CfRNA androgen alopecia early screening method and system based on machine learning model

Abstract

The invention discloses a cfRNA male hormone alopecia early screening method and system based on a machine learning model, and belongs to the field of male hormone alopecia, and the method comprises the following steps: extracting RNA from a sample set to obtain an RNA data set; constructing a library based on the RNA data set; performing quality inspection and NGS sequencing on the library to obtain sequencing data; performing bioinformatics analysis on the sequencing data to obtain significant differential genes; and establishing a prediction model by taking the significantly differential genes as input features of a random forest model, and inputting new sample information into the prediction model to obtain the alopecia risk of the subject corresponding to the sample.

Description

Technical Field

[0001] The present invention belongs to the technical field of androgenic alopecia, and in particular relates to a cfRNA androgenic alopecia early screening method and system based on a machine learning model. Background Art

[0002] Cell-free RNAomics provides a new direction for cancer genome data classification. cfRNAomics for early disease detection has become a new research direction in academia, and several international journals have published papers on early disease detection using cfRNA.

[0003] In addition, the trend of using artificial intelligence in the field of life sciences has been increasing year by year, among which language-based artificial intelligence models have made rapid breakthroughs in recent years. In 2017, the Transformer artificial intelligence model was accepted by the top conference "NIPS". In 2021, OpenAI used the underlying algorithm based on Transformer to superimpose and rebuild the generative pre-trained transformer model (GPT), and commercialized the product ChatGPT. This product has become a major breakthrough in the field of artificial intelligence with its high intelligence, strong understanding ability, and generative characteristics, allowing artificial intelligence to quickly popularize among the general public. In 2022, the MetaA team released the algorithm EsFold based on the BioRivX preprint, and in 2023, the entrepreneurial team published the model ProgOne. By using a language model, it achieved a structural prediction capability comparable to AlphaFold 2 protein, and demonstrated higher effects in specific fields such as unknown proteins. Drug discovery: In 2024, DeepMind launched a new version of AlphaFold, called AlphaFold-3, which can predict protein-drug interactions with higher accuracy. This progress has accelerated the drug screening process, allowing some new drug candidates to enter the clinical trial stage in just a few months. In addition, a company used artificial intelligence to optimize the CRISPR-Cas9 gene editing tool. The AI ​​model it developed can accurately predict the off-target effects of gene editing, significantly improving the safety and effectiveness of gene editing. This breakthrough has achieved initial success in the classification and identification of rare genetic diseases. Another startup launched BioFold, a protein structure prediction model that combines quantum computing and AI technology. BioFold excels in predicting the structure of complex protein complexes, especially in the study of interactions between viruses and host cells, providing new targets for the development of antiviral drugs. A company has developed a new personalized medical platform using AI technology. The platform integrates patients' genomic data, electronic medical records, and lifestyle data, and conducts comprehensive analysis through AI models to provide personalized patient situation analysis.

[0004] So far, many papers have pointed out the effectiveness of artificial intelligence in early classification of hair loss. In 2018, a paper reported that artificial intelligence machine learning methods can identify whether a patient has hair loss. In 2021, a paper reported that machine learning is used to distinguish alopecia areata. In 2024, a paper reported the use of deep learning methods to form an early classification model for hair loss and other scalp diseases.

[0005] According to the traditional method, after the sample cfRNA is subjected to traditional bioinformatics analysis, the results obtained often require human judgment, relying on accumulated experience to judge the results, and the results may not be accurate and reliable, and have a certain degree of subjectivity. Therefore, the present invention makes appropriate adjustments to the shortcomings of the traditional method and proposes a cfRNA androgenic alopecia early screening method and system based on a machine learning model. Summary of the invention

[0006] In order to solve the above technical problems, the present invention proposes a cfRNA androgenic alopecia early screening method and system based on a machine learning model to solve the problems existing in the above-mentioned prior art.

