KR102537092B1 - Soil Pollution Source Prediction Method using Artificial Neural Network Model - Google Patents

Abstract

The present invention relates to a method of predicting a soil pollutant using an artificial neural network model or a computer readable recording medium storing a computer program for executing the prediction method. Using the soil contamination source prediction method of the present invention and a computer readable recording medium recording a computer program for executing the method, it is possible to analyze the contamination cause of contaminated soil easily, economically, and quickly, which will be widely applied in the field of environmental forensic science. can

Description

Soil Pollution Source Prediction Method using Artificial Neural Network Model}

The present invention relates to a method of predicting a soil pollutant using an artificial neural network model or a computer readable recording medium storing a computer program for executing the prediction method.

Hazardous chemicals widely used in industry pose a potential risk of environmental damage in the event of an accidental release. Harmful effects of accidental chemical spills can affect industrial workers, surrounding communities and the natural environment where industrial and residential areas are not clearly separated. For example, in 2012, approximately 8 tonnes of highly toxic hydrogen fluoride leaked from the Hube Global chemical plant in Gumi into the surrounding area. This incident resulted in the death of workers and the deterioration of the natural environment. In addition, acids such as hydrochloric acid and nitric acid are commonly used for chemical treatments such as etching and anodizing in electronic device manufacturing in Korea.

To determine the cause of a chemical accident, testing must be done quickly before the substance can diffuse or decompose. Once the contaminant source is identified, control measures to remove the target material can be precisely formulated and efficiently implemented. To this end, it is necessary to develop environmental forensic technology to track the source of contamination.

Environmental forensic technology is a technology that analyzes causative substances and intermediates produced by biodegradation or chemical reactions. A combination of chromatography, mass spectrometry and stable isotope techniques can be used for analysis. Depending on the analysis method, material analysis may be limited due to the purchase of expensive equipment, the risk of radioactive materials, the skill of human resources, and labor-intensive analysis. In addition, causative substances or intermediates may be lost as a result of evaporation, photoreaction, washing by rainfall, adsorption, biodegradation, and weathering. In addition, if the substance is decomposed into a substance with excessively low mass, most of the characteristics of the causative substance are lost, so the causative substance cannot be identified. In addition, because various reactions occurring in the environment vary according to time and place, the analysis of intermediates is complicated and uncertain.

When an acid/base leakage accident occurs, the acidity of the soil changes rapidly, and it can cause harm to the natural ecology and humans. Unfortunately, acids/bases dissociate quickly in a watery environment and are converted into cations and anions, and are easily freed from contamination by rainwater, surface water, and water generated during washing. In the case of hydrogen fluoride, complicated analysis procedures are required due to its fast reactivity and low boiling point.

Bacterial diversity is high in neutral samples and low in acidic/basic samples. This is because the types of microorganisms that naturally grow are limited under conditions of extremely low or high acidity. Soil pH is an important factor in organizing the bacterial community. The pH range most suitable for bacterial growth is 5.5-9.0. However, there are exceptions: acidophilic microorganisms grow optimally at about pH 3.0, and alkalophilic microorganisms grow optimally between pH 8 and 10.5. Each microbe has an optimal pH, but in an unsuitable pH environment, the activity of enzymes that support the growth of bacterial cells is hindered.

An artificial neural network is an information processing technology method similar to the information analysis activity of the human brain. The network structure includes input, hidden and output layers. Neurons are made up of hidden layers. The specified information of the input layer is passed to the hidden layer. Khan et al (2019) predicted positive or negative Mycobacterium tuberculosis infection with over 94% accuracy. Nikkonen et al. (2019) achieved greater than 90% accuracy in classifying patients for correcting obstructive sleep apnea conditions.

Therefore, there is a demand for a method for accurately, economically, and quickly predicting soil contamination sources in the event of an acid/base spill accident using bacterial diversity and artificial neural network technology.

Accordingly, the present inventors performed metagenome analysis or T-RFLP analysis on microbial DNA in soil samples while studying a method for predicting soil pollutants by combining bacterial diversity and artificial neural networks, and using the data to learn artificial neural networks. The present invention was completed by confirming that it is possible to predict soil contamination sources.

Accordingly, an object of the present invention is to provide a method for predicting a soil pollutant by learning microbial DNA information in a soil sample through an artificial neural network model.

In order to achieve the above object, the present invention includes (1) collecting a soil sample having soil contamination cause information; (2) extracting DNA from the soil sample; (3) PCR amplification of the extracted DNA using primers, and T-RFLP analysis of the PCR product to derive T-RF peak data; (4) training an artificial neural network model with T-RF peak data having soil contamination cause information in step (3); (5) extracting DNA from soil of unknown contamination, and PCR amplifying the extracted DNA using the primers of step (3); (6) T-RFLP analysis of the amplified PCR product to derive T-RF peak data; And (7) using the T-RF peak data of step (6) as an input variable and predicting the cause of soil contamination using the artificial neural network model of step (4); to provide.

In addition, the present invention comprises the steps of (1) collecting a soil sample having soil contamination cause information; (2) extracting DNA from the soil sample; (3) performing metagenome analysis on the extracted DNA and selecting operational taxonomic units of microorganisms present in the soil; (4) calculating relative dominance data of the microbial community in the soil sample using the operational classification unit selected in step (3); (5) training an artificial neural network model with the relative dominance data of step (4); (6) extracting DNA from soil of unknown contamination, performing metagenome analysis, and calculating relative dominance data in the soil sample using the operational taxonomic unit selected in step (3); And (7) using the relative dominance data in the soil sample in step (6) as an input variable and predicting the cause of soil contamination using the artificial neural network model in step (5); provides

In addition, the present invention provides a computer readable recording medium recording a computer program for executing the soil pollutant source prediction method.

