4 cm-1 and 64 to 128 scans are averaged. The background noise is measured through an empty well. The "raw" absorption spectra (as is) are then saved and exported in Jcamp format ("open" format) using a macro routine under OPUS.

[0109] D- Quality test

[0110] In order to evaluate the quality of the spectra according to several parameters: Water vapor, Signal / water ratio, Noise intensity... and identify aberrant spectra (outliers) which do not respect certain criteria. To check the hydration state of the sample, we ensure that the Amide 1 band of the proteins (1650 cm *) is 2 or 3 times larger than the band of (3400 cm1) which essentially reflects liquid water. [YES] E- Baseline Correction

[0112] The variation of the baseline can be caused by a change in the conditions during acquisition or variations related to the instrumentation or the environment (example: temperature).

[0113] F-Normalization

[0114] In order to minimize signal intensity differences that are not related to the sample but to the instrumentation, the raw spectra are normalized by an anti-scattering algorithm MSC (Multiplicative Scattering Correction): this is a spectral correction method (Sun. D-Wet al. 2009).

[0115] G-Filtration

[0116] This treatment consists of choosing the spectral range of interest depending on the sample. On the sample of interest (plasma), a spectral range of 3800 to 940 cm1 is fixed. On the band 2800 to 1800 cm1, the spectrum is truncated because it does not contain any information of interest for the analyses carried out; we mainly find the contribution of atmospheric CO2 which reflects environmental variations.

[0117] H- Second derivatives

[0118] The derivation makes it possible to improve the resolution of the spectra and therefore to limit the effects of band overlap. It should be noted that the transition from the raw spectrum to the second derivative reduces the signal / noise ratio {Martens H et al. 2002}. The second derivatives of the spectra are calculated using 13 points for Savitsky-Golay smoothing by sliding window.

[0119] Spectral data analysis

[0120] A- Statistical methods

[0121] 1. Unsupervised analysis (descriptive analysis)

[0122] PCA: Principal Component Analysis: This is an analysis carried out as a first step, allowing us to understand the structure of the data and to identify possible so-called outlier spectra which present a different spectral profile for technical reasons, poor acquisition for example, or for biochemical reasons.

[0123] 2. Supervised analysis (explanatory analysis)

[0124] R—PLS: Regression by the method of least squares (R-Partial Least Squares): It is a statistical method that allows the modeling of complex relationships between observed quantitative variables, called manifest, and latent variables (MIR spectrum)

[0125] B- Selection of variables

[0126] The selection of variables by genetic algorithm or FADA method makes it possible to identify a subset of discriminating variables to specify the types of biochemical markers modified by the pathology (Trevisan J et al. 2014). It has two advantages:

[0127] o Improve the model's performance in prediction (Jouan-Rimbaud D et al. 1995).

[0128] o Improve the interpretation of models and understand the system studied.

[0129] C- FADA and GLM algorithms

[0130] It is an analysis by LDA / Logistic Regression which makes it possible to identify the most discriminating spectral variables here between the healthy and the different groups of patients. From these most discriminating variables, a progressive selection is carried out to identify the few variables which make it possible to have the best specificities and selectivities.

[0131] D- Prediction principle

[0132] The results of the discriminant analysis tests are conventionally represented in the form of a confusion matrix to be interpreted as shown below in Table 1.

[0133] [Tables 1] Measured (diagnosis) Well classified (%) Positive Negative Prediction IR Positive VP FP VPP = VP / (VP+FP) Negative FN VN VPN = VN / (VN+FN) Se = VP / (VP + FN) Sp = VN / (FP + VN)

[0134] The numbers in bold represent the WELL classified samples, and the underlined ones represent the POOR classified ones.

[0135] VPP and VPN present the classification success percentage

[0136] VP: True positive, VN: True negative, FN: False negative, FP: False positive

[0137] PPV: positive predictive value, NPV: negative predictive value

[0138] Se: Sensitivity, Sp: Specificity.

