Method for spectral characterization of microorganisms
Through the use of a variable condition classifier, the analysis instability problem caused by environmental changes in microbial spectral characterization is solved, and the reliability and accuracy of microbial classification under different conditions is achieved, ensuring the robustness of spectral analysis.
Patent Information
- Application Number
- CN202410989467.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Priority Date
- 2020-02-27
- Filing Date
- 2020-04-10
- Publication Date
- 2024-11-08
AI Technical Summary
In microbial spectral characterization, changes in humidity and temperature affect the reliability of the analysis results, resulting in instability and error of classification. Especially when measuring under different environmental conditions, the prior art is difficult to effectively alleviate this problem.
Using a change condition classifier, the classifier is trained to give greater weight to the spectral characteristics that are not affected by environmental changes by using reference data under different environmental conditions, thereby masking the impact of environmental changes during the measurement period, ensuring the stability of the spectral characteristics and Distinguishable.
It realizes robust classification of microorganisms under different humidity and temperature conditions, avoids the impact of environmental changes on spectral analysis, improves the reliability and accuracy of the analysis, and ensures accurate identification and classification of microorganisms.
Smart Images

Figure CN118914125A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to a method for spectroscopic characterization of microorganisms, in particular an infrared spectroscopic method. Background Art
[0002] The prior art will be explained below with reference to specific aspects. However, this should not be construed as limiting. Useful further developments and modifications known from the prior art may also be used above and beyond the relatively narrow scope of this introduction, and will be apparent to those skilled in the art after reading the following disclosure.
[0003] During the spectral characterization of microorganisms such as bacteria, fungi, yeasts, algae or protozoa (and viruses), the spectral measurements can be subject to variations outside the organism itself, which affect the content of the spectrum and therefore also the quality of the analysis. Outside the organism itself means that the source of these variations is not in the systematic classification of the microorganism.
[0004] An example can be in the infrared spectroscopy characterization under different humidity conditions, which are dominant during the measurement, and in the measurement chamber of the infrared spectrometer, the prepared sample is inevitably exposed to these humidity conditions because of the lack of shielding measures. The humidity of the sample has a direct impact on the rotation and stretching vibrations reflected in the infrared spectrum. For example, the absolute humidity of the air in the warm season is generally higher than that in the colder season (seasonal dependence). Humidity can also depend on the location where the measurement is performed. Consider, for example, a mobile infrared spectrometer in a vehicle of an emergency management organization, which can be used at the coast (humidity tends to be high) and in highlands or even in the mountains (humidity tends to be low).
[0005] When carrying out actual characterization measurement under conditions significantly different from the conditions that dominated during the recording of reference data, this change of humidity becomes a problem. When characterizing microorganisms, reference data are usually used to determine phylogenetic taxa, such as being divided into kingdoms, phyla, classes, orders, families, genera and species at a level from top to bottom. The reliability of characterization may be subject to the differences caused by humidity. This example can be transferred from humidity to the dominant temperature in the measurement chamber, because temperature at least indirectly affects the ability of the prepared sample to absorb humidity, and therefore affects the vibration characteristics of molecules.
[0006] In the above examples, instrument-based measures can be taken to mitigate the listed problems. The infrared spectrometer can be designed with a measurement chamber with a gas seal, or the measurement chamber can be continuously flushed with a gas pre-conditioned for humidity and temperature, so that constant, uniform conditions are maintained in the chamber during the measurement. Such constancy will ensure that there are no significant differences from the conditions that prevailed during the recording of the reference data. However, spectrometer manufacturers have not yet begun to engage in such gas seals.
[0007] Various reports on the spectroscopic characterization of microorganisms are available. Reference is made to the following by way of example:
[0008] US2008 / 0132418A1 explains a method for characterizing microorganisms for medical diagnosis or food control and environmental monitoring purposes. At least one spectral image of a microorganism having a number of pixels is obtained, and one or more spectra are selected from the spectral image having a number of pixels based on a predetermined spectral feature. The selected spectrum contains spectral information characteristics of the microorganism. Identification can be performed by comparing the spectrum of the selected microorganism with the spectrum of a reference microorganism in a database. The database can be created by the following steps: obtaining at least one spectral image having a number of pixels for each of a plurality of reference microorganisms, wherein each pixel includes a signal corresponding to the spectrum of the reference microorganism; and selecting a spectrum from the spectral image having a number of pixels based on a predetermined spectral feature to create a database, the database containing at least one spectrum for each of the reference microorganisms.
