Method for monitoring yeast fermentation process and pre-judging quality of fermented product

Abstract

The invention provides a method for monitoring a yeast fermentation process and pre-judging product quality. The method comprises the following steps: collecting yeast samples and recording fermentation days; the method comprises the following steps: extracting total DNA (Deoxyribonucleic Acid) of a sample, and respectively and quantitatively detecting the copy number of 16S rRNA genes of bacteria and the copy number of 18S rRNA genes of fungi; inputting the bacterial 16S rRNA gene copy number, the fungus 18S rRNA gene copy number and the fermentation day number into a random forest classification model, and outputting a normal or abnormal fermentation state; and pre-judging the product quality based on the fermentation state. The method breaks through the limitation of traditional dependence on sensory experience and a culture method, the microbial biomass is reflected in real time through absolute quantification of the microbial gene copy number, the detection time is remarkably shortened, and the accuracy of the random forest classification model is extremely high.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of brewing, in particular to a method for monitoring the fermentation process of Daqu and predicting the quality of fermentation products. BACKGROUND

[0002] As a typical representative of Chinese traditional distilled liquor, the core of the brewing process of Luzhou-flavor liquor lies in the synergistic effect of the two-stage fermentation of Daqu and fermented grains. Among them, Daqu, as a saccharifying agent carrier, provides an indispensable "microbial community" (microbial community), "enzyme system" (functional enzymes) and "material system" (flavor precursor substances) for the brewing process, which directly determines the yield, quality and style characteristics of the liquor.

[0003] Currently, the production of Luzhou-flavor Daqu mainly adopts a solid-state fermentation mode of natural inoculation: wheat is used as the main raw material, which is soaked, crushed, and pressed into shape, and then relies on the dynamic fermentation of environmental microbial community in the koji room, and then is stored and matured in the koji warehouse before being put into use. In this process, the growth and metabolism of microorganisms is the key factor affecting the quality of Daqu. The microbial community structure of Daqu is complex, covering bacteria, molds, yeasts, actinomycetes and other main groups. The microbial community not only contributes key hydrolytic enzymes such as amylase and protease, but also directly participates in the synthesis of alcohol, aldehyde, acid, ester, ketone and other aroma substances, which has a decisive influence on the formation of koji aroma and flavor quality.

[0004] However, the traditional Daqu fermentation process monitoring method has significant limitations, which seriously restricts the precise control of the production process, which is reflected in the following aspects:

[0005] 1. Strong dependence on sensory experience: the current mainstream method relies on human experience, which is judged by observing the cloth hanging state of the koji skin, the water running condition of the koji, the growth density of the cross-section mycelium, and the smell of the flavor characteristics, etc. Subjective indicators. This method is greatly affected by individual experience differences, lacks objective quantitative standards, and is difficult to achieve cross-batch, cross-line quality stability control.

[0006] 2. In recent years, although some studies have proposed to detect volatile substances (such as Chinese patent CN1144146868) or specific metabolites to indirectly infer the microbial state, but these indicators can only reflect the accumulation results of metabolites, and cannot reflect the dynamic changes of microbial biomass in real time and quantitatively, making it difficult to capture abnormal fluctuations in the early stage of fermentation.

[0007] 3、Traditional microbial quantitative methods have inherent defects, such as relying on the cultivability of microorganisms, long culture period (usually more than 48 hours), and being unable to effectively detect microorganisms with harsh growth conditions (such as anaerobic bacteria) and non-culturable microorganisms (more than 90% of environmental microorganisms), resulting in a serious underestimate of the actual biomass; it is difficult to accurately count the mold groups with complex morphology (such as mycelium), and the equipment cost is high, the pretreatment is complex, which is not suitable for rapid screening in industrial production scenes.

[0008] In summary, the existing means cannot obtain real-time data of microbial biomass, and cannot construct a quantitative correlation model between biomass and fermentation state; because the abnormal fluctuations of microbial growth (such as imbalance of microbial community or sudden decrease of biomass) cannot be identified in time, the fermentation is often out of control and then adjusted passively, resulting in waste of raw materials and loss of quality. SUMMARY

[0009] The present application aims to provide a method for monitoring the fermentation process of Daqu and predicting the quality of fermentation products, which directly quantifies the biomass of microorganisms (such as gene copy number) by qPCR, solves the problem of missed detection of non-culturable microorganisms by traditional culture method; the detection can be completed in a very short time, overcoming the 48-hour lag defect of traditional methods, realizing early warning of fermentation abnormalities, and adjusting the fermentation conditions of Daqu in time based on abnormal state to reduce waste of raw materials.