Using the soil contamination source prediction method of the present invention and a computer readable recording medium recording a computer program for executing the method, it is possible to analyze the contamination cause of contaminated soil easily, economically, and quickly, which will be widely applied in the field of environmental forensic science. can

1 is a diagram showing a soil column containing a soil sample.
Figure 2 is a diagram showing the pH of the soil over time after acid and base leakage.
3 is a diagram showing the pH of soil column effluent over time after acid and base leakage.
Figure 4 is a diagram showing the change in pH by soil location according to acid and base contamination at the start of the experiment (day 0).
5 is a diagram showing the pH change by soil location according to acid and base contamination on the 23rd day after the start of the experiment.
6 is a diagram showing the yield and quality of DNA when DNA was extracted from a soil sample without a pretreatment step.
7 is a diagram showing the change in chromaticity of the supernatant according to the amount of 0.4 M aluminum chloride treatment.
8 is a diagram showing changes in DNA yield and quality when acidic soil samples are pretreated with aluminum chloride.
9 is a diagram showing the visualization of the 16S rRNA-based metagenome analysis results through the main component analysis chart.
10 is a diagram showing grouping by hierarchical dendrogram of 16S rRNA-based metagenome analysis results.
11 is a diagram showing the results of relative dominance data derived by confirming the microbial community structure according to acid and base treatment characteristics.
12 is a diagram showing the prediction rate (Accuracy %) of the artificial neural network model for each homology and dominance analysis criterion for OTU selection.
13 is a diagram showing the prediction rate of an artificial neural network model trained based on species-level relative dominance data to which indicator microorganisms are applied.
14 is a diagram showing the results of principal component analysis of T-RF peak patterns between soil samples using RStudio (Boston, MA, USA).
15 is a diagram showing the results of hierarchical clustering (hierarchical dendrogram) of T-RF peak patterns between soil samples using RStudio (Boston, MA, USA).
16 is a diagram showing the results of training an artificial neural network using data obtained by applying the indicator microorganism results to T-RF peak data and evaluating the prediction rate.

The present invention comprises the steps of (1) collecting a soil sample having soil contamination cause information; (2) extracting DNA from the soil sample; (3) PCR amplification of the extracted DNA using primers, and T-RFLP analysis of the PCR product to derive T-RF peak data; (4) training an artificial neural network model with T-RF peak data having soil contamination cause information in step (3); (5) extracting DNA from soil of unknown contamination, and PCR amplifying the extracted DNA using the primers of step (3); (6) T-RFLP analysis of the amplified PCR product to derive T-RF peak data; And (7) using the T-RF peak data of step (6) as an input variable and predicting the cause of soil contamination using the artificial neural network model of step (4); to provide.

In the present invention, the T-RFLP (Terminal Restriction Fragment Length Polymorphism) is obtained by amplifying and securing 16S rRNA (bacteria) or 18S rRNA (fungus) gene information from an environmental DNA sample by polymerase chain reaction (PCR), followed by restriction enzyme digestion It is a method of obtaining fingerprint information of the microbial community through restriction enzyme digestion. T-RFLP can quickly compare differences in microbial DNA sequences amplified by PCR from environmental samples and has the advantage of being able to analyze a large amount of samples at once (Dunbar et al., 2001). However, there are disadvantages in that accurate information about species cannot be known, and resolution is somewhat low when analyzing species diversity.

In the present invention, the step of extracting DNA in step (2) may further include a pretreatment step of removing substances that interfere with DNA extraction from the soil sample, and a pretreatment step of removing substances that interfere with DNA extraction from the soil sample. When the cause of soil contamination is acidic, it may be a step of pre-treating the soil sample with aluminum chloride (AlCl ₃ ), preferably 0.2 to 0.6 M aluminum chloride (AlCl ₃ ) 466.7 to 866.7 μL per 1 g of the soil sample. It may be a pre-treatment step, and most preferably, it may be a step of pre-treating 666.7 μL of 0.4 M aluminum chloride (AlCl ₃ ) per 1 g of soil sample. By further including the step of pre-treating the acidic soil sample, the quality of the DNA is improved during DNA extraction, so that DNA suitable for PCR can be extracted.

In the present invention, 'primer' refers to template-directed DNA synthesis under suitable conditions (eg, four different nucleoside triphosphates and a polymerizing agent such as DNA, RNA polymerase or reverse transcriptase) in an appropriate buffer solution and an appropriate temperature. It refers to a single-stranded oligonucleotide that can serve as a starting point for The appropriate length of the primer may vary depending on the purpose of use, but is usually 15 to 30 nucleotides, preferably 15 to 25 nucleotides, but is not limited thereto.

In the present invention, the primer in step (3) may be a forward primer of SEQ ID NO: 1 or a reverse primer of SEQ ID NO: 2, preferably a primer of SEQ ID NO: 1. In addition, the primers in step (3) are 70% or more, more preferably 80% or more, still more preferably 90% or more, and most preferably 95% of each nucleotide sequence of any one of SEQ ID NOs: 1 and 2. It may be a nucleotide sequence having the above sequence homology.

In the present invention, the PCR amplification in step (3) may be to amplify the 16s rRNA gene of microorganisms in the soil.

In the present invention, the step of training the artificial neural network model in step (4) may be a step of setting the number of hidden layer neurons in the range of 1 to 151, and preferably, the primer of step (3) is a forward primer of SEQ ID NO: 1 And, if the hidden layer neurons in the step of training the artificial neural network model in step (4) are 143, the soil pollutant may be predicted with a high probability of 78.8%.