[0007] To achieve the above objectives, the present invention provides a cfRNA androgenic alopecia early screening method based on a machine learning model, comprising:

[0008] Extract RNA from the sample set to obtain an RNA data set;

[0009] constructing a library based on the RNA dataset;

[0010] Performing quality inspection and NGS sequencing on the library to obtain sequencing data;

[0011] Performing bioinformatics analysis on the sequencing data to obtain significantly differentially expressed genes;

[0012] The significantly different genes are used as input features of the random forest model to establish a prediction model, and new sample information is input into the prediction model to obtain the hair loss risk of the subject corresponding to the sample.

[0013] Optionally, the process of constructing a library based on the RNA dataset includes:

[0014] The extracted RNA is fragmented and rRNA is removed based on a fragmentation and rRNA removal buffer to obtain a first sample;

[0015] Performing cDNA synthesis on the first sample and adding a linker to obtain a second sample;

[0016] The second sample is subjected to library amplification to prepare a library.

[0017] Optionally, the process of obtaining the second sample further includes product purification, wherein the process of product purification includes:

[0018] Balancing the purified magnetic beads, adding the purified magnetic beads to the first mixed solution with the linker added, mixing and incubating to obtain a second mixed solution;

[0019] The second mixed solution is centrifuged to remove the supernatant, and then ethanol is added for washing and drying to obtain a third mixed solution;

[0020] Add water to the third mixed solution, mix well, let stand, centrifuge, and remove the supernatant to obtain a second sample.

[0021] Optionally, the process of performing bioinformatics analysis on the sequencing data includes:

[0022] Performing quality control on the sequencing data to obtain high-quality sequencing data;

[0023] Comparing the high-quality sequencing data with human genome data to obtain comparison data;

[0024] Counting the number of sequencing reads of each gene in the alignment data to generate gene count data;

[0025] The gene count data were subjected to differential expression analysis to obtain genes with significant differences.

[0026] Optionally, the process of comparing the high-quality sequencing data with human genome data includes:

[0027] Build a genome index based on the human genome reference sequence and annotation files;

[0028] The high-quality sequencing data is compared with the genome index to obtain comparison data.

[0029] Optionally, the process of performing differential expression analysis on the gene count data includes:

[0030] Reading gene count data and preprocessing it to obtain preprocessed data;

[0031] Based on the voom function, the preprocessed data is transformed and normalized to obtain normalized data

[0032] The normalized data were subjected to linear modeling based on lmFit and eBayes functions to extract genes with significant differences.

[0033] The present invention also discloses a cfRNA androgenic alopecia early screening system based on a machine learning model, comprising: a sample collection module, an RNA extraction module, a library preparation module, a high-throughput sequencing module, a bioinformatics analysis module and an artificial intelligence analysis module;

[0034] Wherein, the sample collection module collects samples of the hair loss group and samples of the healthy group;

[0035] The RNA extraction module is used to extract RNA data of the sample;

[0036] The library preparation module is used to prepare a library based on the RNA data;

[0037] The high-throughput sequencing module is used to obtain sequencing data based on the library;

[0038] The bioinformatics analysis module is used to extract significantly different genes from the sequencing data;

[0039] The artificial intelligence analysis module predicts the subject's risk of hair loss based on the significantly differentially expressed genes.

[0040] The present invention also discloses a computer terminal device, comprising:

[0041] one or more processors;

[0042] A memory, coupled to the processor, for storing one or more programs;

[0043] When the one or more programs are executed by the one or more processors, the one or more processors implement a cfRNA androgenic alopecia early screening method based on a machine learning model.

[0044] The present invention also discloses a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, a cfRNA androgenic alopecia early screening method based on a machine learning model is implemented.

[0045] Compared with the prior art, the present invention has the following advantages and technical effects:

[0046] The biological experimental method of the present invention, the cfRNA micro-extraction library construction method has a higher qualitative synthesis efficiency compared with the traditional RNA extraction library construction method.

