RU2798738C1 - Method and system for identifying microorganisms - Google Patents

Abstract

FIELD: measurement technology.

SUBSTANCE: invention is related to a method and a system for identifying microorganisms in a sample by assessing the vibrational spectrum. The method comprises the steps of preparing a database of known spectra and obtaining pre-calculated models from known spectra, the pre-calculated models being associated with hierarchically structured categories and/or classification groups of microorganisms. Spectra of unknown microbial samples are compared with pre-calculated microbial identification models of progressively smaller categories and/or classification groups to provide scores indicating that the unknown sample belongs to the specified microorganism categories and/or classification groups.

EFFECT: increasing the accuracy and reducing the analysis time without using significant computing power.

11 cl, 7 dwg

Description

The field of technology to which the invention belongs

The present invention relates to a method for the identification of microorganisms applicable in medicine, clinical research, veterinary medicine, agriculture, food industry and environmental protection, which uses statistical methods, such as, for example, methods of multivariable analysis of spectral profiles or, alternatively, neural networks, to identify and obtain information regarding the growth of microorganisms in an extremely short time compared to conventional methods.

In particular, the invention provides an analytical tool for identifying microorganisms by analyzing the spectral profile of the vibrational spectrum of an unknown sample and comparing the spectral profile with the spectra of samples previously collected and stored in a database.

State of the art

The use of vibrational spectroscopy techniques for the identification of microorganisms is known in the art.

For example, Fourier transform infrared spectroscopy (FTIR) is a non-destructive analytical technology that provides information on the chemical composition of the analyzed sample. This technology has been used to analyze biological samples since the early 1990s (Diem M. et al., The Analyst [August 27, 2004, 129 (10): 880-885]).

At the same time, the ability of FTIR technology to identify and classify unknown microorganisms was shown (Helm D. et al., Journal of General Microbiology (1991, 137, 69-79) and Marley L. et al., Vibrational Spectroscopy 26.2 , 2001, 151-159).

It is known that various species, subspecies or subclassifications of microorganisms are characterized by a precise biochemical composition with respect to proteins, fats, nucleic acids and polysaccharides, which is reflected in a characteristic vibrational spectrum (ed. Griffiths and Chalmers, 2001 Handbook of vibrational spectroscopy, John Wiley & sons, New York, Volume 5).

Some potential applications of FTIR technology in microbiology include:

(i) identification of pathogens in the medical or veterinary field, for example in a clinical laboratory;

(ii) epidemiological studies, case studies, pathogen screening, hygiene testing, infection chain identification, treatment monitoring and detection of recurrent infections;

(iii) characterization and screening of environmental microorganisms;

(iv) monitoring of biotechnological processes;

(v) microbiological quality control in the food or pharmaceutical industry;

(vi) conservation of collected strains.

Known methods for identifying microorganisms based on the technology of comparative analysis of the spectrum of an unknown sample and the spectra of known types of microorganisms stored in the database.

Examples of known methods for identifying this type are given in documents US5660998A, CN103217398A, US6379920B1, WO2006002537A1, US9551654B2, US20170167973A1, WO2017/210783A1, Sousa et al., European Journal of Clinical Microbiology & Infect ious Diseases (2014) 33: 1345-1353, Whittaker et al. ., Journal of Microbiological Methods 55 (2003) 709-716, Wang et al., International Journal of Food Microbiology 167 (2013) 293-302.

Another method is described in patent application PCT/IT2019/050025 in the name of the applicant.

However, in many known methods, spectrum acquisition is performed in transmission, reflection, or imaging modes.

For example, spectrum acquisition in transmission or reflection mode is documented in US5660998 and US6379920, and imaging modes are discussed in WO2006002537A1, US9551654B2, US20170167973A1.

The disadvantages of these methods are related to the fact that the received signal directly depends on the thickness and morphology of the sample, if the sample thickness is too large, this leads to saturation, if the sample is inhomogeneous, then there will be distortion due to scattering.

In addition, it is known that imaging-based methods, such as multi-pixel technologies, can be characterized by limitations in the resolution of individual spectra due to a lower signal-to-noise ratio per pixel compared to single-point detectors.

Another disadvantage of the prior art methods is that often the spectra can contain signals caused by water or other contaminants that override the specific signals of microorganisms.

In some cases, such as in CN103217398A, drying techniques can be used, but, as noted in WO 2017/210783A1, these techniques can have damaging and irreversible effects on the microorganism, while modifying its spectrum.

In some cases, such as in WO 2017/210783 A1, multiple spectrum requests can be made to remove water components from the spectrum. However, this solution has disadvantages associated with a long analysis time and greater complexity of analytical methods for comparing spectra.

Another disadvantage of the prior art is that reference databases often contain a limited number of microbial species and thus do not cover the wide range of possible species.

For example, the database and associated identification method presented in CN103217398A refers to 13 kinds of bacteria.

Moreover, with the increase in the size of the reference databases, a series of problems arises related to the methods of cultivation and growth of microbial species, the procedures for obtaining the spectrum and the procedures for comparative analysis.

For example, as the size of the database increases, the time to analyze an unknown sample also increases, since the spectrum of the unknown sample must be compared with a large number of spectra contained in the database.

Moreover, the same microorganism species can show great variability in terms of spectral characteristics, making it difficult to compare them with database spectra.