[0139] The results obtained for this study are shown in the following Table 2 and in

[0140] [Tables 2] TEST Fl n 169 Discriminant variables (cm 1 ) Discriminated groups Measured Se Sp Well classified (%) 1330, 1445, 1478, 1493, 1505, 1507, 1520, 1526, 1544, 1571, 1602, 1668, 1674, 1676, 1697, 2852 predicted GO 3197 6 0.99 0.99 99 [G1-G4] 3 2394 99 F2 137 1330, 1478,1520, 1668,1697, 2852 predicted GO 1611 723 0.80 0.79 69 G1 389 2877 88 F3 153 1330, 1445, 1493, 1505, 1507, 1520, 1526, 1571, 1666,1668, 1674 predicted GO 3242 94 0.98 0.94 97 [G1-G3] 58 1706 97 F4 115 1054, 1056, 1122, 1124, 1493, 1520, 1571, 1602, 1666,1668, 1674 predicted GO 3189 39 0.99 0.93 99 G4 11 561 98 F5 23 3270, 3268, 3266, 3264,3192, 3190, 2850, 2840, 1707, 1705, 1664, 1662, 1633, 1631, 1493, 1491, 1489, 1458, 1456, 1256 predicted G3 128 292 0.32 0.70 30 G4 272 708 72 F6 47 3060, 3062, 3396, 3384, 3052 predicted G1 1323 239 0.95 0.20 84 G2 68 61 47 F7 16 1705, 1182, 1174, 1060, 1058, 1056 predicted G2 191 186 0.63 0.38 51 G3 109 114 51 F8 146 3339, 3384, 3062, 3060, 3052 predicted G0 3194 156 0.96 0.90 95 [G1+G2 ] 106 1444 93 F9 122 1668, 1666, 1526,1507, 1505, 1493 predicted G0 3175 238 0.96 0.70 93 [G3+G4 ] 125 562 82 F10 70 3316, 3283, 3281,3256, 3118,3116, 1345, 1343, 1340, 1338 predicted [G1+G2 ] 1376 364 0.86 0.55 79 [G3+G4 ] 224 463 66

[0141] This study establishes that myelodysplastic syndromes and acute leukemias (de novo or secondary) are accompanied by distinct metabolic changes that are revealed through specific IVIIR spectral signatures (specific “barcodes”). This opens up interesting possibilities in terms of early and rapid diagnosis for:

[0142] o The identification of plasma molecules that could reflect the pathophysiology and be interesting biomarkers (interpretation of spectral signatures);

[0143] o An aid in the early detection of myelodysplasia; and

[0144] o Help for monitoring SMD patients.

[0145] Example 2

[0146] The files imported under OPUS are then imported and transposed into a matrix using software written in the R environment:

[0147] ################################################################

[0148] # ImportJcamp.R

[0149] # Directory (Jcamp files and saving the Excel file)

[0150] rm(list=ls()) # clears all memory

[0151] activeRep <- "- / Documents / Rcore / Rdata / Cancer / "

[0152] setwd(activeRep)

[0153] #Name of the exported Excel file?

[0154] excelout <- "mad.xlsx"

[0155] #Number of expected attributes

[0156] N <-2

[0157] ################################################

[0158] #Selection of files of interest

[0159] list_files <- list.files(activeRep, pattem=c("dx","JDX"), full.names=TRUE)

[0160] #liste_files <- list.files(activeRep, full.names=TRUE)

[0161] #Number of files

[0162] n <- length(file_list)

[0163] #Library read JDX

[0164] library(readJDX)

[0165] #Wave number on a data.frame

[0166] NODF <- readJDX(file_list[l])

[0167] NODF <- NODF[[4]]

[0168] FCE <- data.frame(NODF$x)

[0169] #Adding empty columns for attributes

[0170] ATTRIBUTES <- data.frame(c(l:N))

[0171] for(iinl:n){

[0172] A<-c(l:N)

[0173] ATTRIBUTES <- data.frame(cbind(ATTRIBUTES,A))