[0009] MJ Gupta et al. (Transactions of the ASABE, Vol. 49(4), 2006, 1249-1255) reported the identification and quantification of four common enteric bacteria (Escherichia coli; E. coli O26, Salmonella typhimurium, Yersinia enterocolitica and Shigella boydii) in foods in four different food matrices from the principal components obtained from infrared spectroscopy and artificial neural networks (ANNs) with the help of FTIR spectroscopy. The classification accuracy of the ANN for identification and quantification was given as 93.4% and 95.1%, respectively. The ANN was validated using an independent data set obtained from separately cultured and grown enteric bacteria. For this validation, the accuracy of the ANN for detecting Yersinia enterocolitica was given as between 64% and 100%. This work highlights the challenge of filtering background noise in the spectra.
[0010] EP2174116B1 describes a method for typing or identifying microorganisms by vibrational spectroscopy. A microbial sample is subjected to vibrational spectroscopy analysis, wherein signal changes caused by bleachable components in the sample are substantially eliminated from the vibrational spectrum obtained. Bleachable components are identified as those components that produce vibrational spectral bands that exhibit reduced intensity when they are optically bleached. The intention is to provide vibrational spectral information that is corrected for typing and identification purposes.
[0011] US2002 / 138210A1 discloses a method for compensating for the drift of a fingerprint spectrum caused by changes in environmental factors, comprising: culturing a microorganism of interest and a second microorganism under a first set of environmental factors, assuming that the second microorganism is metabolically similar to the microorganism of interest; measuring the fingerprint spectrum of the microorganism of interest cultured under the first set of environmental factors and the fingerprint spectrum of the second microorganism also cultured under the first set of environmental factors; detecting the difference between the fingerprint spectrum of the second microorganism cultured under the first set of environmental factors and the fingerprint spectrum of the second microorganism cultured under the second set of environmental factors; and utilizing the difference between the fingerprint spectra of the second microorganism cultured under the two sets of environmental factors to convert the fingerprint spectrum of the microorganism of interest cultured under the first set of environmental factors into an expected fingerprint spectrum of the microorganism of interest under the second set of environmental factors.
[0012] EP1319176B1 describes a method for characterizing a spectrometer instrument based on instrument variations existing between instruments and / or variations over time within the same instrument, comprising the following steps: providing multiple spectra of known standards from at least one spectrometer instrument; classifying the one or more spectrometer instruments in at least one of a plurality of predefined clusters based on spectral features extracted from the one or more spectra; and providing at least one calibration model for each predefined cluster, wherein each calibration model compensates for the instrument variations of the instruments classified in the corresponding cluster.
[0013] Therefore, there is a need to make the spectral characterization of microorganisms more robust and reliable, including in the presence of variations during the measurement. Other tasks that can be achieved by the present invention will be immediately clear to those skilled in the art by reading the following disclosure. Summary of the invention
[0014] The present invention relates to a method for spectroscopic characterization of microorganisms, in particular an infrared spectroscopic method.
[0015] Many different analytical techniques are available for the classification and identification of microorganisms such as bacteria, fungi, yeasts, algae or protozoa (also viruses), but many of these are based on specific biochemical substances that must be extracted, modified and detected. The most commonly used are test systems that describe the properties and enzyme activities of bacteria, such as bile solubility, catalase, coagulase, deoxyribonuclease, motility, toxins, Optochin, streptomycin, oxidase, Gram staining, Ziehl-Neelsen staining, spores, casings, fermentation capacity, etc. Other commonly used methods are cell fatty acid analysis, protein spectrum, polymerase chain reaction (PCR) and 16S rRNA-gene sequencing, total DNA G+C content, high-resolution gas / liquid chromatography and microscopy to determine cell morphology based on differences in shape and appearance. What all of the above methods have in common is that they use only one specific marker of bacteria for classification. In contrast, infrared spectroscopy simultaneously detects all biomolecules present in bacterial cells because almost every biomolecule absorbs infrared radiation. The spectral information resulting from the interaction between infrared radiation and biomolecules includes all cellular components and can therefore serve as an almost complete phenotypic description of the bacteria analyzed.
