A sketch data quality correction and curation system based on a haemophilus influenzae complete genome benchmark

Abstract

This invention relates to the fields of bioinformatics and data processing technology, and provides a draft data quality correction and optimization system based on the complete genome benchmark of Haemophilus influenzae. The system includes: an automated data acquisition module configured to automatically acquire the dataset and metadata of the target microorganism from a public database; a dynamic benchmark construction module configured to construct a standard feature vector based on the dataset and metadata of the target microorganism; an interference calculation module configured to calculate the fragmentation interference coefficient K of the dataset relative to the standard feature vector; and an adaptive correction and optimization module configured to correct the confidence level of gene detection results according to the interference coefficient K and output an optimized dataset with weight labels. This invention solves the problems of data waste and false negatives caused by hard threshold filtering in existing technologies, and realizes the effective utilization and accurate correction of low-quality genomic data.

Description

Technical Field

[0001] This invention relates to the fields of bioinformatics and data processing technology, specifically a draft data quality correction and optimization system based on the complete genome of Haemophilus influenzae. Background Technology

[0002] With the widespread adoption of high-throughput sequencing technology, public databases such as NCBI have accumulated massive amounts of microbial genome data. However, these data exhibit significant differences in quality: 1. Severe data stratification: Only a small portion of the data is a high-quality "complete genome," while over 90% of the data is a fragmented and pieced-together "draft genome."

[0003] 2. Deficiencies of existing technologies: Current mainstream data cleaning methods typically employ a "hard threshold filtering" strategy (e.g., directly removing sketch data with N50 < 20kb or Contig number > 200; or using quality control tools such as CheckM to set data with Completeness < 90% or Contamination > 5% as unacceptable and discard them): Defect 1 (Data Waste): Most Haemophilus influenzae isolates with significant clinical value often exhibit fragmented assembly results due to insufficient sequencing depth or host contamination. Directly removing them based on hard thresholds would result in the permanent loss of valuable clinical sample information (such as drug resistance phenotypes and isolate source information). Defect 2 (False Negative Misleading): In low-quality data, the failure to detect a certain virulence gene is often due to "false negatives" caused by splice breaks. Current technology cannot distinguish between "true deletion" and "undetectable", which can easily lead to erroneous negative conclusions.

[0004] Therefore, in view of the above situation, there is an urgent need to provide a draft data quality correction and optimization system based on the complete genome of Haemophilus influenzae to overcome the shortcomings in current practical applications. Summary of the Invention

[0005] The purpose of this invention is to provide a draft data quality correction and optimization system based on the complete genome of Haemophilus influenzae, which effectively solves the problems in the background art.

[0006] This invention is implemented as follows: a draft data quality correction and optimization system based on the complete genome benchmark of Haemophilus influenzae, the system comprising: An automated data acquisition module is configured to automatically acquire the target microorganism's dataset and metadata from a public database. The dynamic benchmark building module is configured to construct standard feature vectors based on the target microorganism's dataset and metadata. The interference calculation module is configured to calculate the fragmentation interference coefficient of the dataset to be processed relative to the standard feature vector. ; The adaptive correction and optimization module is configured to adjust the value based on the interference coefficient. The gene detection results are adjusted for confidence level, and a preferred dataset with weighted labels is output; the adaptive adjustment and optimization configuration are based on the fragmentation interference coefficient. The confidence level of gene detection results from Haemophilus influenzae genome draft data is corrected, and a preferred dataset with weight labels is output. The confidence level correction includes confidence level correction for gene deletion and confidence level correction for gene presence. The value of the weight label is matched with the corrected confidence level for gene deletion.

[0007] As a further aspect of the present invention: the dataset to be processed includes all sketch sequences with an assembly level of "Contig" or "Scaffold"; The metadata includes sequencing platform, submission year, and environmental parameters of the isolation source.

[0008] As a further aspect of the present invention: the standard feature vector constructed by the dynamic benchmark construction module includes N50 and the contig number.

[0009] As a further aspect of the present invention: the formula for calculating the fragmentation interference coefficient K by the interference degree calculation module is as follows: ; in, The ratio of the measured N50 of the sample to the baseline N50 reflects the loss of sequence continuity; min() indicates taking the minimum of the two values ​​within the parentheses when the number of sample fragments ( Exceeding the preset threshold ( When ), the system forces the value of this item to be 1, where ; + =1, It is a continuous weight. It is a fragmented weight.