[0174] }

[0175] ATTRIBUTES[c(0:N),c(0:n+l)] <- ""

[0176] #We name the column (on data.frame attributes to simplify after cbind)

[0177] names(ATTRIBUTS)[l] <- "Sample Name"

[0178] #Adding the rest of the other columns (and we name them here too on the data.frame)

[0179] for(iinl:n){

[0180] ACDF <- readJDX(file_list[i])

[0181] ACDF <- ACDF[[4]]

[0182] FCE <- cbind(FCE,ACDF$y)

[0183] names(ATTRIBUTES)[i+l] <- basename(unlist(strsplit(file_list[i],c(".JDX",".dx"))[l]))

[0184] }

[0185] #Inversion of rows and columns (=transposition)

[0186] FCE <- t(FCE)

[0187] ATTRIBUTES <-1(ATTRIBUTES)

[0188] #We bring everything together

[0189] FCE <- cbind(ATTRIBUTES,FCE)

[0190] #Export in Excel format

[0191] library(openxlsx)

[0192] wbl <- createWorkbook()

[0193] sheet <- addWorksheet(wbl,sheetName = "Raw Data")

[0194] writeData(wbl, sheet = "Raw Data",FCE,colNames = FALSE,rowNames = TRUE)

[0195] saveWorkbook(wbl, excelout)

[0196] At the end of this program, an Excel file is created whose first tab contains the transposed matrix (1 sample = One line) of all the samples to be processed.

[0197] The next step is to calculate the second derivatives of each spectrum, smooth these derivatives using the Savitky and Golay sliding window routine over 11 or 13 points and then truncate these derivatives to keep only the frequency domains relevant for the analysis. The spectral domains retained are, in almost all cases, 3800-2800 cm-1 and 1800-700 cm-1. They are then normalized according to the principle of vector normalization (the area of ​​the second derivative is normalized to 1). The matrix of second derivatives is saved in a second tab of the same Excel file. The script below performs these preprocessings.

[0198] #####################################################

[0199] # Pn / ©treatment.R

[0200] #Directory (Jcamp files and saving the Excel file)

[0201] activeRep <- "- / Documents / Rcore / Rdata / CANCER"

[0202] setwd(activeRep)

[0203] #Number of attributes

[0204] N <-2

[0205]

[0206]

[0207]

[0208]

[0209]

[0210]

[0211]

[0212]

[0213]

[0214] #Troncature (T1>T2>T3>T4) "Tl" <- 3600 "T2" <- 2800 "T3" <- 1800 "T4" <- 700 #Taille de la fenetre glissante de lissage W<-11 #Nom du fichier excelout <- "Mad.xlsx" / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / #############

[0215]

[0216]

[0217]

[0218]

[0219]

[0220]

[0221]

[0222]

[0223]

[0224]

[0225]

[0226]

[0227]

[0228]

[0229]

[0230]

[0231]

[0232]

[0233]

[0234]

[0235]

[0236]

[0237] #Packages library(RcppArmadillo) library(prospectr) library(tidy verse) library(dplyr) library(ppls) library(openxlsx) library(hyperSpec) library(plotrix) #Import du fichier Import <- read.xlsx(excelout,colNames = FALSE) write.csv(Import,'Import.csv') NF <- read.csv('Import.csv') file.remove('Import.csv') NF <- NF[,-1] #Tableau d'attributs (Nombre de columns dépendant du nombre d'attributs) MAT.ATTRIBUTS <- data.frame(NF[,l:(N+l)]) #Tableau des données MAT.ABS <- data.frame(NF[,-c(l:(N+l))]) #Derivee seconde et lissage P<- (wl) / 2 LMA <- length(MAT.ABS) MAT.D2 <- rbind(MAT.ABS[l,-c(l:P,(LMA-P+l):LMA)],savitzkyGolay(MAT.ABS[-l,],p=3,w,m =2))

[0238]

[0239] #Normalisation de D2 MAT.D2 <- rbind(MAT.D2[l,],normalize.vector(MAT.D2[-l,]))

[0240] #Truncation D2N

[0241] if(Tl > max(MAT.D2[l,])){ .