This variability may be due to the presence of different strains of each species or differences between different cultures of the same species, caused by possible variations in the chemical composition of the culture medium and (or) growth conditions.

Thus, there is a need to develop new microbial identification methods that can work for a large number of microbial classifications (eg, taxonomic categories or phenotypic or genotypic groups) and with large variability in the reference database.

Thus, one object of the present invention is to provide an identification method that can determine whether samples of unknown microorganisms belong to one or more of the categories and/or groups described above.

Another object of the present invention is to provide a method for the identification of microorganisms, which allows to distinguish between different strains of the same species, subspecies or subclassifications of microorganisms, possibly grown in different cultures on different media or under different environmental conditions in the presence or absence of biological fluids and (or) complex matrices.

Another object of the present invention is to provide a microorganism identification method that is accurate, requires little time for each individual analysis, and does not require significant computing power, even in the case of large databases.

Another object of the present invention is to provide a device that can easily implement a microorganism identification method integrating all the various assay steps and instrumental components.

Another object of the present invention is to provide an apparatus and a method that does not require a continuous or partial connection to the Internet in order to remotely process data and subsequently achieve a result.

The author invented, tested and implemented the present invention in order to overcome the shortcomings of the prior art and achieve these and other goals and advantages.

Disclosure of the essence of the invention

The present invention is presented and characterized by independent claims. The dependent claims describe other features of the invention or variations on the basic idea of the invention.

The present invention relates to a method for identifying microorganisms and, in particular, a method for identifying microorganisms in an unknown sample by comparing the vibrational spectrum of said unknown sample with a pre-calculated model created from the vibrational comparison spectra of known samples that were previously archived in a database.

The method for identifying microorganisms of the present invention comprises the following steps:

- the stage of preparing a database from the comparison spectra of samples of known microorganisms;

- one or more steps for generating a plurality of pre-calculated models associated with categories and/or classification groups of hierarchical organized microorganisms, from broader categories and/or classification groups to smaller categories and/or classification groups, each pre-calculated model containing allowing identification of spectral characteristics associated with the category and (or) classification group of microorganisms;

- one stage of sampling of an unknown sample;

- one or more stages of obtaining the spectrum of an unknown sample;

- one or more stages of processing the spectrum of an unknown sample;

- one or more stages of analysis, in each of which it is proposed to sequentially carry out a plurality of sub-stages of comparing the spectrum of an unknown sample with the indicated pre-calculated models, and during each sub-stage by methods of multi-parameter analysis, the spectrum of an unknown sample is compared with pre-calculated models of gradually decreasing categories and (or) classification groups in order to provide at least scores demonstrating the belonging of an unknown sample to categories and (or) classification groups of microorganisms;

- one or more control steps that request a new step to obtain the spectrum of an unknown sample if the reliability parameters are below predetermined acceptance rates, or provide the end result of the method, which may contain an unsuccessful identification or a successful identification.

According to some embodiments, the step of preparing a database of comparison spectra associated with samples of known microorganisms comprises, for each sample:

- the stage of sampling a known sample, possibly obtained by growing on a solid or liquid medium in the presence or absence of biological fluids and (or) complex matrices;

- one or more stages of obtaining the spectrum of a known sample;

- one or more stages of processing the spectrum of a known sample.

In accordance with the first embodiment of the method, the unknown sample and/or the known sample may be a sample obtained by cultivation using any suitable medium, for example, solid or liquid, optionally in the presence or absence of biological fluids and/or complex matrices.

According to another embodiment, an unknown sample and/or a known sample can be directly obtained from native samples without a growth phase.

According to some embodiments, the database preparation step may be performed only once to create a database that can then be used for each subsequent analysis of unknown samples.

According to some embodiments, the step of preparing the database may also mean updating the database when new categories and/or classification groups of microorganisms are required to be included, thereby expanding the applicability of the method.

Similarly, pre-calculated models, once created, can be used for each subsequent analysis of unknown samples, or they can be updated whenever desired to add new categories and/or classification groups of microorganisms.

In accordance with the present invention, the spectra are obtained using a device capable of detecting the absorption of infrared radiation in direct or indirect mode, such as, for example, an vibrational spectrophotometer.

In accordance with some embodiments, the invention may use a continuous signal readout mode both in the absence and presence of a sample, the continuous mode facilitates, respectively, instrument cleaning operations and an objective assessment of spectral parameters (for example, signal intensity, signal-to-noise ratio, spectrum shape , the optimal level of drying) as a preparation for obtaining the spectrum.

Advantageously, according to the present invention, the monitoring of spectral parameters such as those described above is provided, this allows the quality of the resulting spectrum to be standardized, making the process independent of subsequent evaluation by the operator.

Thus, the spectra obtained in this mode refer to samples that are characterized by virtually the same level of intensity and drying and, thus, are more comparable with each other.

In accordance with some embodiments, the present invention proposes the use of algorithms that automatically identify the most significant spectral characteristics of classification groups required for identification.

This characteristic allows to significantly reduce the time required at the analysis stage, which makes it possible to speed up the process and improve the accuracy of identification.

Advantageously, with respect to the methods of the prior art, the sequential use of many pre-calculated models associated with classification groups of hierarchically organized microorganisms allows to reduce the computing power required to identify an unknown sample, does not require an Internet connection, which is necessary when using distributed or remote computer systems for data processing.