[0010] As a further aspect of the present invention: the interference coefficient The range of values ​​for is 0 ≤ K ≤ 1. A higher value indicates a higher degree of data fragmentation.

[0011] As a further aspect of the present invention, the confidence correction includes correction for gene deletion and correction for gene presence.

[0012] As a further aspect of the present invention: the correction formula for gene deletion is as follows: ; in The adjusted confidence level for gene deletion. This is the preset correction factor.

[0013] As a further aspect of the present invention: the correction factor The value is 0.9.

[0014] As a further aspect of the present invention: the target microorganism is Haemophilus influenzae; the target microorganism is Haemophilus influenzae; when the adaptive correction and optimization module performs gene presence confidence correction, it first identifies the length and sequence edge state of the contig containing the target gene, and then reduces the sequence integrity score of the target gene according to the numerical gradient of the fragmentation interference coefficient K, wherein the sequence integrity score is the core reference index for the corrected confidence of gene presence; the module sets a contig length threshold and a sequence truncation judgment standard. If the length of the contig containing the target gene is lower than the threshold and there are base truncation features at the sequence edge, the influence weight of the fragmentation interference coefficient K on the sequence integrity score is increased; after the adaptive correction and optimization module completes the confidence correction of the gene detection results, it generates a unique confidence weight label for each dataset to be processed. The weight label is assigned based on the corrected confidence value of gene loss, directly mapping the credibility of the gene detection results of the dataset to be processed; the adaptive correction and optimization module is configured with a data filtering and grading unit, which filters and grades data according to the confidence correction results. The datasets to be processed are classified into three levels: high-confidence datasets, suspected false negative datasets, and low-confidence datasets. Different levels of datasets correspond to different downstream analysis access permissions. The criteria for judging suspected false negative datasets are: the initial gene detection result is negative, and the corrected confidence value of gene deletion is less than 0.5. The adaptive correction and optimization module labels such datasets with suspected false negative tags and completely preserves their original data and metadata. The adaptive correction and optimization module is equipped with a structured data output unit, which integrates the sample number, original N50, fragmentation interference coefficient K, original gene detection result, system correction conclusion, and final weight label of the dataset to be processed into a standardized structured table to form the optimized dataset. The adaptive correction and optimization module is equipped with a dynamic weighting unit, which uses the final weight label value of each sample in the optimized dataset as the effective sample coefficient and applies it to downstream statistical analysis such as the Haemophilus influenzae gene prevalence rate, so that the statistical results are presented in a weighted calculation manner, reducing the misleading effect of low-confidence datasets on the statistical results.

[0015] Compared with the prior art, the beneficial effects of the present invention are as follows: 1. Improved data utilization: This invention avoids the sample size loss caused by traditional "one-size-fits-all" filtering and can effectively utilize more than 90% of the sketch data in the public database.

[0016] 2. Correction of bias: By introducing an "interference correction" mechanism, the problem of "false negatives" caused by splicing fragmentation in low-quality data is effectively solved, and the accuracy of virulence gene mining is improved.

[0017] 3. Automation and versatility: This system is built based on publicly available data and general algorithms, does not depend on specific wet experimental conditions, and is suitable for genome analysis of a variety of microorganisms. Detailed Implementation

[0018] The technical solution of the present invention will be clearly and completely described below. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] The present invention will be further explained below with reference to specific embodiments.

[0020] This invention provides a draft data quality correction and optimization system based on the complete genome of Haemophilus influenzae, the system comprising: An automated data acquisition module is configured to automatically acquire the target microorganism's dataset and metadata from a public database. The dynamic benchmark building module is configured to construct standard feature vectors based on the target microorganism's dataset and metadata. The interference calculation module is configured to calculate the fragmentation interference coefficient of the dataset to be processed relative to the standard feature vector. ; The adaptive correction and optimization module is configured to adjust the value based on the interference coefficient. The gene detection results are adjusted for confidence level, and a preferred dataset with weighted labels is output; the adaptive adjustment and optimization configuration are based on the fragmentation interference coefficient. The confidence level of gene detection results from Haemophilus influenzae genome draft data is corrected, and a preferred dataset with weight labels is output. The confidence level correction includes confidence level correction for gene deletion and confidence level correction for gene presence. The value of the weight label is matched with the corrected confidence level for gene deletion.