[0010] To achieve the above purpose, the present application proposes the following technical solutions:

[0011] A method for monitoring the fermentation process of Daqu, comprising the following steps:

[0012] S1 collecting samples in the fermentation process of Daqu and recording the fermentation days;

[0013] S2 extracting total DNA of the sample, and quantitatively detecting the bacterial 16S rRNA gene copy number and the fungal 18S rRNA gene copy number in the sample;

[0014] S3 inputting the detected bacterial 16S rRNA gene copy number, fungal 18S rRNA gene copy number and fermentation days into a pre-trained random forest classification model;

[0015] S4 determining the fermentation state of Daqu as normal or abnormal according to the classification result output by the random forest classification model;

[0016] Wherein, the random forest classification model takes fermentation days, bacterial 16S rRNA gene copy number and fungal 18S rRNA gene copy number as input features, and takes normal or abnormal fermentation state as classification target.

[0017] As a preferred technical solution of the present application, in the S3 step, the construction of the random forest classification model comprises the following steps:

[0018] a. Establish a bacterial quantitative standard curve: amplify the bacterial 16S rRNA gene using the first primer pair 338F / 518R to generate a bacterial quantitative curve equation;

[0019] b. Establish a fungal quantitative standard curve: amplify the fungal 18S rRNA gene using the second primer pair FungiQuant-F / FungiQuant-R and the probe FungiQuant-Prb to generate a fungal quantitative curve equation;

[0020] c. Collect training samples in stages: randomly collect a plurality of Daqu samples at predetermined sampling time points, and record the fermentation days of the current Daqu. The fermentation state of the Daqu, including normal state and abnormal state, is evaluated through sensory and physicochemical detection;

[0021] d. Detect the bacterial 16S rRNA gene copy number and the fungal 18S rRNA gene copy number of the Daqu sample based on the bacterial quantitative curve equation and the fungal quantitative curve equation;

[0022] e. Construct a random forest classification model with fermentation days, bacterial 16S rRNA gene copy number, and fungal 18S rRNA gene copy number as independent variables, and fermentation state as dependent variable.

[0023] As a preferred technical solution of the present application, in the a step, the bacterial quantitative curve equation is Ct=-3.321logN+49.789.

[0024] As a preferred technical solution of the present application, in the b step, the fungal quantitative curve equation is Ct=-2.586logN+35.269.

[0025] As a preferred technical solution of the present application, in the step c, the total amount of Daqu samples collected at each sampling time point is at least 100, of which 80% is used as the training set and 20% is used as the test set.

[0026] As a preferred technical solution of the present application, the random forest classification model is constructed based on at least 100 times of 5-fold cross-validation of the training set;

[0027] The discriminant ability of the random forest classification model is verified based on the test set, and the verification indicators include AUC value, accuracy, sensitivity, and specificity.

[0028] As a preferred embodiment of the present invention, the sampling time points are fermentation days 0, 2, 4, 6, 8, 10, 13, 16, 19, and 21, and no less than 10 spatially parallel samples are collected at each sampling time point.

[0029] This invention also provides a method for predicting the quality of fermented Daqu products, employing the method described above for monitoring the Daqu fermentation process, and directly associating the fermentation state output by the random forest classification model with the quality of the fermented Daqu products, including:

[0030] If the fermentation state of the Daqu is normal, it is predicted that the fermented product meets the quality standards.

[0031] If the fermentation state of the Daqu (a type of starter culture) is abnormal, it is predicted that the fermented product has quality defects.

[0032] As can be seen from the above technical solutions, the present invention provides a method for monitoring the fermentation process of Daqu (a type of starter culture) and predicting the quality of fermented products. It directly quantifies the biomass of microorganisms (such as gene copy number) using qPCR, solving the problem of missed detection of unculturable microorganisms in traditional culture methods. Detection can be completed in a very short time, overcoming the 48-hour lag of traditional methods, enabling early warning of fermentation anomalies, and timely adjustment of Daqu fermentation conditions (such as temperature and humidity in the fermentation room) based on abnormal conditions, reducing raw material waste. A random forest classification model constructed based on at least 100 samples achieves a cross-validation AUC of 0.9946 and a test set accuracy of 98.49%, significantly outperforming traditional human sensory experience judgment. Furthermore, the random forest classification model only requires three input parameters (fermentation days, bacterial gene copy number, and fungal gene copy number), significantly lowering the application threshold in industrial scenarios. The random forest classification model of the present invention has extremely high accuracy, supports standardized quality control across batches and production lines, and solves the quality fluctuation problem caused by reliance on human experience.