[0016] A test microorganism is provided, and its identity is known at the first classification level. The identity of the test microorganism at the first classification level can be predetermined by at least one of the following methods in particular: (i) mass spectrum, (ii) infrared spectrum, (iii) growth on selective culture medium (" API (analytical spectrum index) test ") and (iv) gene sequence analysis. In DE102013022016B4, examples of determining the prior art can be found, wherein in particular, the mass spectrum determination of the species of microorganism is supplemented with the infrared spectrum determination of subspecies. In addition, the method can include separating the test microorganism from the habitat (such as biological and / or chemical matrix). The separation of the test microorganism preferably includes removing the matrix, such as by washing, filtering, centrifuging, evaporating, extracting, precipitating and / or other separation forms. In addition, the provision of the test microorganism can include a proliferation step, such as in a nutrient solution or on a plate nutrient medium (such as agar) to cultivate.
[0017] The test microorganism is preferably sterilized. The sterilization method may include exposing the test microorganism to a metabolically inhibiting liquid, such as an alcohol (such as ethanol or isopropanol), or an acid (such as formic acid), and / or exposure to an energy source, such as heat or high-energy radiation (possibly ultraviolet light). Sterilization refers in particular to the loss of the microorganism's ability to proliferate even under favorable conditions in order to prevent the biological risks associated with unintentional / uncontrolled transmission occurring in analytical laboratories. In some cases, it may not be necessary to sterilize the test microorganism, such as when working in an analytical laboratory with a biosafety level of 2 or higher, because it can be safely assumed that the professionals working there are fully trained.
[0018] Under the condition of the influence of at least one source of variation that allows not to derive from the systematic classification of test microorganisms, obtain spectral measurement data from the test microorganisms of possible sterilization. Said one or more variations can be of atmospheric origin, and particularly comprise different values about at least one of the following aspects: temperature, humidity, pressure and carbon dioxide content of ambient air. The variation of atmospheric origin or atmospheric variation itself show as the variation of the condition of approaching and surrounding the test microorganism sample prepared for spectral measurement, and said test microorganism sample is prepared on sample slide (for example small glass plate) conventionally. One or more variations preferably relate to the variation that can be affected on continuous and limited aspect, particularly those variations that are limited at least at one end (as being absolute zero for temperature), or those variations with lower limit and upper limit, as being 0% to 100% for relative humidity.
[0019] Select a classifier that has been trained to determine the identity of a microorganism on a second classification level that is subordinate to the first classification level. The possible identity of the classifier on the second classification level is assigned to the known identity of the test microorganism on the first classification level. The first and second classification levels can be directly adjacent, such as species and subspecies, respectively. In addition, the first classification level can correspond to species, and the second classification level can include different varieties, such as pathogenic and non-pathogenic varieties, resistant and susceptible varieties, or different strains of species. In order to characterize, these different types of varieties / strains can be grouped into identity subclasses, such as all pathogenic serotypes / strains in the first group and all non-pathogenic serotypes / strains in another group different from the first group.
[0020] The term serotype or serovar (short for serovar) is used to describe individual species within a bacterial subspecies, which can be distinguished by serological tests. They differ in terms of antigens on the cell surface and are recognized in conventional microbiology by specific antibodies. The classification hierarchy of serotypes is as follows: genus => species => subspecies (subsp.) => serotype, e.g., the added binomial species name Salmonella enterica subspecies serovar Typhimurium, the short form is Salmonella Typhimurium.
[0021] Pathotypes (from the Greek word for "disease") are bacterial strains or groups of strains that have the same characteristics, whose pathogenicity distinguishes them from other strains within a species or subspecies. Pathotypes are named by the third or fourth addition of the binomial species name. For example, the bacterium Xanthomonas axonopodis, which can cause citrus canker, has multiple pathotypes with different host specializations: X. axonopodis pv. citri is one of them. The abbreviation "pv." stands for "pathotype." Virulent strains of human pathogens also have pathotypes, but in this case they are named by a prefix before the name. For example, the mostly completely harmless enterobacterium E. coli has the highly dangerous pathotypes enterohemorrhagic E. coli (EHEC), enteropathogenic E. coli (EPEC), enterotoxigenic E. coli (ETEC), enteroinvasive E. coli (EIEC), enteroaggregative E. coli (EAEC) and dispersed adherent E. coli (DAEC). The pathotypes can in turn include different serotypes: for example, EHEC has many known serotypes. About 60% of all recognized EHEC serotypes are O157, O103 and O26. The serotype O157 / H7 is particularly dangerous.