[0021] The target microorganism is Haemophilus influenzae. When the adaptive correction and optimization module performs gene presence confidence correction, it first identifies the length and sequence edge status of the contig containing the target gene, and then reduces the sequence integrity score of the target gene based on the numerical gradient of the fragmentation interference coefficient K. The sequence integrity score is the core reference indicator for correcting the gene presence confidence. The module sets a contig length threshold and a sequence truncation judgment standard. If the length of the contig containing the target gene is lower than the threshold and there are base truncation features at the sequence edge, the influence weight of the fragmentation interference coefficient K on the sequence integrity score is increased. After the adaptive correction and optimization module completes the confidence correction of the gene detection results, it generates a unique confidence weight label for each dataset to be processed. The weight label is assigned based on the corrected confidence value of gene loss, directly mapping the credibility of the gene detection results of the dataset to be processed. The adaptive correction and optimization module is configured with a data filtering and grading unit. This unit grades all datasets to be processed according to the confidence correction results. The results are categorized into high-confidence datasets, suspected false negative datasets, and low-confidence datasets, with different levels of datasets corresponding to different downstream analysis access permissions. The criteria for identifying suspected false negative datasets are: the initial gene detection result is negative, and the corrected confidence level for gene deletion is less than 0.5. The adaptive correction and optimization module labels these datasets with suspected false negative tags and fully preserves their original data and metadata. The adaptive correction and optimization module is equipped with a structured data output unit, which integrates the sample number, original N50, fragmentation interference coefficient K, original gene detection results, system correction conclusions, and final weight labels of the dataset to be processed into a standardized structured table, forming the optimized dataset. The adaptive correction and optimization module is also equipped with a dynamic weighting unit, which uses the final weight label values ​​of each sample in the optimized dataset as effective sample coefficients, applied to downstream statistical analyses such as the prevalence of Haemophilus influenzae genes, presenting statistical results in a weighted calculation manner to reduce the misleading effect of low-confidence datasets on statistical results.

[0022] The confidence correction for gene presence adopts a three-layer execution mechanism of "feature recognition - gradient reduction - weight adaptation", which solves the problem of misjudgment of gene presence caused by contig splicing fragments and sequence truncation in low-quality draft data, and achieves accurate matching between detection results and data quality.

[0023] Prior feature recognition: The module first performs basic feature analysis on the sketch data of the detected target gene to accurately identify the actual length of the contig where the target gene is located. At the same time, it analyzes the sequence edge state of the contig through sequence alignment to determine whether there are quality problems such as base deletion, truncation, and fuzzy matching, providing a basis for subsequent correction. This step is the key to distinguishing between "real and valid gene sequences" and "fragmented spliced ​​sequences".

[0024] K-value-based gradient reduction: After feature recognition, the module uses the fragmentation interference coefficient K as the core quantitative indicator to gradient-reduce the sequence integrity score of the target gene. The K value, as a core parameter characterizing the degree of data fragmentation, indicates a higher degree of data splicing and fragmentation, and a greater risk to gene sequence integrity due to splicing problems. The module pre-defines the gradient correspondence between the K value and the sequence integrity score; as the K value increases stepwise, the sequence integrity score decreases stepwise, ensuring that the score accurately reflects the degree of interference from data quality on gene presence detection. The sequence integrity score, as the core reference indicator for gene presence correction confidence, directly determines the credibility level of the gene presence detection result.

[0025] Differentiated weighting: To further improve the accuracy of the correction, the module pre-sets a contig length threshold and a sequence truncation judgment standard. This standard is based on the baseline characteristics of the complete Haemophilus influenzae genome and aligns with the inherent characteristics of the species' gene sequence. If the contig length of the target gene is lower than the preset threshold, and the sequence edge shows clear base truncation characteristics, it indicates that the gene sequence is significantly more affected by data fragmentation than usual. The module will automatically increase the weight of the fragmentation interference coefficient K on the sequence integrity score, resulting in a greater decrease in the sequence integrity score for this type of data. This avoids false high confidence judgments of gene existence due to low-quality fragmented sequences.

[0026] Through this mechanism, the module achieves a non-one-size-fits-all correction of gene presence detection results. It retains the core conclusion that the target gene was detected, and quantifies the credibility of the results through integrity scoring, thus solving the problem of result distortion caused by the binary judgment of "presence / absence" in traditional technologies.