[0242] RT1 <- 1

[0243] }else{

[0244] RT1 <- which(MAT.D2[l,]<=Tl)

[0245] RT1 <- min(RTl)

[0246] }

[0247] RT2 <- which(MAT.D2[l,]<=T2)

[0248] RT2 <- min(RT2)

[0249] RT3 <- which(MAT.D2[l,]<=T3)

[0250] RT3 <- min(RT3)

[0251] if(T4 < min(MAT.D2[l,])){

[0252] RT4 <- length(MAT.D2[l,])

[0253] }else{

[0254] RT4 <- which(MAT.D2<=T4)

[0255] RT4 <- min(RT4)

[0256] }

[0257] MAT.D2NT <- cbind(MAT.D2[,c(RTl:RT2,RT3:RT4)])

[0258] #Final tables:

[0259] #Table Derived second, normalized and truncated

[0260] D2NT <- cbind(MAT.ATTRIBUTES,MAT.D2NT)

[0261] #Displaying graphics

[0262] pdf(file = "Plot of preprocessing steps.pdf")

[0263] par(mfrow=c(2,l))

[0264] Y1 <- MAT.ABS

[0265] Y1<-Y1[-1,]

[0266] Y1 <- as.matrix(Yl)

[0267] XI <- as.vector(as.numeric(MAT.ABS[l,]))

[0268] RawSpectrum <- new("hyperSpec",spc=Yl,wavelength = XI)

[0269] plotspc(SpectreBrut,wl.reverse = TRUE,wl.range = c(T4 ~ T3,T2 ~ Tl),xoffset = 990,

[0270] title.args = list (xlab = expression("Wavenumber " (cmA-l)), main="Raw data"))

[0271] Y2<-MAT.D2NT

[0272] Y2<-Y2[-l,]

[0273] Y2 <- as.matrix(Y2)

[0274] X2 <- as.vector(as.numeric(MAT.D2NT[l,]))

[0275] SpectrumD2 <- new("hyperSpec",spc=Y2,wavelength = X2)

[0276] plotspc(SpectreD2,wl.range = c(T4 ~ T3,T2 ~ Tl),xoffset = 990,wl.reverse = TRUE,

[0277] title.args = list(xlab = express("Wavenumber " (cmA-l)),main="Truncated normalized derivatives"))

[0278] #Export the plot in PNG

[0279] dev.off()

[0280] #Export in Excel format

[0281] wb <-createWorkbook()

[0282] RD <- "RawData" # raw data tab

[0283] secD <- "D2Norm" # tab derivatives normalized seconds

[0284] sheet <- addWorksheet(wb,sheetName = RD)

[0285] sheet2 <- addWorksheet(wb,sheetName = secD)

[0286] writeData(wb,sheet = RD,NF,colNames = F,rowNames = F)

[0287] writeData(wb,sheet = secD,D2NT,colNames = F,rowNames = F)

[0288] saveWorkbook(wb,excelout,overwrite = T)

[0289] #Export D2NT to .csv format for PCA

[0290] colnames(D2NT) <- D2NT[1,]

[0291] write.csv(D2NT,"D2NT.csv")

[0292] Please note: these steps can be performed with any calculation software. The mathematical operations are standard. However, it is important to respect the order in which they are performed.

[0293] Some authors prefer to work from raw spectra corrected for scattering (Multiple Scattering Correction or MSC routine). The inventors have observed better performance by working from second derivatives.

[0294] These second, truncated and normalized derivatives are used for the calibration of predictive models.

[0295] The predictive model is based on an R-PLS (Least Squares Regression) type analysis which makes it possible to identify the most discriminating spectral variables between two groups. These variables are ordered according to the number of times they have been positively selected during a large number of iterations (usually 100). Manual tests are then carried out to best reduce the variables which will have to be taken into account in the predictive model. Each time (for each combination of variables) a confusion matrix is ​​calculated which makes it possible to identify the well and poorly classified samples.