In the present invention, the soil contaminant may be included without limitation as long as it is a substance that causes change in soil, preferably any one selected from the group consisting of no pollution, ammonia, hydrochloric acid, sulfuric acid, hydrogen fluoride and nitric acid. can be

In addition, the present invention comprises the steps of (1) collecting a soil sample having soil contamination cause information; (2) extracting DNA from the soil sample; (3) performing metagenome analysis on the extracted DNA and selecting operational taxonomic units of microorganisms present in the soil; (4) calculating relative dominance data of the microbial community in the soil sample using the operational classification unit selected in step (3); (5) training an artificial neural network model with the relative dominance data of step (4); (6) extracting DNA from soil of unknown contamination, performing metagenome analysis, and calculating relative dominance data in the soil sample using the operational taxonomic unit selected in step (3); And (7) using the relative dominance data in the soil sample in step (6) as an input variable and predicting the cause of soil contamination using the artificial neural network model in step (5); provides

As used in the present invention, the term "metagenome" is also referred to as "group genome" and refers to the sum of genomes including all viruses, bacteria, fungi, etc. in an isolated area such as soil or animal intestine, It is used as a concept of genome to explain the identification of many microorganisms at once using a sequencer to analyze microorganisms that cannot be cultured. In particular, the metagenome does not refer to the genome or genome of one species, but to a kind of mixed genome as the genome of all species in one environmental unit. This is a term derived from the viewpoint that when defining a species in the process of biological development omically, not only one functionally existing species, but also various species interact with each other to create a complete species. Technically, it is the subject of a technique that uses rapid sequencing to analyze all DNA and RNA regardless of species, identify all species in an environment, and identify interactions and metabolism. In the present invention, metagenome analysis was preferably performed using DNA of microorganisms present in soil.

In the present invention, the step of extracting the DNA in step (2) may further include pre-treating the soil sample with aluminum chloride (AlCl ₃ ) when the source of soil contamination is acidic, preferably It may further include pre-treating 466.7 to 866.7 μL of 0.2 to 0.6 M aluminum chloride (AlCl ₃ ) per 1 g of soil sample, most preferably 0.4 M aluminum chloride (AlCl ₃ ) per 1 g of soil sample. It may further include the step of pre-processing 666.7 μL. By further including the step of pre-treating the acidic soil sample, the quality of the DNA is improved during DNA extraction, so that DNA suitable for PCR can be extracted.

In the present invention, the metagenome analysis in step (3) may be to analyze part or all of the extracted DNA, preferably targeting the 16S rRNA gene V3 and / or V4 region of the microbial DNA. there is.

In the present invention, the step of selecting an operational taxonomic unit in step (3) is performed by metagenome analysis of DNA extracted from a soil sample to analyze the nucleotide sequence of various microorganisms present in the soil, and several analyzed Grouping microbial nucleotide sequences at a level of homology set by a user, and selecting an arbitrary microbial nucleotide sequence in an individual group among the grouped microbial sequences as an operational taxonomic unit. Preferably, in the present invention, the step of selecting the operational taxonomic unit may be a step of selecting the homology between nucleotide sequences at a homology level of 95 to 100%.

In the present invention, the relative dominance data of the microbial community in step (4) is Kingdom, Phylum, Class, Order, Family, Genus or Species ( It may be calculated based on Species, and preferably it may be calculated based on Family, Genus, or Species.

In the present invention, the step (3) selects an operational taxonomic unit with a homology level of 99%, and the relative dominance data of the microbial community in the step (4) may be calculated based on species, , in this case, the soil contamination source may be predicted with a high probability of 80.8%.

In the present invention, the relative dominance data of the microbial community may be relative dominance data of a contaminant-specific indicator microorganism, and preferably, the indicator microorganism is Angustibacter luteus, Runella slithyformis, Cystobacter velatus, Geodermatophilus daqingensis, At least one selected from the group consisting of Pseudonocardia soli, Aquabacterium commune , and Skermanella rubra , When the contaminant is ammonia, at least one selected from the group consisting of Nitrosospira tenuis, Flavihumibacter sediminis, Flavobacterium saliperosum, Brevundimonas naejangsanensis, Actinotalea ferrariae, Nocardioides caricicola, Chryseobacterium montanum, Flavobacterium dankookense, and Pedobacter bauzanensis , and when the contaminant is hydrochloric acid , Mucilaginibacter polysacchareus, Dyella japonica, Actinocorallia aurantiaca, Pedobacter kyungheensi, Flavobacterium resistens, Pedobacter cryoconitis, and Flavobacterium spartansii , and at least one selected from the group consisting of, and the contaminant is In the case of nitric acid, at least one selected from the group consisting of Methylophilus rhizosphaerae, Pseudomonas silesiensis, Ralstonia pickettii, and Cupriavidus campinensis , and in the case of hydrogen fluoride, Bacillus marisflavi, Paenibacillus aceris, Paraburkholderia hospital, Halobacillus profundi, Clostridium puniceum, Bacillus plakortidis, Bacillus aryabhattai , Paenibacillus populi, Rhodanobacter xiangquanii, and Bordetella flabilis , and when the contaminant is sulfuric acid, it may be at least one selected from the group consisting of Micromonospora noduli, Actinospica acidiphila, Desulfohalotomaculum peckii, and Tumebacillus ginsengisoli . When the relative dominance data of the microbial community is used as the relative dominance data of the pollutant-specific indicator microorganisms, noise data is reduced, thereby increasing the training progress speed and increasing the prediction rate.

In the present invention, the step (3) selects an operational taxonomic unit with a homology level of 99%, and the relative dominance data of the microbial community in the step (4) is calculated based on species, and the ( The relative dominance data of the microbial community in step 4) is the relative dominance data of the pollutant-specific indicator microorganisms, and in the step of training the artificial neural network model in step (5), when the hidden layer neurons are set to 7, the soil pollutant is 85.0% It may be predictable with a high probability of

In addition, the present invention provides a computer readable recording medium recording a computer program for executing the soil pollutant source prediction method.

The method according to the present invention may be implemented in hardware, firmware, or software or a combination thereof. When implemented as software, the storage medium includes a storage medium in a form readable by a device such as a computer. For example, a computer readable medium may include read only memory (ROM); random access memory (RAM); magnetic disk storage media; It includes an optical storage medium and a flash memory device.