[0022] In a broader sense, characterization of microorganisms may also include varieties that differ in other medically relevant properties, especially resistance to antimicrobial substances such as antibiotics (especially β-lactam antibiotics and glycopeptide antibiotics) and antifungal agents, but also in toxin formation ("toxivars") or sensitivity to the same or similar bacteriophages ("phage variants"). Usually, if a species or subspecies of microorganisms is selected to have common biological characteristics, the term "biovar" is used. An example of an antibiotic-resistant variety is MRSA: methicillin-resistant Staphylococcus aureus.
[0023] The term "strain" describes a population grown from a single organism and maintained at a (usually in a running) repository of microbial strains. Internationally standardized strain names are added to the nomenclature chain including genus, species, subspecies, and variant. The individual organisms of a strain are genetically identical; different strains differ slightly in their genetic makeup.
[0024] A classifier is applied to the measured data to determine the identity of the test microorganism at a second taxonomic level. The classifier is also variation-conditioned in that it is obtained by training on spectral reference data loaded with directional variations of different known reference microorganisms that exhibit the same identity as the test microorganism at a first taxonomic level and that cover a different identity at a second taxonomic level. The training includes the provision that those spectral features of the first type from the reference data that facilitate the distinction of the different identity at the second taxonomic level are given a greater weight than those spectral features of the second type from the reference data that are subject to directional variations, so as to largely or completely mask the effects of the variation in the characterization of the test microorganism at the second taxonomic level.
[0025] In another aspect of the invention, in order to distinguish the identity of a microorganism at a predetermined taxonomic level, a classifier adjusted for at least one variation is applied to spectral measurement data, the spectral measurement data being acquired from a (optionally sterilized) test microorganism under conditions that allow one or more variations to have an effect. One or more variations do not have their origin in the systematic classification of the microorganism.
[0026] The expert in the field will realize that all embodiments disclosed in connection with this method can also be transferred to this application. BRIEF DESCRIPTION OF THE DRAWINGS
[0027] The present invention may be better understood by reference to the following drawings, which are not necessarily drawn to scale but are intended primarily to illustrate the principles of the invention (mostly schematically):
[0028] Figure 1A Shown is the principal component analysis (PCA) in three-dimensional space of the infrared spectra of different serotypes (SV) of S. pneumoniae as a function of different relative humidity (rH).
[0029] Figure 1B Shown with Figure 1A Same spectral evaluation, but Figure 1B The evaluation was performed by linear discriminant analysis aimed at removing the effect of humidity on changes in spectral features.
[0030] Figure 2 A very simplified diagram of the measuring setup of a Fourier transform infrared spectrometer (FT-IR) in transmission and the acquired spectrum is shown.
[0031] Figure 3 An example sequence of steps for preparing a spectroscopic microbiological sample prior to measurement (counterclockwise) is shown.
[0032] Figure 4Shown is a principal component analysis (PCA) in two-dimensional space of infrared spectra of different serotypes (SV) of Legionella pneumophila as a function of different relative humidity (rH), wherein the serotypes are grouped by pathogenicity. DETAILED DESCRIPTION
[0033] While the invention has been shown and described with reference to several embodiments, those skilled in the art will recognize that various changes in form and details may be made without departing from the scope of the technical teaching as defined in the following claims.
[0034] Figure 1A The principal component analysis (PCA) of the infrared spectra of different serotypes (SV) of the microorganism Streptococcus pneumoniae at four different humidity levels (rH#1 to rH#4) between 10% and 80% relative humidity, which are specially set up for measurement in the measurement chamber of an infrared spectrometer. As expected, the respective accumulation of data points is spread over the serotypes. Figure 1A , for clarity, the four resulting data clouds for two example serotypes are emphasized with different dotted ellipses (SV#1 and SV#2). The measurements are all based on several strains / isolates of one serotype, as reflected in the elongated data clouds for the individual humidity levels. Different strains / isolates are marked by different shapes (such as pyramids, cubes, and balls). In addition, the separation of the measurements for the same serotype at different humidity levels is clearly visible. Surprisingly, it is also obvious that the humidity dependence of the spectral features does not cause the data clouds of different serotypes to merge and / or confuse each other. On the contrary, the individual data clouds of different serotypes remain separated ("layered"), including for different humidity levels, and can therefore be distinguished during the characterization process.
[0035] This finding explains the hypothesis that this property of stable distinguishability, even when variations during spectral measurements are considerable (10% to 80% humidity range), can be used to eliminate this variation in the characterization through advanced evaluation methods (e.g. machine learning methods) and thus avoid the need for complex conversion of the spectrometer for this purpose.