[0296] Once this optimization is done, a validation is carried out by the prediction of samples not used for the calibration of the predictive model. The R script below allows these tasks to be carried out.

[0297] # GA_PLSmultipleRuns.R

[0298] rm(list=ls(all=TRUE))

[0299] setwd("~ / Documents / Rdata / CANCEROPOLE / ") # A RENSEIGNER

[0300] experidata <- read.csv("Evolution2D.csv",sep=";",dec=".",header=TRUE)

[0301] savefile <- "CalEvolution2DcumulVarPLS.txt"

[0302] selindiv <- "CalEvolution2D_Xsomes.txt"

[0303] #head(data)

[0304] library('MASS')

[0305] library('class')

[0306] library('GA')

[0307] library('glm2')

[0308] library(xlsx)

[0309] library('pls')

[0310] library('ChemometricsWithR') [0311 ] library('ChemometricsWithRData')

[0312] source("~ / Documents / Rpackages / routines / init_pop.R")

[0313] source("~ / Documents / Rpackages / routines / pls_fitness_pop.R")

[0314] source("~ / Documents / Rpackages / routines / select_pop.R")

[0315] source("~ / Documents / Rpackages / routines / crossover_pop.R")

[0316] numcol <- dim(experidated)

[0317] maxcol <- numcol[2]

[0318] identity <- experidata[, 1 ]

[0319] VarCible <- experidata[,2]

[0320] #TNR90brut <- experidata[,3]

[0321] #TNR30brut <- experidata[,4]

[0322] #index30 <- expired[,5]

[0323] #index90 <- expired[,6]

[0324] #TNRejaculat <- experidata[,7]

[0325] MIR <- experidata[,3:maxcol] # spectra begin at COL3

[0326] true <- dim(experidata)

[0327] spectrumsizelOpc <- trunc(0.1*tutu[l])

[0328] # Data without the target variable: rawData

[0329] rawData <- MIR

[0330] X <- rawData

[0331] XI<-X

[0332] XS <- X

[0333] # cNames : Variable names

[0334] cln <- colnames(X)

[0335] n <- length(cln)

[0336] cNames <- t(cbind(l:n,colnames(X)))

[0337]

[0338] # creation data.frame with choice of index #plsdata <- data.frame(vary = tnrEjaculat, varx = I(as.matrix(MIR)), row.names = identity)

[0339]

[0340]

[0341]

[0342]

[0343]

[0344]

[0345]

[0346]

[0347]

[0348]

[0349]

[0350]

[0351]

[0352]

[0353]

[0354]

[0355]

[0356]

[0357]

[0358]

[0359]

[0360]

[0361]

[0362]

[0363]

[0364]

[0365]

[0366]

[0367]

[0368]

[0369]

[0370]

[0371]

[0372]

[0373] #attach(plsdata) # R-PLS results plots #(RMSEP(apls), legendpos = 'topright') #plot(apls, ncomp=3, asp=l, line=TRUE) #Sys.sleep(3) # plot(apls, ncomp = 1, asp = 1, line = TRUE) # plot(apls, 'loadings', comps = 1, legendpos = 'topleft', xlab = 'wavenumbers') # abline(h = 0) # Selection stores up to 100 selections (or runs) selection <- matrix(0,n,l) # Y : Target variable Y <- TargetVar # PLOT Average spectrum frame() ymin <- min(XI) ymax <- max(XI) meanS <- apply(XI,2, mean) plot(l:n, meanS, type = "1", Iwd = 3, col = "blue", ylim = c(ymin, ymax), xlab = "Discriminant variables",ylab = "Spectre Moyen") # * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * number_of_runs = 3 # remettre à 100 number_of_passes = 7 max_iterations = 10 # remettre à 50 pop_size =100 for (nrun in l:number_of_runs) { cat("nrun = ",nrun,"\n") for (pass in l:number_of_passes) { cat("pass : ",pass,"\n") pop = list(pop = NULL, size = pop_size, bits = ncol(X), fitness = NULL, P = 0) pop$P = pop$size pop$pop = matrix(0, nrow = pop$size, ncol = pop$bits) pop$pop = init_pop(pop$size, pop$bits) pop$fitness = pls_fitness_pop(pop$pop, X, Y) ,