Redundant content is omitted in consideration of the complexity of the present specification, and terms not otherwise defined in the present specification have meanings commonly used in the technical field to which the present invention belongs.

Hereinafter, the present invention will be described in detail by examples. However, the following examples are merely illustrative of the present invention, and the content of the present invention is not limited by the following examples.

Example 1. Analysis of Soil Changes According to Acid/Base Contaminants

1.1 Preparation of laboratory-scale acid-alkaline exposed soil columns

Farm soil was collected from a garden located in Busan at 35° 14' 38.4" N, 129° 03' 23.4" E. A 21 cm long soil column was prepared using a total of 6 plastic cylinders with an inner diameter of 5 cm. At the beginning of the experiment (day 0), hydrochloric acid (1 N HCl), nitric acid (1 N HNO3), hydrogen fluoride (1 N HF), sulfuric acid (1 N H2SO4), and ammonia (0.01 N NH3) were added to the soil column by volume of soil. was administered in the same way as Tap water was administered as a control group. To simulate the effect of natural rainfall, a volume of tap water equal to that of soil was administered at the top of the column from day 1 to day 6. Next, a drying period was applied. Specifically, after drying the column for 5 days, tap water was added on the 6th day. The rain-drying process was repeated twice. Soil samples were collected from three locations: top (T), middle (M), and bottom (B) at the start of the experiment (day 0) and on day 23 after the start of the experiment, and on other days, samples were taken only at the top of the column. A 21 cm long soil column containing the soil sample is shown in FIG. 1 .

1.2 Analysis of pH change by period after acid and base contamination

Using the acid and base exposed soil column prepared in Example 1.1, the pH change according to acid and base contamination was analyzed. The pH of soil over time after acid and base leakage is shown in FIG. 2 , and the pH of soil column effluent over time after acid and base leakage is shown in FIG. 3 .

As shown in Figure 2, the pH of the soil before the experiment was 7.3, and this value rapidly decreased to less than 4.7 when the acid solution was introduced into the column. The lowest soil pH value was found in the sulfuric acid column and was about 1.8. At the beginning of the reaction, pH values of about 1.9, 2.0, and 2.5 were observed in the nitric acid, hydrochloric acid, and hydrogen fluoride columns, respectively. sulfuric acid, nitric acid, hydrochloric acid 4.1, 5.6 on day 6; It showed pH values of 5.7 and 6.3. Compared to the control column soil's pH of 7.7, the sulfuric acid column showed the lowest soil pH of 4.1 at the end of the experiment, with sulfuric acid having the most adverse effect on soil pH among the four types of acids tested. It was confirmed that the soil leaking ammonia rapidly stabilized from pH 8.9 to pH 7.5-8.0 after 2 days of the experiment.

As shown in Figure 3, the pH trend of the soil column effluent showed a similar trend to the soil pH.

1.3 Analysis of pH change by location of acid and base contaminated soil

Using the acid and base exposed soil column prepared in Example 1.1, the pH change by soil location according to acid and base contamination was analyzed. The soil at the top of the column is most directly affected by the leakage of acids and bases, and the pH changes due to the pH buffering effect of the soil as the acids and bases descend in the vertical direction of the column. The pH change by soil location due to acid and base contamination at the start of the experiment (day 0) is shown in FIG. 4, and the pH change by soil location due to acid and base contamination 23 days after the start of the experiment is shown in FIG. .

As shown in Figure 4, at the start of the experiment, the soil pH was the lowest in the top (T) soil, and it was confirmed that the pH slightly increased in the middle (M) soil and bottom (B) soil.

As shown in Figure 5, after the end of the experiment on the 23rd, the pH of the soil showed little change depending on the location. This reflects the change in soil pH trend by the neutralizing activity of microorganisms.

Example 2. Establishment of a Method for Extracting Microbial DNA in Soil for Prediction of Acid and Base Contaminants

2.1 Soil sample preparation according to the source of contamination

Eighty-four soil samples contaminated by various acid and base contaminants were collected. The pH of the soil samples was measured immediately after collection. For DNA extraction, soil samples were stored in the dark at 4 °C until use. Of the 84 soil samples, 28 soil samples corresponded to samples collected from control and ammonia leaking soils. Soil samples obtained from the hydrochloric acid, nitric acid, sulfuric acid, hydrogen fluoride leaky soil column had acidic pH values ranging from 1.85 to 6.35. Soil sample information is shown in Table 1.

2.2 Comparison of yield and quality of DNA samples extracted without pretreatment

Control, ammonia soil samples, and acidic soil samples (hydrochloric acid, nitric acid, hydrogen fluoride, sulfuric acid) were DNA extracted using FastDNA® Spin Kit for Soil (MP Biomedical, USA) without any pretreatment step. The procedure was performed according to the instructions provided by the manufacturer. The DNA concentration and quality (A260/A280) of the extracted DNA were measured using a NanoDrop ND-1000 UV-Vis spectrophotometer (NanoDrop, Wilmington, DE, USA), and are shown in FIG. 6 .

As shown in Figure 6, in the absence of separate pretreatment, DNA from the acidic soil showed low concentrations between 0.8 and 20.9 ng/μL, while DNA from the control and ammonia soil samples showed higher concentrations (>33.4 ng/μL). ) was shown. DNA quality (A260/A280) was the most stable in the control soil column sample, and the ammonia column soil sample showed a wide range of variation, but showed a level suitable for PCR on average. However, the quality of DNA in acidic soil was found to be rather low, and a method for extracting and purifying the quality of DNA in acidic soil to a level suitable for PCR was required.