[0036] Therefore, the second evaluation Figure 1A The specific purpose of using linear discriminant analysis is to identify and increase the weight of those spectral features (such as principal components) which maximize the uniqueness of the individual known serotypes, while those spectral features affected by the variation within the individual serotype classes, which essentially originate from different strains (although intra-organism variation) and humidity levels during the measurement period, are masked as much as possible.
[0037] The results are as follows Figure 1B As shown. It can be seen that Figure 1AAs shown in , in the layers of four distinct data clouds per serotype, the four different humidity levels are merged into one continuous elongated data cloud per serotype, regardless of humidity, without causing the separate serotype data clouds to be superimposed or overlapped with each other. This result can be used to create a changing condition classifier that can assign newly determined spectral features to known spatial volumes when applied to new spectra of S. pneumoniae of unknown serotype. To this end, the volumes of the multidimensional feature space can be marked, which are below the data clouds, and a check is performed to determine which volume of the feature space is the volume into which the important spectral features of the new spectrum are projected.
[0038] For infrared spectroscopy characterization measurements in transmission, Fourier transform spectrometers (FT-IR) can be used, which provide high resolution. Figure 2 Figure 1. The spectra are usually taken from 4000cm -1 Up to 500cm -1 Measurement. Figure 2 The dashed line in the middle is 1350cm -1 Left and right and 700cm -1 The highlighted area between left and right (bottom) is considered to be particularly effective for spectral specificity. At an acquisition rate of twenty spectra per second, several hundred spectra were measured and summed in order to improve the signal-to-noise ratio.
[0039] Infrared spectroscopy is based on thousands of vibrations of functional groups and polar bonds in biological materials; these in turn originate from all components of the microbial cell, such as DNA, RNA, proteins, internal structures, membranes and cell walls, up to energy storage. There is no clear molecular assignment of individual features in the spectrum, even if certain spectral ranges can be preferentially assigned to certain molecular species: fatty acids in the range 3050-2800 cm -1 , where the CH2 and CH3 groups vibrate, and the amide range is 1750-1500 cm -1 , where the peptide bond vibrations for polysaccharides range from 1200-900 cm -1 . From 900 to 700cm -1 The range is sometimes called the fingerprint range because it contains something from all molecules and is very important for distinguishing species.
[0040] In a slightly modified embodiment, infrared spectra can also be measured in reflected light. In this case, for example, they are prepared on a metallic reflective substrate made of aluminum. Raman spectroscopy can also be used, which has the advantage that the spectra of the prepared microorganisms can also be measured in liquids and also requires smaller amounts of sample material.
[0041] from Figure 1A and Figure 1BThe acquired knowledge can be implemented to create a change condition classifier by following the steps below:
[0042] (i) Preparing a reference microorganism and specifying the variation(s).
[0043] The first task is to specify the class to be distinguished. If it is assumed that the species (such as Streptococcus pneumoniae) as the identity on the first taxonomic level is known, the purpose can be to determine the corresponding serotype as a possible identity on the second subordinate taxonomic level. As an example, the 23 serotypes of Streptococcus pneumoniae most frequently found in clinical trials can be selected. Reference biomass of these microorganisms can be obtained from public-operated repositories (e.g., the Leibniz Institute DSMZ in Braunschweig, the German Collection of Microorganisms and Cell Cultures GmbH (microorganisms and cell culture institutions)).
[0044] In order to fully account for the variation within organisms, a representative selection of microorganisms of the class to be distinguished can be considered. Depending on availability, in the example of S. pneumoniae, this can be three to six different strains per serotype; in the case of the 23 most common serotypes, 69 to 138 strains can be used to compile reference data and create a classifier.
[0045] The next task is to specify the parameter, the variation of which will be imposed on the recording of the reference data and for which the variable occurrence during the infrared spectroscopy measurement is possible. This can be an atmospheric variation, such as humidity, pressure, gas concentration or temperature. In principle, more than one variation parameter can be considered when recording reference data (such as humidity and temperature). However, a wider coverage of conceivable variations is also associated with a corresponding increase in the work required to measure the reference data, since different representative values or reference values of the variation parameter must be recorded in combination with each other. A list of reference points for the variation parameter(s) covering all actual conditions during the spectral measurement is selected. It should be possible to interpolate between the values of this representative selection of reference points.