[0374] last_idx = length(pop$fitness)

[0375] idx = order(pop$fitness)

[0376] pop$pop = pop$pop[idx,]

[0377] pop$fitness = pop$fitness[idx]

[0378] worst_fitness = pop$fitness[last_idx]

[0379] for (i in l:max_iterations)

[0380] {

[0381] #cat(" i = ",i,"\n")

[0382] idx_parents = select_pop(pop$P)

[0383] offspring_pop = crossover_pop(pop$pop, idx_parents, pop$bits)

[0384] offspring_fitness = pls_fitness_pop(offspring_pop, X, Y)

[0385] if (offspring_fitness[l] < worst_fitness)

[0386] {

[0387] pop$pop[last_idx,] = offspring_pop[l,]

[0388] pop$fitness[last_idx] = offspring_fitness[l]

[0389] idx = order(pop$fitness)

[0390] pop$pop = pop$pop[idx,]

[0391] pop$fitness = pop$fitness[idx]

[0392] worst_fitness = pop$fitness[last_idx]

[0393] }

[0394] if (offspring_fitness[2] < worst_fitness)

[0395] {

[0396] pop$pop[last_idx,] = offspring_pop[2,]

[0397] pop$fitness[last_idx] = offspring_fitness[2]

[0398] idx = order(pop$fitness)

[0399] pop$pop = pop$pop[idx,]

[0400] pop$fitness = pop$fitness[idx]

[0401] worst_fitness = pop$fitness[last_idx]

[0402] }

[0403] cat("Pass ", pass," of ", number_of_passes," itération ",i," of ",max_iterations, best= ",pop$fitness[l],"\n")

[0404] #flush.console()

[0405] }

[0406] idx = which(pop$pop[l,] != 0)

[0407] if (pass == 1)

[0408] {

[0409] best_ever_variables = cNames[,idx]

[0410] best_ever_fitness = pop$fitness[l]

[0411] } else {

[0412] if (pop$fitness[l] <= best_ever_fitness)

[0413] {

[0414] best_ever_variables = cNames[,idx]

[0415] best_ever_fitness = pop$fitness[l]

[0416] }

[0417] }

[0418] X = X[,idx]

[0419] cNames = cNames[,idx]

[0420] }

[0421] best_ever_variables = t(best_ever_variables)

[0422] colnames(best_ever_variables) = c("Index", "Name")

[0423] best_ever_variables

[0424] best_ever_fitness

[0425] variables <- colnames(MIR)

[0426] ind <- rep(O,length(variables))

[0427] for (i in l:length(best_ever_variables[,2])) {

[0428] id<- which(best_ever_variables[i,2]==variables)

[0429] ind[id] <- 1

[0430] }

[0431] ligne <- c(best_ever_fitness,trunc(ind))

[0432] white <- matrix(0,nrow=n,neol=l)

[0433] # Tracé des var discri sur le graphe *****************

[0434] w <- dim(best_ever_variables)

[0435] bw <- w[l]

[0436] for (bs in l:bw)

[0437] {

[0438] sel <- as.double(best_ever_variables[bs,l])

[0439] selection[sel,l] <- selection[sel,l]+1.0000

[0440] white[sel,l] <- 1

[0441] }

[0442] gfactor <- (ymax-ymin) / max(selection)

[0443] plot.new()

[0444] #pdf(filename="TempHisto.pdf", 20, 20)

[0445] plot(l:n, meanS, type = "1", Iwd = 3, col = "blue", ylim = c(ymin, ymax),

[0446] xlab = "Discriminant Variables", ylab = "Average Spectrum")

[0447] by(bg=NA)

[0448] for (bev in l:n)

[0449] {

[0450] y2 <- ymin + selection[bev] * gfactor

[0451] segments (bev, ymin, bev, y2, col = "red")