[0047] The present invention uses machine learning to construct an early screening model that can predict the high-risk population of potential disease. The model can reduce the inaccuracy of traditional single bioinformatics analysis plus human subjective judgment verification, reduce human interference to a certain extent, and improve the credibility and accuracy of the test results. Therefore, the model involved in the present invention has a certain universal method, and the model has better performance, low cost and short cycle. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] The drawings constituting a part of the present application are used to provide a further understanding of the present application. The illustrative embodiments and descriptions of the present application are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0049] Figure 1 is a flow chart of bioinformatics analysis according to an embodiment of the present invention;

[0050] Figure 2 This is a flow chart of model training in an embodiment of the present invention;

[0051] Figure 3 The result diagram of whether hair loss occurs or not in the embodiment of the present invention, wherein the ROC-AUC Curve (AUC=0.70);

[0052] Figure 4 The ROC-AUC Curve prediction result diagram of the severity of hair loss according to an embodiment of the present invention (AUC=0.85);

[0053] Figure 5 It is the difference in expression of the key factors of the embodiment of the present invention. DETAILED DESCRIPTION

[0054] It should be noted that, in the absence of conflict, the embodiments and features in the embodiments of the present application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0055] It should be noted that the steps shown in the flowcharts of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and that, although a logical order is shown in the flowcharts, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0056] Embodiment 1

[0057] This embodiment provides a cfRNA androgenic alopecia early screening method and system based on a machine learning model, comprising the following steps:

[0058] In the plasma processing step, the addition of a specific lysis buffer and the vortex mixing step ensure that the sample is fully mixed and reacted, thereby improving the efficiency and purity of RNA extraction. RNA extraction reagents are used to remove organic phase impurities; by adjusting the amount of precipitant and centrifugation, the total RNA in the sample is more thoroughly separated and collected, thereby obtaining higher purity and more types of RNA samples, including large molecular weight RNA (28S rRNA, 18S rRNA, mRNA) and small molecular weight RNA (miRNA, siRNA, shRNA, snRNA, etc.). Centrifugal columns are used to collect samples by adsorption, while ensuring the purity of RNA. In the RNA collection step, the stability and efficient collection of RNA are ensured by adjusting the added volume of the solution and the incubation method.

[0059] The library construction process includes RNA fragmentation, rRNA removal, conventional ds-cDNA synthesis, and library amplification. The configuration of multiple reaction systems ensures the uniformity of RNA fragments, the removal of rRNA, and the efficiency and accuracy of cDNA synthesis. In the PCR reaction of adding adapters and labels, the concentration of adapters and labels is adjusted to improve the specificity and efficiency of the reaction, and to avoid problems such as adapter contamination and primer dimers. Magnetic bead treatment and magnetic separation technology are used multiple times, and 80% ethanol washing is used to ensure the efficiency and purity of the purification process. Buffer treatment and magnetic bead treatment are used to accurately remove ribosomal RNA, reduce the content of Mt-rRNA (mitochondrial rRNA) in the sequencing results (less than 30%), and improve the quality of sequencing data. By adjusting the number of cycles of double-strand amplification, the high quality and concentration of the constructed library are ensured.

[0060] According to traditional methods, after traditional bioinformatics analysis of sample cfRNA, the results often require human judgment and rely on accumulated experience to judge the results. The results may not be accurate and reliable and have a certain degree of subjectivity.

[0061] This invention makes appropriate adjustments to the shortcomings of traditional methods and proposes a cfRNA androgenic alopecia early screening method based on a machine learning model. As a specific implementation of this embodiment, it includes the following steps:

[0062] S1. Sample collection

[0063] Inclusion criteria: Participants in the hair loss group provided the area of ​​their hair loss, the length of the hair loss area, the width of the hair loss area, and the width of the hairline. According to the severity of hair loss, they were divided into the healthy group, the low hair loss group, the medium hair loss group, and the high hair loss group, totaling 106 people. The proportion of people in each group is shown in Table 1. The criteria include individuals with other skin diseases, those who use hair growth drugs or antibacterial shampoos or hair care products.

[0064] Table 1

[0065] Healthy control group Low middle high total 41(38.67％) 18(16.98％) 31(29.25％) 16(15.09％) 106(100％)

[0066] S2. RNA extraction

[0067] S2.1 Sample preparation: Blood samples were centrifuged to separate plasma and transferred to new storage tubes.