[0046] (ii) Recording reference data
[0047] First, strains of reference microorganisms can be prepared in a standardized manner. For example, after cultivation on or in a suitable culture medium, and if necessary, after sterilization to prevent biological contamination, they can be deposited on a sample slide for infrared spectroscopy in several repetitions and then introduced into a measurement chamber. The measurement chamber is maintained at a constant predetermined value for (one or more) variable parameters, such as 10% relative humidity at 20°C. After the sample slide is introduced, it is preferred to wait for a certain length of time, such as five to ten minutes, so that the biomass of the prepared reference microorganism can adapt to the preset conditions.
[0048] After all parameters have been determined, reference data for the prepared reference microorganisms can be recorded under predetermined conditions. The procedure is repeated under appropriately varying conditions, i.e., for example, at 30%, 55% and 85% relative humidity and a constant 20°C. Each change in the variation value should be followed by an adaptation period of several minutes to allow transient processes to decay and to obtain reproducible stable results.
[0049] This method of recording reference data can be supplemented by measuring dependent variations, such as those resulting from slightly different culture conditions (biological replicates) or preparation conditions (e.g. technical replicates, use of different batches of reagents / agents or chemicals), or from measurements performed on different spectrometers to allow for instrument variations. The reference data thus recorded are checked for completeness, obvious outliers (e.g. using local outlier factorization methods (LOF)) and / or plausibility, and the reference data are corrected and / or re-recorded if necessary.
[0050] (iii) Training of the changing condition classifier
[0051] Preferably, machine learning methods are used, such as artificial neural networks (ANN) or linear discriminant analysis (LDA). With regard to class affiliation, such as serotype #1, serotype #2, ..., serotype #23 in the aforementioned example of Streptococcus pneumoniae, the training is supervised. However, with regard to changing conditions, i.e. different environmental conditions (e.g. humidity) or other influencing factors (e.g. changing cultivation, preparation, spectrometer), the training of the classifier is unsupervised. This is equivalent to the requirement of emphasizing the importance of those spectral features in the reference data, which maximize the uniqueness of the individual classes (here the serotypes of the species Streptococcus pneumoniae), while those spectral features that are strongly affected by the changes have lower weights and are therefore actually masked. For example, the spectral characteristics can present themselves as principal components.
[0052] Briefly, and for the purpose of illustration (without claiming any strict scientific correctness), the machine learning algorithm identifies those partial volumes in the generally multidimensional, multivariate feature space, each of which is assigned to one of the classes being distinguished (i.e., identity at the second taxonomic level). An unexpected aspect of this essentially known approach to accounting for interference is that atmospheric variations (such as relative humidity) do not cause the spectral signature of one serotype / strain to overlap with the spectral signature of other serotypes / strains when humidity varies, but rather they remain separated, thus ensuring distinguishability in the space of spectral signatures also under such varying conditions.
[0053] As is usual in such a training phase using reference data, there is the option of testing the efficiency of the resulting classifier by cross-validation. If appropriate, the machine learning algorithm can be adjusted based on the results of the cross-validation in order to further improve the accuracy of the classifier.
[0054] (iv) Verification (optional)
[0055] When the taxonomic assignment of one or more test microorganisms is known, a validation test run can be conducted under conditions that allow for expected variations (e.g., varying relative humidity) in order to validate the efficacy based on external data.
[0056] This process of creating classifiers can be repeated to create a varying condition classifier database with a very wide range of microorganisms, which in turn can be identified at different taxonomic levels. Preferably, reference data can be obtained and processed from pathogens that occur with the greatest frequency in clinical settings.
[0057] After creating the variation condition classifier, the method for characterizing microorganisms can be performed as follows, see Figure 3 An illustrative sequence in:
[0058] First, the identity of the test microorganism must be known or must have been determined using a mass spectrometer (e.g., MALDI (Bruker Daltonik GmbH, Bremen, Germany)) is determined on the first classification level (such as species). On this basis, a change condition classifier suitable for the determined identity is selected. As an example, the method described in EP3083981A1 is paid attention to in this context.
[0059] In order to obtain enough biomass, the test microorganism can be cultivated in a nutrient solution or on a plate nutrient medium. Then, the microbial cells so grown can be removed from the nutrient medium, for example by separating them from the nutrient solution, for example by centrifugation or filtration, or by sampling from an agar plate. For sterilization, the microorganisms thus harvested can be resuspended in an activity-inhibiting liquid, such as ethanol (e.g., 70% v / v).