2.3 Establishment of a pretreatment method for acidic soil samples

2.3.1 Selection of aluminum chloride pretreatment concentration

Since acidic soil adversely affects DNA extraction, an aluminum chloride (AlCl ₃ ) pretreatment step was constructed to improve the quality of DNA extraction from soil. Cationic metals such as aluminum (Al ³⁺ ) can scavenge humic acid due to their antagonistic charge. Indeed, Al ³⁺ can bind to and aggregate both humic acid and DNA. Therefore, since excess or insufficient aluminum chloride can lead to loss of DNA yield, an experiment was conducted to calculate the optimal dosage for extracting DNA from acidic pH soil. Specifically, for 56 acidic soil samples (pH < 6.5) per 0.3 g of soil sample, 100, 200, 400, 500, and 600 μL of 0.4 M aluminum chloride solution were treated, and the supernatant by 0.4 M aluminum chloride treatment Chromaticity change is shown in FIG. 7 .

As shown in FIG. 7, it was confirmed that a clear brown supernatant was produced under the condition of 200 μL or more of aluminum chloride per 0.3 g of soil sample. Therefore, 666.7 μL of 0.4 M aluminum chloride per gram of soil sample was selected as the treatment concentration for acidic soil pretreatment.

2.3.2. Measurement of DNA Yield and Quality Changes in Acidic Soil Samples with Aluminum Chloride Pretreatment

Experiments were conducted to measure DNA yield and quality changes by aluminum chloride pretreatment. Acid soil samples pretreated with 666.7 μL of 0.4 M aluminum chloride per gram of soil sample were DNA extracted using FastDNA® Spin Kit for Soil (MP Biomedical, USA). The procedure was performed according to the instructions provided by the manufacturer. The DNA concentration and quality (A260/A280) of the extracted DNA were measured using a NanoDrop ND-1000 UV-Vis spectrophotometer (NanoDrop, Wilmington, DE, USA), and are shown in FIG. 8 .

As shown in FIG. 8, the DNA yield by aluminum chloride pretreatment was 11.8 to 14.0 ng/uL, significantly lower than before pretreatment, but there was no problem in performing PCR. The quality of the aluminum chloride pre-treated sample ranged from 1.7 to 2.1, which was confirmed to be close to 1.8, which is known as the optimal value.

Combining these results, we completed the microbial DNA extraction method in soil for the prediction of the following acid and base contaminants. The control and ammonia soil samples did not undergo a separate pretreatment step, and in the case of acidic soil samples, a step of pretreatment with 666.7 μL of 0.4 M aluminum chloride per 1 g of soil sample was performed. Then, DNA was extracted using the FastDNA® Spin Kit for Soil (MP Biomedical, USA), and the extracted DNA was purified and concentrated using the Dneasy Power clean Pro Cleanup Kit (Qiagen, USA) to perform DNA PCR. A sample was prepared.

Example 3. Establishment of method for predicting acid and base contaminants using metagenome analysis

3.1 Calculation of microbial relative dominance data to predict acid and base contamination sources

The bacterial 16S rRNA gene V3 and V4 regions were sequenced on the Illumina MiSeq platform (Illumina, San Diego, CA, USA) for DNA samples extracted by the DNA extraction method of Example 2 from 36 soil samples from laboratory-scale exposure experiments. The sequence was analyzed (Macrogen, Korea). The remaining high-quality sequences from the non-chimeric clusters were selected as Operational Taxonomic Units (OTUs) within 97% homology using a greedy algorithm. Information from phylum to species level was searched by RDP (Ribosomal Database Project) using QIIME (Quantitative Insight Into Microbial Ecology) pipeline. The microbial diversity index (chao1, Shannon-Wiener, Simpson) was calculated. A 16S rRNA gene library of soil bacterial communities was constructed by Illumina MiSeq sequencing, and a total of 5,074,018 sequences were identified from 36 DNA samples. Primer information used in Illumina MiSeq sequencing is shown in Table 2, and the results of sequencing of DNA samples are shown in Table 3. The visualization of the 16S rRNA-based metagenome analysis results through a principal component analysis chart is shown in FIG. 9, and the hierarchical dendrogram grouping of 36 samples is shown in FIG. The relative dominance data results obtained by confirming the microbial community structure according to the method are shown in FIG. 11 .

sequence number Base sequence (5' -> 3') 27F (Forward) SEQ ID NO: 3 TCGTCGGCAGCGTCAGATGTGTATAAGAGACAGCCTACGGGNGGCWGCA 518R (reverse) SEQ ID NO: 4 GTCTCGTGGGCTCGGAGATGTGTATAAGAGACAGGACTACHVGGGTATCTAATCC

As shown in Figure 9, overall, the control and ammonia samples were found to be clearly separated from the acid samples.

As shown in FIG. 10, as a result of hierarchical clustering of 36 soil samples, Group 1 (control and ammonia samples) and Group 2 (acidic soil samples) were generated. Group 1 was clearly separated from group 2, confirming that the soil bacterial community structure changed due to the decrease in soil pH. Group 2 was classified into three subgroups as follows: subgroup 1 (D0.NT, D0.NM, D0.ST, D0.FT, D4.FT), subgroup 2 (D0.Cl.T, D0). .Cl.M, D0.FM, D0.SM), subgroup 3 (D4.Cl.T, D4.NT, D4.ST, D11.Cl.T, D11.NT, D11.FT, D11.ST, D23.Cl.T, D23.Cl.M, D23.NT, D23.NM, D23.FT, D23.FM, D23.ST, D23.SM). These results confirm that the unique bacterial community in the soil changes according to the distance and contamination type in terms of space and time as leakage accidents occur.

As shown in FIG. 11, Proteobacteria was confirmed as the most abundant phylum in the relative dominance map at the phylum level, and the ratio was found to be the most abundant in almost all ammonia column soil samples and acid-treated soil samples (23 days). It was confirmed that it occupied more than 39.5% of It was confirmed that Proteobacteria were present in the range of 39.5 to 46.7% in the control and ammonia samples. Subsequently, 19.9 to 23.0% of Actinobacteria, 5.8 to 9.3% of Acidobacteria, and 7.2 to 13.2% of Bacteroidetes were found to dominate the ammonia soil column. It was confirmed that Firmicutes were dominant at 35.1 ~ 94.8% in almost initial acidic samples (0, 4 days) with a pH value lower than 4.7. They are widely distributed in the soil environment and have a major impact on agriculture. As a result of the relative dominance confirmed at the class level, within Proteobacteria, Alphaproteobacteria accounted for 16.8-19.8%, confirming that they were the most abundant, followed by 9.5-17.4% of Gammaproteobacteria, 8.2-11.0% of Betaproteobacteria, and 3.1-4.1% of It was confirmed that the order of Deltaproteobacteria was dominant.