[0068] S2.2 Plasma treatment: Add plasma / serum to a centrifuge tube. Add buffer and vortex to mix.

[0069] S2.3 Centrifugal separation: Perform centrifugation and collect the supernatant. Transfer the supernatant to a new centrifuge tube, add the precipitant, and mix well.

[0070] S2.4 Waste liquid treatment: transfer the liquid to the centrifuge column, centrifuge, and discard the waste liquid. Add washing buffer, centrifuge separately and discard the waste liquid

[0071] S2.5 RNA collection: Place the centrifuge column into the collection tube, add solution, incubate, and collect RNA after centrifugation.

[0072] S3. Library preparation

[0073] S3.1 Fragmentation and removal of ribosomal rRNA: Prepare fragmentation and removal of rRNA buffer, and run the reaction program on a PCR instrument. After the reaction is completed, the sample is cooled to obtain the first sample.

[0074] S3.2 First-strand cDNA synthesis: Prepare the mixed solution, add it to the reaction tube, vortex, and centrifuge.

[0075] S3.3 Second-strand cDNA synthesis and adding adapters:

[0076] (1) Solution preparation: Prepare PCR reaction 1 mixture and add it to the reaction tube. Run the reaction program on the instrument. After the reaction is completed, cool the sample.

[0077] (2) Configuring the connector connection system: Configuring the connector connection mixed solution, adding it to the reaction tube. The instrument runs the reaction program. After the reaction is completed, the sample is cooled to obtain the first mixed solution.

[0078] (3) Product purification: Add the purified magnetic beads to the first mixed solution with the linker, mix and incubate to obtain a second mixed solution; centrifuge the second mixed solution to remove the supernatant, then add ethanol to wash and dry to obtain a third mixed solution; add water to the third mixed solution, mix and let stand, centrifuge and remove the supernatant to obtain a second sample. Balance the magnetic beads and mix. Add some purified magnetic beads, mix and incubate. Centrifuge, separate the magnetic beads and liquid with a magnetic rack, and remove the supernatant. Add ethanol to wash, repeat once, and dry the magnetic beads. Add an appropriate amount of water, mix and let stand, centrifuge and remove the supernatant to obtain a second sample.

[0079] S3.4 Library amplification:

[0080] (1) PCR2 mixed solution: Prepare PCR2 mixed solution and perform PCR.

[0081] (2) PCR2 library purification: Balance the magnetic beads and mix well. Add some purified magnetic beads, mix well and incubate. Centrifuge, separate the magnetic beads and liquid with a magnetic rack, and remove the supernatant. Add ethanol to wash, repeat once, and dry the magnetic beads. Add appropriate amount of water, mix well and let stand, centrifuge and remove the supernatant.

[0082] S4. High-throughput sequencing

[0083] S4.1 Library quality inspection: The library concentration is tested by fluorescence quantification instrument. The library fragment distribution is analyzed by bioanalyzer. Samples that pass the quality inspection are sequenced.

[0084] S4.2 NGS sequencing: Calculate the library concentration and dilute to the specified concentration. Denature and dilute the library. Inject the sample into the sample well of the kit. Import the sample information, clean the chip, insert the kit and chip. Confirm the machine settings and start sequencing.

[0085] S5. Bioinformatics analysis

[0086] In this step, bioinformatics methods are used to preprocess and analyze the sequencing data. The expression differences of cfRNA groups between hair loss patients and healthy controls are compared based on the data after bioinformatics analysis. Figure 1 shown.

[0087] S5.1 Sequencing data quality control: fastp v0.23.3 was used for data quality control and low-quality data was cleaned up. The quality of the original sequencing data was assessed and low-quality sequences were removed. Then, the adapter sequences were detected and removed to finally generate cleaned high-quality sequencing data.

[0088] S5.2 Sequencing data alignment: STAR v2.7.10b was used to align sequencing data with human genome data. First, a genome index was constructed using the human genome reference sequence and annotation file. Second, the cleaned sequencing data (FASTQ file) was aligned with the constructed genome index to generate an alignment file (BAM format).