These characteristics make it possible to increase the size of the database, which allows the identification of many species, subspecies, classifications or subclassifications of different microorganisms, different strains of the same species, subspecies or subfamily, also possibly grown in different cultures on different media or under different environmental conditions, possibly in the presence of biological fluids and (or) complex matrices.

The present invention also relates to a device for performing the method for identifying microorganisms described here, comprising a device for detecting a vibrational profile, which contains a radiation source, a detector, an information collection area, defined as a place where the analyzed sample should be located, and an electronic device, stationary or a portable type associated with the detection device in which the information processing system and the data display system are installed.

In accordance with some embodiments, pre-calculated models are stored in an electronic device, these models are associated with classification groups of hierarchically organized microorganisms, from broader to smaller classification groups, each pre-calculated model contains spectral characteristics associated with a category and / or classification group of microorganisms.

Brief description of the drawings

These and other aspects, characteristics and advantages of the present invention will be understood with reference to the following description, drawings and the appended claims. The drawings, which form an integral part of the description, show some embodiments of the present invention and, together with the description, illustrate the principles of the disclosure.

The present invention can be better described and demonstrated with the help of the following drawings:

fig. 1 is a schematic representation of a possible device according to the method of the present invention;

fig. 2 is a block diagram showing, by way of example, the steps of one embodiment of the method of the present invention;

fig. 3a and 3b show experimental data collected with FTIR-ATR technology; in fig. 3a shows the collected raw data, in which each species is represented by its own color, and in Fig. 3b shows the converted data;

fig. 4a and 4b are respectively graphical representations of the sub-steps for classifying an unknown sample obtained in step E, and the corresponding table shows the scores generated in step F to determine identification reliability;

fig. 5 is a confounding matrix resulting from cross-validation of the method according to the embodiments described here with database spectra and used to evaluate the classification of microorganisms according to the category "genus".

For the purpose of ease of understanding, wherever possible, the same reference numbers are used throughout the drawings to refer to identical common elements. It should be understood that the elements and characteristics of one embodiment can be freely included in other embodiments without further specification.

Implementation of the invention

Next, we will describe in detail the various embodiments of the present invention, one or more examples of which are shown in the attached drawings. Each example is provided to illustrate the invention and should not be construed as limiting the invention. For example, features presented or described as part of one embodiment may be adapted or used in connection with other embodiments to provide yet another embodiment. It is clear that the present invention should contain all such modifications and variations.

Before considering these embodiments, we should also clarify that the present description does not limit the application to the details of the design and arrangement of components, as shown in the following description using the accompanying drawings. Other embodiments may be suggested herein and may be made or made in various other ways. It should be clarified that the phraseology and terminology used here is for description only and should not be considered limiting.

Unless otherwise indicated, all technical and scientific terms used hereinafter have the same meaning as commonly used by those skilled in the art to which the invention pertains.

Although methods and materials similar or equivalent to those described herein may be used in practice or during testing to test this invention, the methods and materials described herein are by way of example. In case of conflict, the present application containing the definitions shall prevail. The materials, methods, and examples are illustrative only and should not be considered limiting.

Here and below, by sample we mean the minimum amount of any substance in general containing microbiological organisms, such as, for example, microorganisms, but not only, for analytical purposes.

Sometimes, if necessary, we will designate a sample as an unknown sample or a known sample if the composition of the microorganisms, respectively, is unknown or known.

According to some embodiments, microorganisms may be of medical, clinical, veterinary, agricultural, food, and environmental interest.

In FIG. 1 schematically shows a device 10 for identifying microorganisms in accordance with the present invention.

In accordance with some embodiments, for direct or indirect registration of absorption of infrared radiation in the device, a device is used, such as, for example:

- infrared (IR) spectrophotometer, possibly performing Fourier transform of IR radiation;

- Raman spectrometer;

- photothermal spectrometer;

- photoacoustic spectrometer;

- combination of one of the above instruments with resonant technologies, including those not for signal amplification, such as resonant Raman, surface enhanced, surface enhanced Raman spectroscopy (SERS), surface enhanced absorption in the IR region (SEIRA), resonant surface enhanced absorption in the IR region (resonant SEIRA), tip-enhanced Raman spectroscopy (TERS).

The applicability of the present invention is, in fact, not limited by the type of technology used to obtain the vibrational profile of an unknown sample, and the method or apparatus described herein can be used in combination with any type of vibrational spectrum.

For example, this device contains an FTIR type vibrational spectrophotometer that can use any mode of sample vibrational spectrum acquisition, such as ATR (Attenuated Total Reflectance, Attenuated Total Reflectance), in which case it will be called an FTIR-ATR spectrophotometer. (Fig. 1).

The spectrophotometer can be combined with a system for processing and displaying data on the screen.

The data processing and output system may, for example, be integrated into a single processing and output system and/or may be contained in an electronic device 15 such as, for example, a screen-equipped personal computer or a portable device such as a cellular phone or tablet computer.

In accordance with some embodiments, the processing system includes a computer program that can be stored in a readable medium by the electronic device 15 and that contains instructions that can be executed by the device 10 whenever required.