3.2 Selection of indicator microorganisms by acid and base contamination characteristics

Based on the "multipatt (function "multipat", nperm 1000, p≤0.01)" function of the "indicspecies" package provided by CRAN (Comprehensive R Archive Network), indicator microorganisms were selected according to acid and base contaminants. Dominance data according to the metagenome results calculated in Example 3.1 was used as input data. Statistically significant species (p<0.01) were selected. IndVal is an index value representing the relationship between acid/base and microbial species. The calculation of IndVal is defined by Dufrene and Legendre (1997), and the highest IndVa value indicates the strongest correlation. Statistical significance values were used to judge the reliability of this method (Dufrene et al., 1997). A total of 41 species of indicator microorganisms according to the contamination source were selected by the above method, and these are shown in Table 4.

As shown in Table 4, Nitrosospira tenuis was selected as the most dominant ammonia indicator microorganism (p < 0.001), and in addition to Nitrosospira tenuis , 8 indicators related to ammonia pollutants were Flavihumibacter sediminis, Flavobacterium saliperosum, Brevundimonas naejangsanensis, Actinotalea ferrariae, and Nocardioides. caricicola, Chryseobacterium montanum, and Flavobacterium dankookbacbau were selected.

In addition, four types of acid contamination (hydrochloric acid, nitric acid, hydrogen fluoride, sulfuric acid) and corresponding indicator microorganisms were identified. Seven of the 10 hydrogen fluoride indicator microorganisms were identified as belonging to the Bacilli class (p≤0.01). In the case of nitric acid , Pseudomonas silesiensis was identified as the most dominant indicator microorganism (dominance rate 30.54%, p<0.001). For hydrochloric acid, seven microorganisms, including Dyella japonica, were selected as indicator microorganisms. Sulfuric acid selected four microorganisms, including Micromonospora noduli, as indicator microorganisms.

3.3 Artificial Neural Network Modeling for Contamination Source Prediction Using Relative Dominance Data

In the method of calculating relative dominance data through metagenome analysis of Example 3.1, relative dominance data was calculated by varying the standard homology, and based on this, the artificial neural network was trained and tested. The contaminants of each soil sample were set to six grades (control, ammonia, hydrochloric acid, nitric acid, hydrogen fluoride, and sulfuric acid), and the accuracy of the prediction of the neural network was verified. For optimization of the predictive model, the number of hidden layer neural networks was applied in the range of 1 to 151, respectively. In addition, in order to mitigate the model overfitting problem, the principle of 10-fold cross-validation was applied, and prediction was performed by alternating training data and verification data within the data group. In each round, the input data was randomly split into a training set (90%) and a test set (10%) to test the effectiveness of the neural network. The overall accuracy of the network is the average accuracy over 10 rounds of cross-validation. The software used MATLAB, and the MATLAB package (2019b, MathWorks, Natick, MA) was used together as a neural network toolbox.

In order to improve the artificial neural network prediction rate, the artificial neural network model was trained and evaluated with different reference homology for OTU calculation, and the prediction rate (Accuracy %) of the artificial neural network model is shown in FIG. 12.

As shown in FIG. 12, when the reference homology is set to 99, 97, 95, 93, and 91%, the prediction rates of 70.0%, 72.5%, and 80.8% are the highest in the number of hidden layer neurons of 59, 31, and 22, respectively. was In particular, it was confirmed that the prediction rate was lowered when the level of homology used in OTU selection was lowered.

3.4 Artificial Neural Network Modeling for Pollution Source Prediction Using Indicator Microbial Species

The indicator microorganisms derived in Example 3.2 were applied to the relative dominance data at the species level, and as a result, the biomarkers of the dominance data were reduced from 403 species to 41 species. Since this method reduces noisy data, it can speed up the training process and increase the prediction rate. The artificial neural network model was trained by varying the hidden layer neurons using the relative dominance data to which the indicator microorganism was applied or the relative dominance data to which the indicator microorganism was not applied, and the prediction rate was evaluated. The prediction rate evaluation results are shown in FIG. 13 .

As shown in FIG. 13, the prediction rate of the artificial neural network model that showed the highest accuracy in the species-level relative dominance data to which the indicator microorganism was applied was 85.05% when the number of neurons in the hidden layer was 7. It was confirmed that the prediction rate increased by 4.2% compared to the highest prediction rate of the model trained with the data.

Example 4. Establishment of Acid and Base Contamination Source Prediction Method Using T-RFLP (Terminal Restriction Fragment Length Polymorphism)

4.1 T-RF (Terminal Restriction Fragment) Peak Analysis

For the amplification of the 16S rRNA gene, PCR was performed on the DNA sample of the soil sample prepared by the method of Example 2 using the primer pair 27F and 518R. In order to use the PCR product for T-RFLP, forward primer 27F and reverse primer 518R were labeled for FAM and HEX fluorescence, respectively. For PCR, prepare a total of 50 μL of PCR sample consisting of 25 μL of DreamTaq DNA polymerase (Thermo Fisher, USA), 1 μL of each primer (10 μM), 2 μL of DNA template and 21 μL of deionized water. did The PCR operating conditions were as follows: 1 cycle at 93 °C for 3 min, 30 cycles at 93 °C for 30 sec, 60 °C for 30 sec, 72 °C for 30 sec, and finally 1 cycle at 72 °C for 10 min. For this purpose, a SimpliAmp Thermal Cycler (Thermo Fisher, USA) was used. PCR products were purified using the QIAquick PCR purification kit (Qiagen, USA) and fragmented with the restriction enzyme BsuR I (10 U) (Thermo Fisher, USA) for 3 hours at 37°C. T-RFLP analysis was performed by SolGent (Solution for Genetic Technology) in Korea. The size and relative density of the distal presegment were analyzed using Peak Scanner 2.0 software (Applied Biosystems, USA). Peaks of less than 50 bp were considered errors. T-RF peak data was finally derived by aligning the distal fragment (T-RF) through the T-Align web-based program and generating a profile of each sample. Sequence information of primer pairs for amplification of the 16S rRNA gene is shown in Table 5.