[0027] The module configures data filtering and grading units, using confidence correction results and weight labels as the core basis to standardize and grade all datasets to be processed. At the same time, it configures differentiated downstream analysis access permissions for datasets of different grades, realizing the "classification and utilization, precise adaptation" of data, which not only ensures the accuracy of downstream analysis, but also maximizes the preservation of data resources.

[0028] The three-tiered classification standard divides the dataset into three levels: high-confidence datasets, suspected false negative datasets, and low-confidence datasets. The classification standard closely aligns with the core relationship between test results and data quality: High-confidence datasets are those with clear gene detection results, high gene deletion correction confidence scores (usually preset to ≥0.5), and low data fragmentation. Their test results are less affected by data quality and have high reliability. Suspected false negative datasets are those with initial negative gene detection results and gene deletion correction confidence scores less than 0.5. Due to their high fragmentation, the probability of false negatives is significantly increased, making them the core target of the module's "soft retention." Low-confidence datasets are those with extremely high data fragmentation, ambiguous gene detection results, and extremely low correction confidence scores. Their test results are severely affected by data quality and have limited reference value.

[0029] Specialized handling of suspected false negative datasets: For suspected false negative datasets, the module has developed specific handling rules. On the one hand, these datasets are labeled with a unique suspected false negative identifier to make them stand out in subsequent analyses and remind analysts to pay attention to the uncertainty of the results. On the other hand, the original data and metadata of these datasets are fully preserved, including key information such as sequencing platform, submission year, and isolation source, to avoid losing valuable clinical and epidemiological sample information due to data quality issues, thus preserving a foundation for subsequent re-analysis, data completion, or wet experimental validation.

[0030] Differentiated downstream usage permissions: Configure corresponding downstream analysis usage permissions for datasets of different levels. High-confidence datasets can be directly applied to core analysis scenarios such as Haemophilus influenzae gene prevalence statistics, virulence gene mining, and strain evolution analysis, serving as core data support; suspected false negative datasets can be used as reference data in analysis, or used after verification with other data. Their weight label values ​​will be used as effective sample coefficients in quantitative calculations, but will not be directly used as the basis for qualitative conclusions; low-confidence datasets are only retained as archived data and will not participate in regular downstream analysis to avoid misleading the analysis results.

[0031] This tiered screening mechanism achieves a precise match between data quality and analysis scenarios, ensuring the accuracy of core downstream analysis while solving the problem of data waste in traditional techniques through the "soft retention" of suspected false negative datasets, thus maximizing the utilization value of sketch data.

[0032] In this embodiment, the system uses a data processing method of "using superior data to correct inferior data with weighted adjustments" to calculate the "fragmentation interference degree" of the draft data based on the complete genome, instead of directly eliminating low-quality data. Based on this, the system dynamically corrects the confidence level of the gene detection results.

[0033] In a more specific example, the automated data acquisition module is configured to automatically retrieve two types of data on target microorganisms (such as Haemophilus influenzae) through a standard API interface of a public database (such as NCBIGenBank): 1. Dataset to be processed: Contains all sketch sequences with an assembly level of "Contig" or "Scaffold", where assembly level is the assembly level, Contig is the contig, and Scaffold is the frame.

[0034] 2. Metadata: Includes environmental parameters such as sequencing platform, submission year, and isolation source.

[0035] In a more specific example, the operational logic of the dynamic benchmark building module is as follows: The system automatically selects a set of strains with an assembly level of "Complete Genome".

[0036] Benchmark feature extraction: Calculate the feature vector of the benchmark set (including N50 benchmark value, core gene integrity contig number, etc.).

[0037] In this embodiment, "benchmark" does not refer to a single specific strain, but rather to the set of statistical characteristics (such as N50 length and contig) that the species should possess in its complete assembly state.

[0038] In a more specific example, the interference calculation module is configured to calculate the "quality degradation" of the dataset to be processed relative to the baseline.

[0039] Interference coefficient Algorithm: The system calculates the fragmentation interference coefficient of each sketch data based on the following formula. (0≤K≤1): ; in, The ratio of the measured N50 of the sample to the baseline N50 reflects the loss of sequence continuity; min() indicates taking the minimum of the two values ​​within the parentheses, used to set the "saturation boundary" of the interference, when the number of sample fragments ( Exceeding the preset threshold ( When this condition is met, the system forces the value of this item to be 1 to prevent overflow of the calculation result (i.e., K>1) due to extreme fragmentation (e.g., Contig>1000), ensuring that the interference coefficient always converges within the effective interval [0,1]. ; , The (weighting coefficient) is used to adjust the relative importance of continuity and the number of fragments. + =1, It is a continuous weight (default is 0.9). It is a fragmented weight (default is 0.1).