[0452] }

[0453] #dev.off()

[0454] # Update selection frequencies in savefile (file overwritten on each run)

[0455] cat("nrun = ",nrun,"\n",file = savefile, append=FALSE)

[0456] cat("best fit = ",best_ever_fitness,"\n",file = savefile, append=TRUE)

[0457] for(sfinl:n)

[0458] {

[0459] cat(substring(cln[sf],2),"\t", selection[sf], file = savefile, fill = TRUE, append=TRUE)

[0460] }

[0461] # Update your individual selections in selindiv

[0462] if (nrun==l) {

[0463] cat("Best RMSEP", substring(cln,2), "\n", file = selindiv, fill = FALSE, sep = "\t", append = FALSE)

[0464] }

[0465] cat(line,"\n", file = selindiv, fill = FALSE, sep = "\t", append=TRUE)

[0466] X <- XS

[0467] cln <- colnames(X)

[0468] n <- length(cln)

[0469] cNames <- t(cbind(l:n,colnames(X)))

[0470] }

[0471] #savePlot(filename = "TempHistoPlot", type = c("pdf")

[0472] The results are presented in the form of second derivative spectrum, identification of markers (discriminating variables) and confusion matrix as identified in [fig.l].

[0473] The invention is not limited to the embodiments presented and other embodiments will become clear to those skilled in the art.

Claims

Claims

1. Method for determining, in vitro, the risk for an individual of developing a malignant hemopathy from a biological sample of blood, or a by-product of blood, or bone marrow, from this individual, said method comprising the following steps: - exposing said biological sample to mid-infrared radiation (MIR) of wavelength varying from 4000 cm 1 to 400 cm 1 to obtain a spectrum characteristic of said sample, said spectrum being processed in order to obtain a spectral signature consisting of absorbance peaks characteristic by their position, or wave number, and their intensity, or absorbance, of the nature and relative concentrations of the different molecules present in said sample;- comparing said spectral signature obtained in the previous step with one or more reference spectral signatures, said one or more reference spectral signatures being obtained from a population of reference individuals from a biological sample of the same nature; and - concluding that; * if the intensities of a first group of peaks of the spectral signature of said individual are significantly different from the intensities of these same peaks obtained in the reference spectral signature(s), then the individual is likely to develop a hemopathy, the first group of peaks corresponding to the wave numbers of the following first group: 1330 cm1, 1445 cm1, 1478 cm ', 1493 cm1, 1505 cm1, 1507 cm1, 1520 cm1, 1526 cm1, 1544 cm1, 1571 cm1, 1602 cm1, 1668 cm1, 1674 cm1, 1676 cm1, 1697 cm1, and 2852 cm1, * otherwise the individual is not likely to develop a blood disease.

2. A method according to claim 1, wherein, when the individual is susceptible to developing a blood disease, it is further concluded that: * if the intensities of a second group of peaks of the spectral signature of said individual are significantly different from the intensities of these same peaks obtained in the reference spectral signature(s), then the individual is likely to develop leukemia, the second group of peaks corresponding to the wave numbers of the following first group: 3316 cm1, 3283 cm1, 3281 cm1, 3256 cm1, 3118 cm1, 3116 cm1, 1345 cm1, 1343 cm1, 1340 cm1 and 1338 cm1, * otherwise the individual is likely to develop myelodysplastic syndrome. plastic.

3. Method according to claim 2, wherein when the individual is likely to develop a myelodysplastic syndrome, it is concluded that * if the intensities of a third group of peaks of the spectral signature of said individual are significantly different from the intensities of these same peaks obtained in the reference spectral signature(s) the individual is likely to develop a low-risk myelodysplastic syndrome, the third group of peaks corresponding to the wave numbers of the following first group: 3060cm ', 3062cm 3396cm 3384cm1 and 3052cm1, * otherwise the individual is likely to develop a high-risk myelodysplastic syndrome.