[0089] S5.3 Gene counting: Gene counting was performed using feature Counts v 2.0.6. First, the BAM file generated by the alignment was aligned with the gene annotation file (GTF format). Second, the number of sequencing reads for each gene was counted to generate gene count data.

[0090] S5.4 Differential expression analysis: Use limma-voom for differential expression analysis. Read gene count data and perform preprocessing; use voom function to transform and normalize data; use lmFit and eBayes functions for linear modeling and differential expression analysis; extract significantly differentially expressed genes.

[0091] S5.5 Batch effect removal & normalization: Model the batch effect, and use the voom function to transform the variables that affect the batch effect in the sample information, such as age and sampling location information, to remove the batch effect. Finally, use PCA for visualization analysis to see if there is any residual batch effect.

[0092] S6. Artificial intelligence analysis: Use machine learning random forest algorithm to predict whether or not hair loss occurs and the severity of hair loss. First, the significantly different genes obtained from the above bioinformatics difference analysis are used as characteristic genes, and the samples are divided into corresponding training sets (50%), validation sets (25%), and test sets (25%). The decision tree calculates the accuracy of the prediction, and after calculations of multiple decision trees, the fitting average is finally taken as the prediction accuracy result, so as to make a preliminary judgment on the test results. The trained model can judge the newly input data and assess the risk of hair loss.

[0093] The specific process of model training (random forest machine learning) is as follows Figure 2 As shown. The model is trained using a labeled gene expression statistical data set. The model input is the key differentially expressed genes of cfRNA screened out by the above bioinformatics analysis, so as to make a preliminary judgment on the test indicators and then predict whether the subject is at risk of hair loss.

[0094] S6.1 Use labeled gene expression data: Screen out key significantly differentially expressed genes through bioinformatics analysis, and use these genes as input features to more specifically judge and predict the risk of hair loss. This method may improve the prediction accuracy of the model by selecting specific biomarkers.

[0095] S6.2 Training set: The training set data is used to train the model. The model learns the relationship between sample features and labels in the training set and adjusts its own parameters so that it can accurately predict new samples. The training set of this model accounts for 80% of the total data set.

[0096] S6.3 Validation set: During the training process, the validation set is used to adjust the model's hyperparameters (such as the degree of randomness, the number of decision trees, the maximum depth, etc.). The model calculates the optimal hyperparameters based on the validation set data. The model sets the optimal parameters for itself and then retrains. After adjusting the parameters, it is retrained to effectively avoid overfitting or underfitting and ensure that the model performs better in practical applications.

[0097] S6.4 Test set: The test set is not involved in the training and validation process at all, and is only used in the final evaluation of model performance, ensuring the fairness and reliability of the evaluation results. Traditional methods sometimes do not strictly distinguish between training and test data, which may lead to high evaluation results.

[0098] S6.5 Model evaluation

[0099] (1) Performance indicators: Use indicators such as accuracy (ROC-AUC), sensitivity, and specificity to evaluate model performance.

[0100] (2) Feature importance analysis: using the interpretation method of the random forest model (to evaluate the contribution of each cfRNA feature to model prediction and identify key markers).

[0101] (3) Biological validation: Experimental validation of the biological functions of key cfRNA markers and their association with FAGA.

[0102] (4) The innovation of experimental methods has significantly improved the stability and success rate of cfRNA micro-extraction and library construction.

[0103] (5) The hair loss-related biomarkers discovered in the present invention are reported for the first time, based on the experimental and sequencing system independently developed by the inventors.

[0104] In summary, the model established by the present invention has better performance, low cost, and short cycle, and can be widely used in the early diagnosis and screening of androgenic alopecia in women.

[0105] The accuracy, sensitivity, specificity, ROC curve and AUC of the model established in the present invention are respectively as follows: Figure 3 and Figure 4 shown.