[0060] Microorganisms are very sensitive to changes in growth conditions (such as different culture media, temperature, nutrients, changes in gas supply (oxygen and other gases), humidity, culture time, etc.). These factors can cause changes in cell composition and changes in metabolism, which can be detected using infrared spectroscopy. In order to cultivate, the cell material of pure single colonies can be spread on agar plates using a scraper to produce confluent growth. This technology enables cells to be sampled in a very reproducible mixture of different growth stages that are always present in the colony. For most clinically relevant strains, the optimal culture period is about 16 to 24 hours, and the culture temperature commonly used for bacteria is about 35°C to 37°C. Sample material of the cultured test microorganism can be harvested directly from the center of the cell layer, for example, using a calibrated platinum loop with a diameter of one millimeter (step A).
[0061] When the test microorganism is grown on a flat nutrient medium (such as agar), the biomass can be sampled from one or more bacterium colonies and directly deposited on a spectral sample slide. It is important to ensure uniform distribution, and the biomass is sterilized (for example, in the case of Streptococcus pneumoniae) by irradiating the biomass with ultraviolet light. Alternatively, the biomass can be resuspended in a metabolism-inhibiting liquid (step B) equally. The liquid can also be deionized water, which does not usually exert any metabolism-inhibiting effect. In this case, the test microorganism can also be sterilized by ultraviolet radiation or other energy sources (for example, heat) after being placed on the test position of the sample slide.
[0062] It must be noted that no nutrient medium residues that would interfere with the measurement results adhere to the test microorganisms extracted or taken out from the nutrient medium. In order to achieve a uniform distribution of the biomass of the test microorganism in the suspension, a small cylinder or bead of a reaction inert material (such as steel) can be added to the suspension, and then the sealed suspension container can be shaken (step C). The suspension is then divided into equal parts and repeated multiple times, for example, by a pipette with a plastic tip and gently applied to the sample slide (step D), and the number of repetitions can vary with the scheme. The uniform application of uniform layer thickness guarantees the best measurement results (step E). After all the samples studied are applied to the sample slide, they are left to stand for several minutes at a specific temperature (such as 37° C.), such as ten to thirty minutes, to dry the suspension (step F). If the test microorganism is applied to the sample slide as soon as it is harvested from the culture container, without any further resuspension, drying can be completely omitted, or at least the time for its performance can be shorter.
[0063] The sample slide thus prepared can then be introduced into the measuring chamber of the spectrometer and measured sample by sample under conditions allowing for the influence of at least one source of variation. Several locations on the sample slide can also be coated with test standard biomass to check the technical performance of the spectrometer, for example according to the applicant's method explained in EP3392342A1.
[0064] The recorded spectra can be analyzed with a pre-created change condition classifier after the usual processing steps (such as baseline subtraction, smoothing and calculation of the second derivative). As mentioned above, only those spectral features that are not affected by the change or are only affected to a slight extent by the change are considered here (or at least mainly considered), while those spectral features that show high changes caused by the change are largely or completely masked.
[0065] Processing the measured data with the changing condition classifier results in the investigated spectrum being assigned to one of the possible identities at the second taxonomic level. In the example of Streptococcus pneumoniae, this refers to one of the mentioned serotypes. Only in rare cases is a reliable characterization not possible, for example due to unforeseen disturbances during cultivation, sample preparation or measurement, or because the identity of the test microorganism sought at the second taxonomic level is not included in the reference data (e.g. in the case of very rare serotypes, which are hardly relevant in clinical practice).
[0066] Figure 4 Another example showing the consistent distinguishability of serotypes in principal component space even at varying humidity levels is shown here in two dimensions for clarity (PC 2, PC 3). The basis is the bacterial species Legionella pneumophila, fifteen of whose serotypes are considered in the diagram of its spectral characteristics. However, the serotypes are not classified individually, but grouped into strongly pathogenic and weakly pathogenic serotypes. This would result in the first serotype SV#1 (data cloud on the right), which is responsible for 70-80% of all Legionella in the person, being distinguished from the next fourteen SV#2 to SV#15 (if any), which are much less pathogenic. Similarly, susceptible (sensitive) and resistant serotypes / strains can also be grouped together.