4. Method according to claim 2, wherein when the individual is likely to develop leukemia, it is concluded that: * if the intensities of a fourth group of peaks of the spectral signature of said individual are significantly different from the intensities of these same peaks obtained in the reference spectral signature(s) the individual is likely to develop secondary leukemia, the fourth group of peaks corresponding to the wave numbers of the following first group: 3270cm1, 3268cm1, 3266cm ', 3264cm1, 3192cm1, 3190cm1, 2850cm1, 2840cm1, 1707cm1, 1705cm1, 1664cm1 , 1662cm1, 1633cm1, 1631cm1, 1493cm1, 1491cm1, 1489cm1, 1458cm', 1456cmy and 1256 cm1, * otherwise the individual will be likely to develop de novo leukemia.

5. A method according to any one of claims 1 to 4, wherein said biological sample is a blood plasma sample.

6. A method according to any one of claims 1 to 5, wherein the spectrum and the control spectrum are obtained by the second derivative of the infrared spectroscopy data.

7. Use of a blood plasma sample from an individual for the determination, in vitro, of the risk of developing a haemopathy of said individual, where * if the intensities of peaks corresponding to a first group of wave numbers of a spectral signature obtained, for said sample, by infrared spectroscopy are significantly different from the intensities of the same peaks obtained from the spectral signature of one or several witness individuals, then the individual is likely to develop a blood disease, said first wave number group corresponding to the following wave numbers: 1330 cm1, 1445 cm1, 1478 cm1, 1493 cm1, 1505 cm1, 1507 cm1, 1520 cm1, 1526 cm1, 1544 cm1, 1571 cm1, 1602 cm1, 1668 cm1, 1674 cm1, 1676 cm1, 1697 cm1, and 2852 cm1, * otherwise the individual is not likely to develop a blood disease.

8. Use of a sample according to claim 7, or if the individual is likely to develop a blood disease, and * if the peak intensities corresponding to a second group of wave numbers of a spectral signature obtained, for said sample, by infrared spectroscopy are significantly different from the intensities of the same peaks obtained from the spectral signature of one or more control individuals, then the individual is likely to develop leukemia, said second group of wave numbers corresponding to the following wave numbers: 3316 cm1, 3283 cm1, 3281 cm1, 3256 cm1, 3118 cm1, 3116 cm1, 1345 cm1, 1343 cm1, 1340 cm1 and 1338 cm1, * otherwise the individual is likely to develop myelodysplastic syndrome.

9. Use of a sample according to claim 8, wherein - if the individual is likely to develop myelodysplastic syndrome, * if the peak intensities corresponding to a third group of wave numbers of a spectral signature obtained, for said sample, by infrared spectroscopy are significantly different from the intensities of the same peaks obtained from the spectral signature of one or more control individuals, the individual will be likely to develop a low-risk myelodysplastic syndrome, said third group of wave numbers corresponding to the following wave numbers: 3060 cm1, 3062 cm1, 3396 cm1, 3384 cm1 and 3052 cm -i * otherwise the individual will be likely to develop high-risk myelodysplastic syndrome, and - if the individual is likely to develop leukemia, * if the peak intensities corresponding to a fourth group of wave numbers of a spectral signature obtained, for said sample, by infrared spectroscopy are significantly different different intensities of the same peaks obtained from the spectral signature of one or more control individuals, the individual will be likely to develop secondary leukemia, said fourth wave number group corresponding to the following wave numbers: 3270 cm1, 3268 cm1, 3266 cm ', 3264 cm ', 3192 cm1, 3190 cm ', 2850 cm ', 2840 cm ', 1707 cm ', 1705 cm ', 1664 cm ', 1662 cm ', 1633 cm ', 1631 cm ', 1493 cm ', 1491 cm ', 1489 cm ', 1458 cm ', 1456 cm1 and 1256 cm ', * otherwise the individual will be likely to develop de novo leukemia.

10. A computer program product comprising portions, means or instructions of program code for carrying out the steps of the method according to any one of claims 1 to 6 when said program is executed on a computer.