[0106] Prediction of hair loss or not: Sample accuracy rate Figure 3 As shown in Table 2, the (AUC) reached 70%, revealing the high accuracy and effectiveness of the model in making decisions on given data; the model prediction results included 6 key factors: RNY1, B2M, EEF1A1, TMSB4X, ATP6 and CYTB, as shown in Table 2, and the cDNA sequences of the key factors are shown in Table 3.

[0107] Table 2

[0108] Gene name EnsemblID EntrezID Predicted value RNY1 ENSG00000201098 6084 0.1826 B2M ENSG00000166710 567 0.1739 EEF1A1 ENSG00000156508 1915 0.1739 TMSB4X ENSG00000205542 7114 0.1652 ATP6 ENSG00000198899 4508 0.1652 CYTB ENSG00000198727 4519 0.1391

[0109] Table 3

[0110]

[0111]

[0112]

[0113]

[0114]

[0115]

[0116]

[0117]

[0118]

[0119]

[0120]

[0121]

[0122] Prediction principle: These six key factors are biologically significant factors derived from the above bioinformatics analysis and have a high correlation with the disease. The gene expression data corresponding to the phenotypes of the six key factors are used as the input for model training. After the model is trained, the output results include prediction accuracy and prediction value. The prediction value is used as an evaluation indicator to predict whether hair loss occurs or not. The higher the prediction value, the higher the probability of hair loss.

[0123] Prediction results: The accuracy of the model prediction is 0.70 using the phenotypes of the six key factors to predict whether hair loss occurs or not. The predicted value is the output of the model prediction result.

[0124] Verification of prediction results: Extract the importance scores of the six key factors, i.e. the predicted values, from the trained model.

[0125] The specific relationship between the predicted value and the predicted result: The predicted value indicates the contribution of the feature to the model prediction. The score is related to the correlation between the factor itself and the disease. The score is the predicted value output after the model is trained. The higher the predicted value, the greater the correlation with the disease, and vice versa, the lower the predicted value, the smaller the correlation with the disease. Among the key factors used for prediction, RNY1 is the most important, and the importance of the predicted result is also the highest. The final prediction result is the judgment of whether there is hair loss or not, which is determined by the predicted value. When the predicted value reaches a significant value. After inputting new data into the trained model, the model outputs "aga" to indicate that there is a risk of hair loss. If the value is not reached, the model outputs "none" to indicate that there is no risk of hair loss. The output statement is the final evaluation result.

[0126] like Figure 5 As shown in the figure, the sample expression levels corresponding to the 6 key factors are plotted using drawing software. Each large column contains 6 small columns; the small columns represent the expression levels of the 6 key factors in the same sample). The horizontal axis represents all samples, which are divided into severe hair loss (upper), moderate (mid), mild (lower) and control (control). The vertical axis represents the expression level of the gene in the sample. As shown in the figure, CYTB has a high expression level in the upper sample: 13269.96, and in a control group sample, the expression level of ATP6 is relatively high: 8126.39).

[0127] Hair loss severity prediction: This model can also predict the severity of hair loss based on the hair loss severity index (AGA Index), and the prediction results are output as "low", "medium" or "high".

[0128] Prediction principle: The hair loss severity index is calculated by calculating the relevant variables of the sample information. The index is used as a feature quantity to train the model. The trained model can directly output the severity of hair loss.

[0129] Prediction results: The accuracy of the model using this feature input reached 0.85.

[0130] Prediction result verification: Extract the importance scores of the six key factors from the trained model, i.e. the predicted value. The prediction result does not need further verification, and the output prediction value is the prediction result.

[0131] The specific relationship between the predicted value and the predicted result: The predicted value represents the predicted result of the feature for the model prediction.

[0132] The specific key factor (biomarker) determined by the present invention has never been reported before, and its originality relies on the experimental and sequencing system, which is original to the present invention.

[0133] Embodiment 2

[0134] This embodiment discloses a computer terminal device, including:

[0135] one or more processors;

[0136] A memory, coupled to the processor, for storing one or more programs;

[0137] When the one or more programs are executed by the one or more processors, the one or more processors implement a cfRNA androgenic alopecia early screening method based on a machine learning model.