sequence number base sequence (5' -> 3') 27F (Forward) SEQ ID NO: 1 AGAGTTTGATCMTGGCTCAG 518R (reverse) SEQ ID NO: 2 ATTACCGCGGCTGCTGG

4.2 Verification of the feasibility of using T-RF (Terminal Restriction Fragment) peak as a pollutant source prediction indicator

It was confirmed whether the T-RF peak data derived in Example 4.1 could be used as a pollutant source prediction indicator. Specifically, to confirm the similarity of T-RF peak patterns among soil samples, principal component analysis and hierarchical dendrogram were performed using RStudio (Boston, MA, USA). Component analysis results are shown in FIG. 14 , and hierarchical clustering results are shown in FIG. 15 .

As shown in FIG. 14, as a result of main component analysis, grouping according to the contamination source for each soil sample was not clearly made.

As shown in FIG. 15, it was confirmed that the soil microbial community in the soil sample could be classified according to the type of contaminant, and the T-RF peak could be used as a biomarker for identifying the causative agent.

4.3 Pollution Source Prediction Artificial Neural Network Modeling Using T-RF Peak Data

Using the T-FR peak data derived from the 27F forward primer and the T-FR peak data derived from the 518R reverse primer of Example 4.1, the artificial neural network was trained in the same way as the artificial neural network modeling method for predicting contamination in Example 3.3, and the prediction rate was evaluated. In addition, the artificial neural network was trained using the data obtained by applying the T-RF result of the indicator assumed as the indicator microbial species to the T-RF peak data. Specifically, as the application data, starting from the relative dominance of each T-RF calculated using the T-Align program, the indicator T-RF was selected through the Indicspecies program among all T-RFs. That is, the indicator T-RF can be assumed to be an indicator microbial species. The relative dominance of the selected index T-RF was input into the artificial neural network model. The predictive rate evaluation results according to this are shown in FIG. 16 .

As shown in FIG. 16, as a result of applying the number of neurons from 1 to 151, the prediction rate was 19.2 to 24.2% in the range of 1 to 10, and the prediction rate was 67.4 to 77.4% in the range of more. The highest prediction rate of 78.8% was confirmed in the artificial neural network model trained with 143 hidden layer neurons using T-RF peak data derived from forward primers.

Example 5. Comparison of acid and base contamination source prediction methods using metagenome analysis and acid and base contamination source prediction methods using T-RFLP

Contamination source prediction artificial neural network model using T-RF peak data completed in Example 4.3 (T-RF peak data calculated with forward primer applied, 143 hidden layers, indicator T-RF not applied) and 85.05% prediction rate completed in Example 3.4 The specificity and sensitivity of an artificial neural network model (7 hidden layers, indicator microbial data applied) for predicting acid and base contaminants using metagenome analysis with . Sensitivity was defined as the ratio of true positive results divided by the number of total positive results, and specificity was defined as the ratio of true negative results divided by the number of total negative results. The specificity and sensitivity analysis results are shown in Table 6.

As shown in Table 6, it was confirmed that both the acid and base contaminant prediction artificial neural network model using metagenome analysis and the contaminant prediction artificial neural network model using T-RF peak data had excellent specificity and sensitivity.

In the above, specific parts of the present invention have been described in detail, and for those skilled in the art, it is clear that these specific descriptions are only preferred embodiments, and the scope of the present invention is not limited thereby. something to do. Accordingly, the substantial scope of the present invention will be defined by the appended claims and their equivalents.

<110> Pusan National University Industry-University Cooperation Foundation <120> Soil Pollution Source Prediction Method using Artificial Neural Network Model <130> pusan1-407P <160> 4 <170> KoPatentIn 3.0 <210> 1 <211> 20 <212> DNA <213> artificial sequence <220> <223> T-RFLP forward primer <400> 1 agagtttgat cmtggctcag 20 <210> 2 <211> 17 <212> DNA <213> artificial sequence <220> <223> T-RFLP reverse primer <400> 2 attaccgcgg ctgctgg 17 <210> 3 <211> 49 <212> DNA <213> artificial sequence <220> <223> metagenome forward primer <400> 3 tcgtcggcag cgtcagatgt gtataagaga cagcctacgg gnggcwgca 49 <210> 4 <211> 55 <212> DNA <213> artificial sequence <220> <223> metagenome reverse primer <400> 4 gtctcgtggg ctcggagatg tgtataagag acaggactac hvgggtatct aatcc 55

Claims (23)