In the following, for ease of demonstration, reference will be made to the electronic device 15 to denote a set of software and hardware systems capable of controlling the device 10, as well as processing and displaying data.

The spectrophotometer, itself known as its main components, comprises at least a radiation source 13 17 and a detector 14.

For simplicity, in FIG. 1 does not show other components of the spectrophotometer, such as the monochromator, chopper, interferometer.

Source 13 may emit any type of radiation 17 suitable for excitation of molecules present in the sample, for example, in the case of an IR, FTIR or FTIR-ATR spectrophotometer, this may be an infrared source 17.

According to some embodiments, the source 13 may be a Globar black body emitter or a quantum cascade laser (QCL) or a conventional infrared laser.

In accordance with some embodiments, for example, if the spectrophotometer is based on Raman technologies, the source 13 may be a laser source 13 of monochromatic light, possibly in the infrared region or even higher, depending on the particular type of Raman technology used.

In accordance with some embodiments, the radiation 17 emitted by the source 13, that is, the incident radiation 17a can be directed, possibly by means of reflective elements 16, to the sample located in the zone 18 of the information collection.

In embodiments where the ATR mode is used, incident radiation 17a may be directed at the reflective elements 16 to affect the internal reflective element.

In accordance with some embodiments, the internal reflective element may, for example, be a crystal 12.

In accordance with some embodiments, the crystal 12 may be a crystal 12 with a high refractive index.

In accordance with some embodiments, the crystal 12 may be a crystal 12 made of diamond, ZnSe, silicon or germanium.

Advantageously, if the crystal 12 is a diamond crystal 12, it has the advantage of greater reliability than ZnSe, silicon, or germanium when used for comparable measurements, and has a larger transparency range.

Reflection of radiation 17 inside the crystal 12 creates an evanescent field on the surface of the crystal 12, on which the data acquisition zone 18 can be defined.

This evanescent field can penetrate the sample to a certain depth, which, depending on the situation, can reach several microns.

In particular, this depth is a function of the angle of incidence and wavelength of the incident radiation 17a, as well as the refractive index of the material used for the crystal 12, and thus can be considered virtually constant for all analyzed samples.

For this reason, advantageously, the ATR mode produces spectra that are independent of the optical path through the sample and thus independent of the sample thickness, which guarantees higher reproducibility than either transmission or reflection approaches.

Moreover, this mode allows you to limit or prevent the phenomenon of scattering and / or signal saturation, which can adversely affect the quality of the spectrum.

Thus, this mode guarantees greater measurement repeatability and easier sample preparation.

The output radiation 17, after interaction with the sample, i.e. the outgoing radiation 17b, is then directed by the reflective elements 16 to the detector 14.

According to some embodiments, detector 14 may be a DLaTGS (triglycine sulfate supplemented with deuterated L-alanine) detector.

In accordance with alternative embodiments, detector 14 may be a DTGS detector. According to alternative embodiments, the detector 14 may be an MCT (Mercury Calcium Telluride) detector, either as a single device or as part of an array.

In accordance with alternative embodiments, the detector 14 may be a CCD detector.

According to alternative embodiments, the detector may be a bolometer or microbolometer, either as a single device or as part of an array.

For example, the detector 14 converts the optical information contained in the outgoing radiation 17b into electrical signals that are transmitted to the electronic device 15.

According to some embodiments, the device is positioned to take spectra in NIR (near IR) and/or MIR (mid IR) and/or FIR (far IR).

In accordance with some embodiments, the device 10 can operate in a continuous collection of information, that is, displaying on the screen of the electronic device 15 in real time the information that is requested each time in the collection area 18.

The present invention also relates to a method for identifying microorganisms in a sample composition by evaluating the vibrational profile shown schematically in FIG. 2. The method contains the following steps:

- stage A of preparing a database of comparison spectra of known samples of known microorganisms, containing:

- one or more stages A1 of the selection of a known sample;

- one or more steps A2 of obtaining the spectrum of a known sample;

- one or more steps A3, possibly automated, of processing the spectrum of a known sample in order to at least identify spectral regions in which to evaluate the presence of identification-enabled spectral characteristics of each category and/or classification group;

- one or more steps A4 of creating a plurality of pre-calculated models associated with categories and (or) classification groups of hierarchically organized microorganisms, from broader categories and (or) classification groups to smaller categories and (or) classification groups, each pre-calculated model contains allowing identification of spectral characteristics associated with the category and (or) classification group of microorganisms;

- stage B of sampling of an unknown sample;

- one or more steps C to obtain the spectrum of an unknown sample;

- one or more processing steps D of the spectrum of the unknown sample based on what is determined in step A3;

- one or more stages E, each of which is designed to sequentially carry out a plurality of sub-stages of comparing the spectrum of an unknown sample with the specified pre-calculated models, and during each sub-stage, using multi-parameter analysis methods, the spectrum of an unknown sample is compared with pre-calculated models of gradually decreasing categories and (or) classification groups in order to provide at least scores demonstrating the belonging of an unknown sample to categories and (or) classification groups of microorganisms;

- one or more control steps F, which request a new step to obtain spectrum C of an unknown sample if the reliability parameters are below predetermined acceptance rates, or provide the end result of the method, which may contain unsuccessful identification J or successful identification G, H, I.