[0138] Embodiment 3

[0139] This embodiment discloses a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, a cfRNA androgenic alopecia early screening method based on a machine learning model is implemented.

[0140] The above are only preferred specific implementations of the present application, but the protection scope of the present application is not limited thereto. Any changes or substitutions that can be easily thought of by a person skilled in the art within the technical scope disclosed in the present application should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A cfRNA androgenic alopecia early screening method based on a machine learning model, characterized in that: The following steps are involved: Extract RNA from the sample set to obtain an RNA dataset; constructing a library based on the RNA dataset; Performing quality inspection and NGS sequencing on the library to obtain sequencing data; Performing bioinformatics analysis on the sequencing data to obtain significantly differentially expressed genes; The significantly different genes are used as input features of the random forest model to establish a prediction model, and new sample information is input into the prediction model to obtain the hair loss risk of the subject corresponding to the sample.

2. The cfRNA androgenic alopecia early screening method based on a machine learning model according to claim 1, characterized in that: The process of constructing a library based on the RNA dataset includes: Fragmenting the extracted RNA and removing rRNA using a fragmentation and rRNA removal buffer to obtain a first sample; synthesizing cDNA from the first sample and adding a linker to obtain a second sample; The second sample is subjected to library amplification to prepare a library.

3. The cfRNA androgenic alopecia early screening method based on a machine learning model according to claim 2, characterized in that: The process of obtaining the second sample further includes product purification, wherein the process of product purification includes: Balancing the purified magnetic beads, adding the purified magnetic beads to the first mixed solution with the linker, mixing and incubating to obtain a second mixed solution; The second mixed solution is centrifuged to remove the supernatant, and then ethanol is added for washing and dried to obtain a third mixed solution; Water is added to the third mixed solution, mixed and allowed to stand, and then centrifuged and the supernatant is removed to obtain a second sample.

4. The cfRNA androgenic alopecia early screening method based on a machine learning model according to claim 1, characterized in that: The process of performing bioinformatics analysis on the sequencing data includes: Performing quality control on the sequencing data to obtain high-quality sequencing data; Comparing the high-quality sequencing data with human genome data to obtain comparison data; Counting the number of sequencing reads for each gene in the aligned data to generate gene count data; The gene count data were subjected to differential expression analysis to obtain significantly differentially expressed genes.

5. The cfRNA androgenic alopecia early screening method based on a machine learning model according to claim 4, characterized in that: The process of comparing the high-quality sequencing data with human genome data includes: Build a genome index based on the human genome reference sequence and annotation files; The high-quality sequencing data is aligned with the genome index to obtain aligned data.

6. The cfRNA androgenic alopecia early screening method based on a machine learning model according to claim 4, characterized in that: The process of performing differential expression analysis on the gene count data includes: Read gene count data and preprocess it to obtain preprocessed data; The preprocessed data is transformed and normalized based on the voom function to obtain normalized data The normalized data were subjected to linear modeling based on lmFit and eBayes functions to extract significantly differentially expressed genes.

7. A cfRNA androgenic alopecia early screening system based on a machine learning model, characterized by: include: Sample collection module, RNA extraction module, library preparation module, high-throughput sequencing module, bioinformatics analysis module, and artificial intelligence analysis module; Wherein, the sample collection module collects samples from the hair loss group and the healthy group; The RNA extraction module is used to extract RNA data of the sample; The library preparation module is used to prepare a library based on the RNA data; The high-throughput sequencing module is used to obtain sequencing data based on the library; The bioinformatics analysis module is used to extract significantly differentially expressed genes from the sequencing data; The artificial intelligence analysis module predicts the subject's risk of hair loss based on the significantly differentially expressed genes.

8. A computer terminal device, characterized in that: include: one or more processors; a memory, coupled to the processor, for storing one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors implement the cfRNA androgenic alopecia early screening method based on the machine learning model as described in any one of claims 1-6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the cfRNA androgenic alopecia early screening method based on a machine learning model as described in any one of claims 1 to 6 is implemented.

Images