[0067] Reference data of the reference microorganism Legionella pneumophila of different serotypes were obtained at four different relative humidity (arid 10%, semi-arid 30%, humid 55% and tropical 85%), with the base strains coded with different symbols (such as triangles, squares and circles). In the figure, this change is essentially manifested as an elongation of the data cloud along the principal component axis PC2. However, it is obvious that, regardless of the change, the data cloud belonging to serotype 1 (SV#1) is sufficiently removed from the data cloud of the grouping of other serotypes SV#2 to SV#15 to ensure distinguishability based on spectral features. If the number of clinically experienced diarrheal cases increases, such as using the established mass spectrometry MALDI-TOF method, such diarrheal cases can be attributed to the bacterial species Legionella pneumophila in the first analysis, then a suitably trained classifier can be used in the subsequent infrared spectral analysis of the isolated and cultured pathogen to distinguish the specific pathogenic serotype SV#1 from other less dangerous serotypes SV#2 to SV#15, so as to start a specific treatment in the case of a positive result. Of course, this process can be transferred to other microorganisms. The flexibility of the classifier creation described here is unlimited.
[0068] Starting from the above method, during the spectral measurement, for a plurality of possible microorganisms and a plurality of possible sources of variation, individually or in combination, a variation condition classifier is determined. In the case where the identity of the microorganism to be characterized is known at a first taxonomic level, a spectral sub-characterization of the identity at a second subordinate taxonomic level can therefore be robustly and reliably performed by selecting an appropriate variation condition classifier.
[0069] In addition to the embodiments described by way of example, other embodiments of the present invention may be envisioned. Based on the knowledge of this disclosure, those skilled in the art can easily design additional advantageous embodiments, which will be covered by the protection scope of the claims (including any equivalents, as the case may be).
Claims
1. A test system for spectral characterization of microorganisms, comprising: - a mass spectrometer for determining the identity of the test microorganisms at the first taxonomic level; an infrared spectrometer having a measurement chamber in which a sample slide carrying biomass of the test microorganism is placed during acquisition of infrared spectral measurement data from the test microorganism; and - a processing device having access to a database comprising a plurality of classifiers of varying conditions, each classifier being trained to determine the identity of a test microorganism at a second classification level subordinate to the first classification level, wherein the possible identities of the classifier at the second classification level are assigned to the identity of the test microorganism at the first classification level, and further, each classifier being conditional with respect to at least one variation in order to distinguish between the identities of microorganisms at the second classification level, acquiring infrared spectroscopy measurement data from the test microorganism under conditions that allow at least one variation to have an effect, wherein, said at least one change does not have its origin in the systematic classification of said microorganism, Therein, the processing device selects a classifier from the database based on the identity of the test microorganism determined at the first classification level, and applies it to the infrared spectrum measurement data obtained from the test microorganism to determine its identity at the second classification level.
2. The test system according to claim 1, wherein the mass spectrometer is a MALDI type.
3. The test system according to claim 1, wherein the test microorganism is selected from the group consisting of: bacteria, fungi, yeast, algae, protozoa, and viruses. The testing system of claim 1 , wherein the first taxonomic level corresponds to a species. The test system of claim 1 , wherein the second taxonomic level corresponds to a subspecies.
6. The testing system of claim 1, wherein the infrared spectrometer is of Fourier transform type.
7. The testing system according to claim 1, wherein the infrared spectrometer is of Raman type.
8. The testing system of claim 1, wherein the infrared spectrometer operates in transmission.
9. The testing system of claim 8, wherein the slide comprises a glass plate.
10. The testing system of claim 1, wherein the infrared spectrometer operates in reflection.
11. The testing system of claim 10, wherein the slide comprises a metallic reflective substrate.
12. The testing system of claim 11, wherein the metallic reflective substrate comprises aluminum.
13. The testing system of claim 1, wherein the at least one change is of atmospheric origin.
14. The testing system of claim 13, wherein the at least one change comprises a different value with respect to at least one of: temperature, humidity, pressure, and carbon dioxide content of the ambient air.
15. The test system according to claim 1, wherein: The classifiers in the database are obtained and trained with the help of one or more machine learning methods.
16. The test system according to claim 15, wherein: The machine learning method includes at least one of an artificial neural network (ANN) or a linear discriminant analysis (LDA).
17. The test system according to claim 1, wherein: The at least one change involves a change having an effect in a continuous and finite aspect.
18. The test system of claim 17, wherein the variation of the at least one is limited at at least one end.
19. The testing system of claim 18, wherein the at least one variation has a lower limit and an upper limit.