In accordance with some embodiments, these steps A1, A2, A3 can be repeated when it is desired to include a new known sample in the database.

According to some embodiments, in the sampling steps B, A1, the sample may be subjected to preliminary procedures for adding nutrients to the microorganisms.

In accordance with some embodiments, the sample can be grown on a solid culture medium.

In accordance with some embodiments, the sample can be grown in Petri dishes.

According to some embodiments, the sample can be grown in a liquid culture medium and then centrifuged to obtain a concentrated pellet.

According to some embodiments, the sample can be grown in liquid culture medium or in liquid culture broth and then filtered.

According to some embodiments, the sample can be grown in liquid culture medium or in liquid culture broth and then subjected to concentration or enrichment procedures to increase the concentration of microorganisms present.

In accordance with other embodiments, the analyzed sample can be obtained directly from a native sample without a growth phase.

In accordance with some embodiments, the sample may contain biological fluids and/or complex matrices of human, animal, environmental, agricultural, and food origin. According to some embodiments, the spectrum acquisition steps C and A2 may comprise a preliminary step of cleaning the surfaces of the sample contacting device, such as the acquisition area 18 shown in FIG. 1.

In accordance with some embodiments, it is possible to obtain background spectra, for example, in a continuous acquisition mode, in order to check the effectiveness of such cleaning, for example, by the disappearance of absorption bands associated with impurities from the acquisition zone 18.

In accordance with some embodiments, in steps C and A2, it is also proposed that the sample for analysis be placed in the collection zone 18, possibly located on the crystal 12 of the spectrophotometer.

In accordance with some embodiments, the sample for analysis is placed in the collection zone 18 in solid form, for example, by removing it and placing it with a disposable stick.

In accordance with some embodiments, the sample can be pressed against the collection zone 18, for example, using a torque press.

In accordance with some embodiments, the sample dryness level and spectral parameters (eg, signal intensity, signal to noise ratio, spectrum shape) are monitored spectroscopically prior to obtaining the spectrum.

In accordance with some embodiments, it is possible to estimate the content of water leaving the device, in a continuous collection of information, tracking the decrease, up to the possible complete disappearance, of the absorption bands related to the spectral characteristics of water, in the respective ranges of wave numbers.

In accordance with the present invention, a sample spectrum is obtained when a predetermined standard dryness level and spectral parameter values are reached.

Advantageously, this feature overcomes or at least limits the problem of the prior art whereby the spectrum may contain water-induced signals that overlap or interfere with microorganism-specific signals.

During the spectrum acquisition steps C, A2, it is also proposed to record the spectrum obtained during the analysis of the sample.

In accordance with some embodiments, spectrum acquisition provides a predetermined number of consecutive spectra, which are then averaged to improve the signal-to-noise ratio.

In accordance with some embodiments, the number of spectra that can be obtained and averaged for each analysis is in the range from 8 to 512, preferably from 32 to 256 for each analysis, more preferably from 64 to 128 for each analysis.

Spectrum processing or spectrum averaging steps D, A3 can provide the following:

- identification, possibly automatic, of spectral regions in which an assessment is made of the presence of spectral characteristics that identify each category and (or) classification group;

- use of algorithms for linear and (or) non-linear interpolation and correlation of spectral profiles;

- calculation of the first and second derivatives of the spectral profile;

- normalization of derivatives using vector normalization algorithms in the entire spectral range. For example, in FIG. 3a shows the spectra that can be collected from various samples, and FIG. 3b shows the spectra after processing.

By repeating steps A1, A2, A3 for several known samples, a database of comparison spectra of known microorganisms can be obtained.

However, in accordance with some embodiments of the present method, it is also contemplated that the database may contain spectra of known microorganisms obtained by other methods.

In accordance with some embodiments, the database contains spectra related to monomicrobial cultures belonging to various classification groups of known microorganisms.

In accordance with some embodiments, the database contains spectra related to various strains of known microorganisms for each category and/or classification group.

In accordance with some embodiments, the database contains spectra related to samples grown on culture media, for example, but not limited to, agar, CNA agar, CLED agar, blood agar, chromogenic agar, Sabouraud agar.

In accordance with some embodiments, the database contains spectra relating to samples grown in liquid growth broth, which are then centrifuged or filtered or enriched to obtain a precipitate of concentrated samples.

In accordance with some embodiments, the database contains spectra related to samples obtained without a growth phase, that is, obtained from native samples, directly from any substance or material.

Advantageously, measurement of spectra obtained during growth in a liquid broth and subsequent precipitation or filtration or enrichment allows the spectra to be identified directly in the presence of biological fluids, eg body fluids and/or complex matrices.

This heterogeneity and variability of the spectra contained in the database allows for a general analysis that enables the identification of microorganisms, taking into account the effects associated with growth on different culture media.

In accordance with embodiments, when each spectrum is obtained several times, several individual spectra and/or an average spectrum calculated for repetitions can be included in the database.

Step A4 involves the creation, starting from the spectrum from the database, of pre-calculated models, that is, reference libraries containing spectral characteristics that allow for the identification of spectral characteristics associated with classification groups of hierarchically organized microorganisms.

In accordance with some embodiments, pre-calculated models may be stored in the electronic device 15 of the device 10.