[0033] It should be understood that all combinations of the foregoing concepts and the additional concepts described in more detail below can be considered part of the inventive subject matter of this disclosure, provided that such concepts do not contradict each other.

[0034] The foregoing and other aspects, embodiments, and features of the teachings of the present invention will be more fully understood from the following description in conjunction with the accompanying drawings. Other additional aspects of the invention, such as features and / or beneficial effects of exemplary embodiments, will become apparent from the following description or may be learned through practice of specific embodiments according to the teachings of the present invention. Attached Figure Description

[0035] The accompanying drawings are not drawn to scale. In the drawings, each identical or nearly identical component shown in the various figures may be denoted by the same reference numeral. For clarity, not every component is labeled in each figure. Embodiments of various aspects of the invention will now be described by way of example and with reference to the accompanying drawings, wherein:

[0036] Figure 1 The gene copy number of microorganisms in samples under different fermentation states in this embodiment of the invention;

[0037] Figure 2 This is a cross-validation ROC curve of the random forest classification model in this embodiment of the invention.

[0038] Figure 3 This is a distribution chart of the cross-validation performance metrics of the random forest classification model in this embodiment of the invention.

[0039] Figure 4 This is a heatmap of the confusion matrix of the random forest classification model in an embodiment of the present invention.

[0040] Figure 5 This is the ROC curve of the random forest classification model in this embodiment of the invention on the test set;

[0041] Figure 6 This is the prediction result of the random forest classification model in this embodiment of the invention on the test set. Detailed Implementation

[0042] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the described embodiments of the present invention without creative effort are within the scope of protection of the present invention. Unless otherwise defined, the technical or scientific terms used herein should have the ordinary meaning understood by those skilled in the art.

[0043] The terms "first", "second", and similar terms used herein and in the claims are used to distinguish one element from another, and are not necessarily used to describe a sequential or chronological order. Also, unless the context clearly indicates otherwise, the use of the terms "a", "an", or "the" or similar referents includes a singular number or plural number, unless the context clearly indicates otherwise. The terms "comprises", "comprising", "includes", "including" and the like can mean the presence of one or more features, integers, steps, operations, elements, and / or components listed in the specification that follow "comprises", "comprising", "includes", "including" and the like, but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups thereof.

[0044] I. Establishing a microbial load quantification method

[0045] A microbial load quantification method for bacteria and fungi is established, and the specific operation steps are as follows:

[0046] In the embodiments of the present application, the genomic DNA of Bacillus subtilis strain commonly found in Daqu is used as the bacterial quantification template, and the genomic DNA of Saccharomyces cerevisiae strain is used as the fungal quantification template. The former is amplified using specific primers 338F and 518R, and the latter is amplified using specific primers FungiQuant-F, FungiQuant-R, and probe FungiQuant-Prb (Table 1). Among them, Bacillus is one of the dominant genera of microorganisms in Daqu, and Bacillus subtilis is a common and typical species in this genus, which is easy to isolate from Daqu samples and has a bacterial universal conserved amplification region. Therefore, Bacillus subtilis is preferred as a representative bacterial species in the embodiments of the present application. Similarly, Saccharomyces is the dominant genus of fungi in Daqu, and Saccharomyces cerevisiae is the dominant species, which is easy to isolate, and its amplification region can cover more than 90% of the fungal groups in Daqu. Therefore, Saccharomyces cerevisiae is preferred as a representative fungal species. Of course, in other specific embodiments, the genomic DNA of other strains can also be used as a template, as long as it belongs to a common bacterial or fungal species in Daqu and has a 16S rRNA gene and an 18S rRNA gene fragment, respectively.

[0047] In addition, primers 338F and 518R target the V3-V4 hypervariable region of the bacterial 16S rRNA gene, which can exclude archaeal interference; primers FungiQuant-F, FungiQuant-R, and probe FungiQuant-Prb target the conserved region of the fungal 18S rRNA gene, which has broad amplification capability for major groups of Daqu, including Ascomycota (containing yeast) and Basidiomycota.