(1) collecting soil samples with acid or base contamination information;
(2) extracting microbial flora DNA from the soil sample;
(3) PCR amplification of the extracted DNA using primers, and T-RFLP analysis of the PCR product to derive T-RF peak data;
(4) Training in the range of 1-151 hidden layer neurons by using the T-RF peak data in step (3) and the acid or base contamination information of the soil sample in step (1) as inputs and outputs of the artificial neural network model step;
(5) extracting microbial flora DNA from soil of unknown contamination, and PCR amplifying the extracted DNA using the primers of step (3);
(6) T-RFLP analysis of the amplified PCR product to derive T-RF peak data; and
(7) using the T-RF peak data in step (6) as an input variable of the artificial neural network model trained in step (4) to predict the contamination cause of the soil whose contamination source is unknown; Containing, soil contaminant prediction method.
The method of claim 1, wherein the step of extracting the DNA in step (2) further comprises a pretreatment step of removing substances that interfere with DNA extraction from the soil sample.
The method of claim 2, wherein the pretreatment step of removing DNA extraction-interfering substances from the soil sample is a step of pre-treating the soil sample with aluminum chloride (AlCl ₃ ) when the soil contamination source is acidic.
The method of claim 3, wherein the step of pre-treating the soil sample with aluminum chloride (AlCl ₃ ) is a step of pre-treating 466.7 to 866.7 μL of 0.2 to 0.6 M aluminum chloride (AlCl ₃ ) per 1 g of the soil sample. method.
The method of claim 1, wherein the primer in step (3) is a forward primer of SEQ ID NO: 1 or a reverse primer of SEQ ID NO: 2.
The method of claim 1, wherein the primer in step (3) is a forward primer of SEQ ID NO: 1.
The method of claim 1, wherein the PCR amplification in step (3) is to amplify the 16s rRNA gene.
delete The soil pollutant prediction method according to claim 1, wherein the primer in step (3) is a forward primer of SEQ ID NO: 1, and the step of training the artificial neural network model in step (4) is a step of setting hidden layer neurons to 143.
The method of claim 1, wherein the soil pollutant is any one selected from the group consisting of no pollution, ammonia, hydrochloric acid, sulfuric acid, hydrogen fluoride, and nitric acid.
(1) collecting soil samples with acid or base contamination information;
(2) extracting microbial flora DNA from the soil sample;
(3) performing metagenome analysis on the extracted DNA and selecting operational taxonomic units of microorganisms present in the soil;
(4) calculating relative dominance data of the microbial community in the soil sample using the operational classification unit selected in step (3);
(5) training in the range of 1-151 hidden layer neurons by using the relative dominance data in step (4) and the acid or base contamination information of the soil sample in step (1) as inputs and outputs of the artificial neural network model;
(6) extracting the DNA of the microbial flora from soil of unknown contamination, performing metagenome analysis, and calculating relative dominance data in the soil sample using the operational taxonomic unit selected in step (3); and
(7) using the relative dominance data in the soil sample in step (6) as an input variable of the trained artificial neural network model in step (5) to predict the contamination cause of the soil whose contamination source is unknown; , Methods for predicting soil pollutant sources.
The method of claim 11, wherein the step of extracting the DNA in step (2) further comprises a pretreatment step of removing substances that interfere with DNA extraction from the soil sample.
13. The method of claim 12, wherein the pretreatment step of removing DNA extraction interference substances from the soil sample is a step of pretreating the soil sample with aluminum chloride (AlCl ₃ ) when the soil contamination source is acidic.
The method of claim 13, wherein the step of pre-treating the soil sample with aluminum chloride (AlCl ₃ ) is a step of pre-treating 466.7 to 866.7 μL of 0.2 to 0.6 M aluminum chloride (AlCl ₃ ) per 1 g of the soil sample. method.
The method of claim 11, wherein the metagenome analysis in step (3) targets the 16S rRNA gene V3 and V4 regions of the extracted DNA.
The method of claim 11, wherein step (3) is a step of selecting an operational classification unit at a homology level of 95 to 100%.
The method of claim 11, wherein the relative dominance data of the microbial community in step (4) is calculated based on family, genus, or species.
The method of claim 11, wherein step (3) selects an operational taxonomic unit with a homology level of 99%, and the relative dominance data of the microbial community in step (4) is calculated based on species. , Methods for predicting soil pollutant sources.
The method according to any one of claims 11 to 18, wherein the relative dominance data of the microbial community in step (4) is the relative dominance data of a contaminant-specific indicator microorganism.
The method of any one of claims 11 to 18, wherein the relative dominance data of the microbial community in step (4) is the relative dominance data of a contaminant-specific indicator microorganism, and the index microorganism is Angustibacter luteus in the case of soil without a contaminant , Runella slithyformis, Cystobacter velatus, Geodermatophilus daqingensis, Pseudonocardia soli, Aquabacterium commune, and Skermanella rubra At least one selected from the group consisting of, When the contaminant is ammonia, at least one selected from the group consisting of Nitrosospira tenuis, Flavihumibacter sediminis, Flavobacterium saliperosum, Brevundimonas naejangsanensis, Actinotalea ferrariae, Nocardioides caricicola, Chryseobacterium montanum, Flavobacterium dankookense, and Pedobacter bauzanensis , and when the contaminant is hydrochloric acid , Mucilaginibacter polysacchareus, Dyella japonica, Actinocorallia aurantiaca, Pedobacter kyungheensi, Flavobacterium resistens, Pedobacter cryoconitis, and Flavobacterium spartansii , and at least one selected from the group consisting of, and the contaminant is In the case of nitric acid, at least one selected from the group consisting of Methylophilus rhizosphaerae, Pseudomonas silesiensis, Ralstonia pickettii, and Cupriavidus campinensis , and in the case of hydrogen fluoride, Bacillus marisflavi, Paenibacillus aceris, Paraburkholderia hospital, Halobacillus profundi, Clostridium puniceum, Bacillus plakortidis, Bacillus aryabhattai , At least one selected from the group consisting of Paenibacillus populi, Rhodanobacter xiangquanii and Bordetella flabilis , and when the pollutant is sulfuric acid, at least one selected from the group consisting of Micromonospora noduli, Actinospica acidiphila, Desulfohalotomaculum peckii and Tumebacillus ginsengisoli Method for predicting soil pollutants.
The method of claim 20, wherein the step of training the artificial neural network model in step (5) is a step of setting the number of hidden layer neurons to 7.
A computer readable recording medium on which a computer program for executing the method according to any one of claims 1 to 7, 9 and 10 is recorded.
A computer readable recording medium on which a computer program for executing the method according to any one of claims 11 to 18 is recorded.

Images