In accordance with some embodiments, spectral characteristics allowing for identification may, for example, contain shapes, sizes, intensities and areas of peaks, correlations and relationships between intensities or between areas of different peaks, frequencies of maximum peak values.

In accordance with some embodiments, spectral characteristics allowing identification may contain spectral profiles in the spectral zones from 950 to 1280 cm ^-1 (associated with nucleic acids, carbohydrates and polysaccharides), from 1280 to 1480 cm ^-1 (associated with starches, for example, proteins, methyl and methylene, e.g. lipids), from 1700 to 1800 cm ^-1 (associated with carbonyl groups, e.g. lipids), from 2800 to 3000 cm ^-1 (associated with aliphatic chains, e.g. lipids), shown, for example, in fig. 3a and 3b.

According to some embodiments, classification groups may contain taxonomic categories such as, for example, domain, biological kingdom, phylum, class, order, family, tribe, genus, species, subspecies.

According to some embodiments, classification groups may, for example, be prokaryotes, eukaryotes, archaea, gram-positive bacteria, gram-negative bacteria, yeasts, filamentous fungi.

According to further embodiments, classification groups may, for example, be taxonomic categories and/or phenotypic groups and/or genotypic groups.

In accordance with some embodiments, classification groups may be hierarchically organized from broader classification groups to smaller classification groups, such that broader classification groups contain many smaller classification groups, for example, one microorganism genus may include several species, and one microorganism species may include several subtypes.

In accordance with some embodiments, pre-calculated models may contain a list of identifying spectral characteristics related to each category and/or classification group.

In step A4, the spectra from the database are then compared to identify identifying spectral characteristics that are common to classification groups of microorganisms.

According to some embodiments, an automatic recognition system can be provided to identify identifyable spectral features associated with classification groups.

In accordance with some embodiments, spectral characteristics associated with classification groups that allow identification can be determined using the technologies and / or statistical analysis approaches mentioned below, based on neural networks and / or machine learning.

According to some embodiments, it is thus possible to create an arbitrary number of pre-calculated models based on the classification groups.

Thus, each time the spectrum of the unknown sample is compared with the model in analysis step E, this comparison can only be made for the spectral intervals and/or for the spectral features of greatest interest, and not for the entire spectral range and/or for all spectral characteristics.

For example, if you want to compare the spectrum of an unknown sample obtained with a resolution of 1 cm ^-1 over the entire spectral range with a database that contains, for example, 18,000 comparison spectra obtained with the same resolution, then in accordance with the methods of the prior art, this the comparison would require an evaluation of 65 million points.

The author has found that when using precalculated models in accordance with the present invention, this comparison will require an estimate of the number of points, which is from 20 to 80 times less than according to the methods of the prior art.

Thus, the use of pre-calculated models makes it possible to reduce the required computing resources, which thereby eliminates remote data analysis and (or) distributed computing, even in the presence of large databases.

This feature thus allows the size of databases to be greatly increased so that they contain a large number of classification groups, which improves and expands the predictive performance and identification performance of the method of the present invention.

This feature also makes it possible to work with spectra even at high resolution, increasing the accuracy and precision of the method.

Moreover, for example, the presence of a list may make it redundant to use least squares methods, and (or) possible calculations of root mean square values are not required to be carried out in a whole range of wave numbers for all spectra.

According to some embodiments, analysis step E proposes to perform a plurality of successive comparison sub-steps in which the spectrum of the unknown sample is compared to pre-calculated models.

In particular, at each sub-step, the spectrum of the unknown sample is compared with pre-calculated models of gradually decreasing classification groups and at least several points are obtained that characterize the belonging of the unknown sample to the classification groups of microorganisms.

In accordance with some embodiments of the present invention, statistical and/or chemometric methods of multivariable analysis can be used for comparison, such as, for example, principal component analysis (PCA) or linear discriminant analysis (LDA), as well as linear discriminant partial least squares. (LDPLS), Quadratic Discriminant Analysis (QD), Hierarchical Cluster Analysis (HCA), Random Tree Ensemble, or any combination of these and other technologies not explicitly mentioned.

In accordance with some embodiments, methods, technologies or algorithms that implement approaches based on neural networks can be used for said comparison.

According to some embodiments, methods, technologies or algorithms that implement approaches based on artificial (machine) learning can be used for said comparison.

In accordance with some embodiments of the present invention, said comparison may also involve a comparison of first and/or second derived spectra and models.

In accordance with some embodiments, statistical weights can be used in various ways to account for different regions of the spectrum.

In accordance with some embodiments of the present invention, the comparison between spectra and model can be performed by determining the distance between the spectrum and the model, for example, using methods based on the method of least squares.

In accordance with some embodiments, the distance can be used as a metric to perform statistical and/or chemometric analysis.

In accordance with some embodiments, the difference between the spectra can be used for PCA and/or LDA analysis.

According to some embodiments, a combination of statistical methods may be used, which involves performing PCA followed by linear discriminant principal component analysis by LDA.

According to these embodiments, if the PCA result is subjected to subsequent LDA analysis, the separation of the resulting PCA clusters is enhanced, which also allows the identification of species whose spectral characteristics are very similar to each other.