[0048] The amplification products of Bacillus subtilis and Saccharomyces cerevisiae were respectively ligated to pMD19-T vector, and after the construction of recombinant plasmids, they were transformed into E. coli competent cells (DH5a competent cell) for culture. Positive clones were picked and plasmids were extracted. The plasmid copy number N was calculated according to the following formula:

[0049]

[0050] wherein m is the mass of the plasmid (ng), L is the length of the plasmid (bp), and the average molecular weight of double-stranded DNA is 660 D / bp.

[0051] qPCR was used to quantitatively analyze the two kinds of plasmids, and the ABI 7500 / 7500 Real-Time PCR system was used to target the plasmid insert for amplification. The standard curve was established with plasmid mass from 10 2 ng, 10 1 ng, 10 0 ng, 10 -1 ng, 10 -2 ng, 10 -3 ng, 10 -4 ng, and the function relationship was established according to the threshold cycle number (Ct, Cycle threshold) and the plasmid copy number N, respectively, to obtain the bacterial quantitative standard curve equation: Ct=-3.321 log N+49.789 and the fungal quantitative standard curve equation: Ct=-2.586 log N+35.269 (Table 1).

[0052] Table 1 Quantitative analysis of microbial load

[0053]

[0054] II. Preparation of Daqu samples

[0055] In the embodiment of the present application, in order to cover a larger sample collection range as much as possible, random sampling is performed in 10 rooms, and the sampling time is the day when the Daqu enters the room (day 0) and the 2nd, 4th, 6th, 8th, 10th, 13th, 16th, 19th, 21st day after fermentation (covering the main fermentation period, the wet fire period and the large fire period of Daqu fermentation), 10 samples are randomly selected from different positions in each room at each time point, and a total of 100 samples are collected. Whether the 100 samples are normal or not is judged by experienced professionals through sensory evaluation and physicochemical detection indexes, such as shown in Table 1. Figure 1

[0056] After each sample is crushed, an appropriate amount of each sample is accurately weighed and added to sterilized water, shaken and mixed, and then the supernatant is centrifuged to collect the bacterial precipitate. The precipitate of each sample is ground in liquid nitrogen and then transferred to a sodium laurate buffer, sequentially extracted with phenol-chloroform-isopropyl alcohol and phenol-chloroform, and then precipitated with isopropyl alcohol, rinsed with ethanol and dried to obtain a Daqu genomic DNA sample.

[0057] Based on the established microbial load quantification method, the ABI 7500 / 7500 Real-Time PCR system and specific primers are used to amplify the target bacterial 16S rRNA gene and fungal 18S rRNA gene sequences. The copy number of the bacterial 16S rRNA gene and the fungal 18S rRNA gene in the sample is calculated according to the bacterial quantification standard curve equation and the fungal quantification standard curve equation, respectively.

[0058] III. Constructing a random forest classification model

[0059] In the embodiment of the present application, the data set is divided according to the fermentation days of Daqu, and 80% of all samples from the same time point are randomly selected as the training set, and the remaining 20% are used as the test set. For example, 100 parallel samples (covering different rooms and different positions in the same room) are collected on the 0th day of Daqu fermentation, of which 80 samples are used as the training set and 20 samples are used as the test set. The collection and division of the remaining fermentation days (such as the 2nd, 4th, 6th, 8th, 10th, 13th, 16th, 19th, 21st day) are the same as those on the 0th day.

[0060] The R package randomForest is used for 100 times of 5-fold cross-validation of the training set. The specific operation method is as follows: the samples in the training set are randomly divided into 5 parts, 4 of which are randomly selected for training, and the remaining one is used as the test set. The fermentation days, the copy number of the bacterial 16S rRNA gene and the copy number of the fungal 18S rRNA gene are used as independent variables, and the fermentation state (normal / abnormal) is used as the dependent variable to construct a random forest classification model. Rotate the five samples as the test set and construct a random forest classification model. Repeat the above process 100 times, as shown in Table 2. Figure 2 ​As shown, 500 cross-validation ROC curves (Receiver Operating Characteristic Curves) were obtained, along with three key performance indicators in cross-validation: AUC (Area Under the ROC Curve), sensitivity, and specificity. Figure 3 As shown, the cross-validation results show that the random forest classification model has an average AUC of 0.9946, an average sensitivity of 0.988, and an average specificity of 0.899, indicating that the model has a very strong classification ability and can accurately identify most samples.

[0061] like Figure 4 , Figure 5 As shown, the model's performance on the test set demonstrates excellent discriminative ability (AUC = 0.9968) and extremely high accuracy (98.49%), precision (99.4%), and recall (98.8%), exhibiting good predictive ability. The performance difference between the training and test sets is 0.0022 < 1%, indicating no significant overfitting. The model exhibits low variance in cross-validation (ROCSD = 0.005), indicating strong robustness. Figure 6 As shown, the prediction results indicate that, except for a few samples, the model has a good predictive effect on samples with a wide range of microbial copy numbers and different fermentation states.