In accordance with some embodiments, the comparison sub-steps may be performed to compare the first unknown spectrum with pre-calculated models of broader classification groups in order to identify the broader category and/or classification group to which the unknown sample belongs; then the unknown spectrum is compared with pre-calculated models of progressively decreasing classification groups to determine the progressively decreasing classification group to which the unknown sample belongs.

Advantageously, this feature allows fewer comparisons between spectrum types than other known methods.

For example, if the database contains 45 spectrum types associated with 45 different species, known prior art methods would attempt to identify the unknown sample by comparing it to all 45 species spectrum types available.

On the other hand, according to embodiments of the method of the present invention, 45 species can be classified according to broader classification macrogroups (eg 4) with similar characteristics.

Each macrogroup contains, for example, 3 smaller classification groups, each of which in turn contains 2 even smaller classification subgroups, the latter, for example, are associated with different species or subspecies.

In this case, 3 sub-stages can be performed, and at the first sub-stage, the unknown sample is assigned to the first classification macrogroup, at the second sub-stage, to the second classification group, and at the third stage, it is assigned to the third classification group.

Advantageously, this feature allows the spectra to be separated into similar macrogroups, simplifying the data matrix and allowing, once the first step is completed, to work with a less complex data matrix in which it is easier to identify features that separate, for example, specific similar species or subspecies.

Moreover, compared to known methods, this feature allows a faster and more efficient method, both in terms of time and in terms of consumption of computing resources.

In accordance with some embodiments, at each comparison sub-step, multivariable analysis techniques provide scores demonstrating belonging to each of the classification groups against which the unknown spectrum is compared.

These membership scores can be obtained in known ways using multivariate analysis techniques as described above, and can be normalized and expressed as a percentage.

For example, in accordance with some embodiments, membership scores may be associated with a linear discriminant between principal components.

In accordance with some embodiments, at control step F, it is proposed to check the reliability of the identification performed at analysis step E.

This control can be carried out, for example, by checking that at each sub-step, at least one classification group is identified to which the unknown sample belongs, having a membership score higher than a predetermined acceptance rate, for example, higher than 70%.

In this case (block G in Fig. 2), the unknown sample is identified as belonging to a smaller classification group.

If in one or more comparison sub-steps one of the membership scores is less than a predetermined acceptance index, according to the method, it is proposed to perform the second step of obtaining the spectrum C, the second processing step D and the second analysis step E, this is the second attempt to analyze the unknown sample.

If two analysis steps E assign the unknown sample to the same classification groups (block H of FIG. 2), the unknown sample is identified according to the smallest classification group assigned by the two analysis steps E.

Otherwise, a third analysis, a third spectrum acquisition step C, a third processing step D, and a third analysis step E are assumed.

If, in the third analysis step E, the unknown sample is assigned to the same classification groups as in at least one of the previous analysis steps E (block I in Fig. 2), the unknown sample is identified according to the smallest classification group determined by the two steps analysis E with the corresponding results.

Otherwise, a message about the impossibility of identification is displayed (block J in Fig. 2).

In accordance with one embodiment, a computer program, or software, is provided that is used to collect data and analyze it against a database that can be stored in a computer-readable environment and that contains instructions that, when executed on an analysis device for classification and identification microorganisms carry out the implementation of the method in accordance with the present invention.

Example 1

In FIG. 4a is a schematic flow diagram of a process performed on an unknown spectrum to successively determine the various categories of interest by applying a plurality of predictive models (step E) in accordance with the present invention. The unknown spectrum is compared with a plurality of predictive models developed in steps A1, A2, A3, A4 based on a database of known spectra. Each model is based, for example, on a PCA-LDA analysis to assign a specific type among all possible for the defined category. In particular, for each panel of FIG. 4a, each point corresponds to the coordinates assigned during the PCA-LDA calculation to the spectrum present in the database of known samples (step A4); each point in the set A, B, C reflects the specific type to which it belongs for the defined category. According to model 1, in which category 1 is assigned to the different types represented, all spectra of known samples are grouped into the corresponding membership sets A, B, C (first panel of Fig. 4a). Thus, for model 1, three different types are possible, each of which is associated with a specific part of the space defined by the model.

For each unknown spectrum, step E predicts the spatial coordinates it occupies in the space described by the model and, by estimating its location in that space, determines the type of category of interest. For example, in the first panel of FIG. 4a, it can be seen that for model 1, its coordinates are similar to those of the spectrum of the model belonging to type A, and thus it is assigned to the same type.

The hierarchical subdivision of the models from broader to narrower allows, through the application of sub-steps in which the application of multiple models is proposed, to obtain more and more specific information regarding the type of microorganism of interest (second and third panels of Fig. 4a).

It should be noted that belonging to a specific type within the scope provided by category 1 determines which model will be the next model applied to the unknown sample.

Thus, the unknown spectrum is compared with pre-calculated models of gradually decreasing classification groups in order to identify more and more detailed categories of the unknown sample (Fig. 4a).

For each model applied in the analysis step E of the unknown sample, a score is also calculated indicating belonging to the associated type, possibly in the form of a percentage. This value indicates how many coordinates calculated for the unknown spectrum are similar to those determined in the spectrum model of the same category (table in Fig. 4b).

Then, in step F, shown schematically in FIG. 4b, the membership scores associated with each of the types identified in the various sub-steps of the process are evaluated to assess the reliability of the result. For each model, acceptability indicators are determined, above which the result can be considered reliable. If, in the course of the comparison described above, all membership score values are higher than predetermined values for each model, the result is reported (step G), otherwise a new analysis is performed.

Example 2

According to one embodiment of the method, it can be used to identify any possible taxonomic characteristic described above, such as, for example, the genus of a microorganism. See fig. 5, which shows the mixing matrix obtained during the validation of the method with the spectrum from the database of the present invention, in this matrix one can observe the correspondence between actual and predicted values from 97.5% to 100%.

It is understood that modifications and/or additions of details or steps can be made to the method and/or apparatus described above without departing from the scope of the present invention. It is also understood that, although the present invention has been described with reference to some specific examples, a person skilled in the art will undoubtedly be able to obtain many other equivalent forms of the method and/or apparatus, while using the characteristics described in the claims, and thus thus, they all fall within the protection area defined here.

Claims (21)

1. A method for identifying microorganisms in a sample by assessing the vibrational spectrum, which includes the following steps: - stage (A) of preparing a database with comparison spectra of known samples of known microorganisms, containing: - one or more steps (A1) of taking a known sample; - one or more steps (A2) of obtaining a spectrum of a known sample; - one or more stages (A3) processing the spectrum of a known sample in order to at least identify the spectral regions in which to evaluate the presence of spectral characteristics that identify each category and/or classification group; - one or more steps (A4) of creating a plurality of pre-calculated models associated with categories and/or classification groups of hierarchically organized microorganisms, from broader categories and/or classification groups to smaller categories and/or classification groups, each pre-calculated model contains the specified allowing identification of spectral characteristics associated with the category and / or classification group of microorganisms; - step (B) sampling an unknown sample; - one or more steps (C) to obtain the spectrum of an unknown sample; - one or more steps (D) processing the spectrum of the unknown sample based on what is determined in step (A3); - one or more stages (E), at each of which a plurality of sub-stages of comparing the spectrum of the unknown sample with the specified pre-calculated models are sequentially performed, and at each sub-stage, the spectrum of the unknown sample is compared using multi-parameter analysis methods with pre-calculated models of gradually decreasing categories and / or classification groups , in order to provide at least scores showing that the unknown sample belongs to the specified categories and / or classification groups of microorganisms; - one or more control steps (F) that compare the membership scores determined in step (E) with predetermined acceptance scores and, if the scores are higher than the acceptance scores, present an end result that contains a successful identification (G), and if the membership scores are below predetermined acceptance indicators, the second stage of obtaining the spectrum (C) of the unknown sample is requested, and based on the results of the processing stages (D) and the analysis stages (E), a successful identification (H) is presented or the third stage of obtaining the spectrum is requested ( C) an unknown sample, and the results of the processing steps (D) and the analysis steps (E) provide the final result of the method, which contains a successful identification (I) or unsuccessful identification (J). 2. The method according to claim 1, characterized in that said steps (A3) of processing the spectrum of a known sample are automated. 3. The method according to claim 1, characterized in that at said control step (F) it is checked that at each of said comparison sub-steps at least one classification group is identified to which the unknown sample belongs, with a membership score higher than specified in advance predetermined acceptance rate. 4. The method according to any one of paragraphs. 1-3, characterized in that in step (A4) of creating a plurality of pre-calculated models, combined statistical methods are used to perform a principal component analysis followed by a linear principal component discriminant analysis (PCA-LDA); said precalculated models are then used to predict the categories and/or classification groups in step (E) of analyzing the unknown sample. 5. The method according to any one of paragraphs. 1-4, characterized in that the unknown sample and/or the known sample are grown using any suitable culture medium, in the presence or absence of biological fluids and/or complex matrices. 6. The method according to any one of paragraphs. 1-4, characterized in that the unknown sample and/or the known sample is a sample obtained directly from a native sample that has not undergone a growth phase. 7. The method according to any one of paragraphs. 1-6, characterized in that at the stage of analysis (E) methods based on neural networks are used. 8. The method according to any one of paragraphs. 1-7, characterized in that at steps (C, A2) a continuous recording mode is performed both in the absence and in the presence of a sample, this continuous mode contributes, respectively, to the instrument cleaning operations and an objective assessment of spectral parameters before standardized spectrum acquisition. 9. Method according to claim 8, characterized in that said spectral parameters are signal intensity, signal-to-noise ratio, spectrum shape, optimal drying level. 10. A device for implementing a method for identifying microorganisms in a sample according to any one of paragraphs. 1-9, containing a device for detecting a vibrational spectrum, which contains a radiation source (13) (17), a detector (14), a certain information collection area (18) designed to locate the analyzed sample on it, and an electronic device (15) of a stationary or portable type, which is associated with a device for detection and in which an information processing system and a data display system are installed, while in the specified electronic device (15) pre-calculated models are stored associated with the classification groups of hierarchically organized microorganisms, from larger groups to smaller ones. classification groups, with each pre-calculated model containing the identification spectral characteristics associated with the classification group of microorganisms, so that there is no need for an Internet connection to perform the computational operations necessary to obtain the result. 11. A computer-readable medium containing a computer program or software used to collect data and analyze it against a database and which contain instructions that, when executed on a device for identifying microorganisms, determine the execution of a method according to any one of paragraphs. 1-9.

Images