[0062] Sensitivity: The proportion of samples that the model correctly predicts as anomalous out of the actual anomalous samples.

[0063] Specificity: The proportion of samples that the model correctly predicts to be normal out of the actual normal samples.

[0064] Accuracy: The proportion of samples that the model correctly predicts out of the total sample.

[0065] Precision: The proportion of samples that the model predicts to be anomalous that are actually anomalous.

[0066] Recall: The proportion of samples that the model correctly predicts as anomalous out of the actual anomalous samples.

[0067] IV. Application of Random Forest Classification Model

[0068] To use a trained random forest classification model to predict new samples, follow these steps: First, ensure that the new sample data format is consistent with the training data and complete the corresponding preprocessing; then, load the model using the readRDS function; next, call the predict function to predict the new samples; finally, determine the fermentation status of the starter culture based on the prediction results and take appropriate measures.

[0069] Embodiment 1

[0070] The method for monitoring the fermentation process of Daqu provided by the embodiment 1 of the present application specifically comprises the following steps:

[0071] S1 collects samples in the fermentation process of Daqu, randomly collects a plurality of Daqu samples in any selected Daqu room, and records the fermentation days of the Daqu in the Daqu room, such as the 0th, 2nd, 4th, 6th, 8th, 10th, 13th, 16th, 19th and 21st days.

[0072] S2 extracts total DNA of the samples, uses the fluorescence quantitative PCR technology, respectively takes the bacterial 16S rRNA gene and the fungal 18S rRNA gene as targets, and uses specific primer pairs to quantitatively detect the copy numbers of the bacterial 16S rRNA gene and the fungal 18S rRNA gene. The pretreatment mode of the samples is completely consistent with the above-mentioned Daqu sample processing steps, and will not be described here.

[0073] S3 inputs the fermentation days, the copy numbers of the bacterial 16S rRNA gene and the fungal 18S rRNA gene into the above-mentioned random forest classification model which has been pre-trained, and loads the model file through the R language readRDS function;

[0074] S4 judges the fermentation state to be normal or abnormal based on the classification result output by the random forest classification model. Specifically, the predict function is executed to output the classification result, if the output result is “abnormal”, it is determined that the Daqu fermentation is abnormal, and the intervention measures such as adjusting the temperature and humidity of the Daqu room can be triggered accordingly; otherwise, if the output result is “normal”, it is determined that the Daqu is in a normal fermentation state, and the existing fermentation conditions of the Daqu room can be maintained.

[0075] Embodiment 2

[0076] Daqu is a saccharifying and fermenting agent for Luzhou-flavor liquor, and its fermentation state directly determines the product quality. Under the normal fermentation state, the number of bacteria (such as Bacillus) and fungi (such as Saccharomyces) is within a certain range (the number of the two cannot be too high or too low), and when the fermentation is abnormal (such as the number of bacteria and fungi is too high or too low), it will cause the acidification of fermented grains, abnormal pH, insufficient activity of fungi, and thus the conversion rate of ethanol is reduced and the liquor yield is reduced. Therefore, the embodiment of the present application also provides a method for predicting the product quality of Daqu fermentation.

[0077] The method is for predicting the quality of the fermentation product (such as Luzhou-flavor liquor) obtained by the fermentation of the Daqu on the basis of the embodiment 1. Specifically, the method for predicting the product quality of Daqu fermentation provided by the embodiment 2 of the present application comprises the following steps:

[0078] S1-S4 steps are the same as those of the embodiment 1.

[0079] S5: directly link the fermentation state output by the random forest classification model to the fermentation product quality of the Daqu. If the Daqu fermentation state is determined to be an abnormal state, it can be predicted that the quality of the obtained fermentation product has quality defects, and it is evaluated as "bad", "poor", "unqualified", "not up to standard", etc. If the Daqu fermentation state is determined to be a normal state, it can be predicted that the quality of the obtained fermentation product meets the quality standard, and it is evaluated as "good", "excellent", "qualified", "up to standard", etc.

[0080] Therefore, by using the method for monitoring the Daqu fermentation process provided in the embodiments of the present application, in actual detection, the obtained new sample can be used as a test set, and the existing random forest classification model is used for classification, the fermentation days, the bacterial 16S rRNA gene copy number and the fungal 18S rRNA gene copy number are input, and then the model outputs the fermentation state result (normal / abnormal), and then the fermentation product quality is predicted based on the fermentation state. The modeling and classification process is mainly realized by means of the functions in the R package, and non-professionals can easily operate after simple training. By determining the microbial gene copy number in the Daqu sample and using the model for classification, non-professionals can monitor the Daqu fermentation state in a simple, efficient and accurate manner, and then intervene in time to ensure the normal development of Daqu fermentation.

[0081] Although the present application has been disclosed with the preferred embodiments as above, it is not intended to limit the present application. Those skilled in the art without departing from the spirit and scope of the present application can make various modifications and improvements. Therefore, the protection scope of the present application shall be subject to the definition of the claims.

Claims

1. A method for monitoring the fermentation process of Daqu (a type of starter culture), characterized in that, Includes the following steps: S1 randomly collects Daqu samples during the fermentation process and records the number of fermentation days; S2 extracts total DNA from the sample and quantitatively detects the bacterial 16S rRNA gene copy number and fungal 18S rRNA gene copy number in the sample; S3 inputs the detected bacterial 16S rRNA gene copy number, the fungal 18S rRNA gene copy number, and the fermentation days into a pre-trained random forest classification model. S4 determines whether the fermentation state of Daqu is normal or abnormal based on the classification results output by the random forest classification model. The random forest classification model uses fermentation days, bacterial 16S rRNA gene copy number, and fungal 18S rRNA gene copy number as input features, and normal or abnormal fermentation status as the classification target.

2. The method for monitoring the fermentation process of Daqu (a type of starter culture) according to claim 1, characterized in that, In step S3, the construction of the random forest classification model includes the following steps: a. Establishing a bacterial quantitative standard curve: Amplify the bacterial 16S rRNA gene using the first primer pair 338F / 518R to generate a bacterial quantitative curve equation; b. Establish a standard curve for fungal quantification: Amplify the fungal 18S rRNA gene using the second primer pair FungiQuant-F / FungiQuant-R and the probe FungiQuant-Prb to generate a fungal quantification curve equation; c. Collect training samples in stages: Randomly collect multiple Daqu samples at preset sampling time points, record the fermentation days of the Daqu, and evaluate the fermentation status of the Daqu through sensory and physicochemical tests, including normal and abnormal states. d. Based on the bacterial quantitative curve equation and the fungal quantitative curve equation, detect the bacterial 16S rRNA gene copy number and the fungal 18S rRNA gene copy number of the Daqu sample; e. Using fermentation days, bacterial 16S rRNA gene copy number, and fungal 18S rRNA gene copy number as independent variables and fermentation state as the dependent variable, a random forest classification model is constructed.

3. The method for monitoring the fermentation process of Daqu (a type of starter culture) according to claim 2, characterized in that, In step a, the equation for the bacterial quantitative curve is Ct = -3.321logN + 49.

789.

4. The method for monitoring the fermentation process of Daqu (a type of starter culture) according to claim 2, characterized in that, In step b, the equation for the quantitative curve of fungi is Ct = -2.586logN + 35.

269.

5. The method for monitoring the fermentation process of Daqu (a type of starter culture) according to claim 2, characterized in that, In step c, the total number of samples collected at each sampling time point is at least 100, of which 80% are used as the training set and 20% are used as the test set.

6. The method for monitoring the fermentation process of Daqu (a type of starter culture) according to claim 5, characterized in that, The random forest classification model is constructed by performing at least 100 rounds of 5-fold cross-validation based on the training set. The discriminative ability of the random forest classification model was verified based on the test set. The verification metrics included AUC value, accuracy, sensitivity, and specificity.

7. The method for monitoring the fermentation process of Daqu (a type of starter culture) according to claim 2, characterized in that, The sampling time points are day 0, 2, 4, 6, 8, 10, 13, 16, 19, and 21 of fermentation, and no fewer than 10 spatially parallel samples are collected at each sampling time point.

8. A method for predicting the quality of Daqu fermentation products, characterized in that, The method for monitoring the fermentation process of Daqu as described in any one of claims 1-7, which directly correlates the fermentation status output by the random forest classification model with the quality of the fermented Daqu product, includes: If the fermentation state of the Daqu is normal, it is predicted that the fermented product meets the quality standards. If the fermentation state of the Daqu (a type of starter culture) is abnormal, it is predicted that the fermented product has quality defects.

Images