[0040] In a more specific example, the adaptive correction and optimization module is configured to adjust based on the interference coefficient. The confidence level of the gene testing results is corrected, and an optimal dataset is generated.

[0041] The correction logic is as follows: Correction for "gene deletion": If the target gene (such as a virulence gene) is not detected in the sample, the system does not directly output "0" (not present), but instead outputs "corrected confidence level": ;in The adjusted confidence level for gene deletion. The preset correction factor represents the "severity" of the system. It is set to 0.9 to set a "safe confidence baseline". Even for the worst quality sample (K≈1), the system retains about 10% of the basic weight to prevent the potential value of extreme data from being completely eliminated due to the algorithm being too absolute.

[0042] Correction for "gene presence": If the target gene is detected, but the contig containing the sequence is extremely short and truncated at the edges, the system will adjust accordingly. The value lowers its sequence integrity score.

[0043] In summary, the specific applications of this system are as follows: Step 1. Parameter definition and calculation logic (refer to Table 1): Table 1. Parameter Indicator Definitions and Calculation Logic ; Step 2. Experimental Sample Preparation Baseline sample: GCF_020736045.1 (Assembly: ASM2073604v1, FDA standard strain) N50 ref =1.83Mb, fully assembled genome Sample A to be processed: GCA_023224445.1, N50 sample =195.2kb,ContigNum sample =55.

[0044] Sample B to be processed: GCA_023225465.1, N50 sample =150.7kb,ContigNumsample =41.

[0045] Step 3. Calculation of interference (two-sample comparison) The system operates based on the set parameters (alpha=0.9, beta=0.1, ContigNum). threshold =100) Perform calculation For sample A (GCA_023224445.1): 1. N50 interference calculation:

[0046] 2. Calculation of Contig interference terms (using the min function): min(1, 55 / 100) = 0.55

[0047] 3. Total interference K A : K A =0.804 + 0.055 = 0.859 For sample B (GCA_023225465.1): 1. N50 interference:

[0048] 2. Contig interference items: min(1, 41 / 100) = 0.41

[0049] 3. Total interference K B : 0.826 + 0.041 = 0.867 Step 4. Target virulence gene detection and result correction To verify the system's ability to correct false negative results, this embodiment selects the hpd gene, a core virulence factor of Haemophilus influenzae, as the detection target.

[0050] Scenario setting: Given that the N50 values ​​of samples A and B are both far below the baseline (<200kb), there is an extremely high risk of gene truncation.

[0051] System correction: The system activates the confidence correction module and performs a weighted correction (γ=0.9) on the "negative" conclusion based on the interference coefficient K calculated in step 3.

[0052] Activate the confidence correction module (set sensitivity factor γ=0.9).

[0053] Correction result for sample A: PA =1-(K A

[0054] Output: Negative confidence level 22.7%.

[0055] Correction result for sample B: P B =1-(K B 0.9) = 1 - (0.867) 0.9) = 0.220 Output: Negative confidence level 22.0%.

[0056] Results Analysis: Although sample B had fewer fragments (41) than sample A (55), sample A achieved a slightly higher confidence score (22.7% > 22.0%) in the weighted model of this system due to its better N50 continuity (195kb > 150kb). This demonstrates that this system can specifically protect data with "good continuity but many fragments" by adjusting the weight parameters, which meets the scientific needs of virulence gene mining.

[0057] Step 5. Optimize Output and Dynamically Weight The system summarizes the above correction results and generates the final Haemophilus influenzae genome weighted optimization dataset.

[0058] The differences between this system and existing technologies are as follows: 1. Decision-making logic (retention vs. elimination): Existing technical solutions typically set a hard threshold (e.g., N50 < 20kb) to physically remove sample A (GCA_023224445.1) and sample B (GCA_023225465.1) from the database, resulting in the permanent loss of valuable clinical isolate information and undermining the sample diversity of epidemiological surveys.

[0059] This system solution: retain samples A and B completely, but label them with "reduced weight".

[0060] 2. Specific output format: The system attaches a "confidence weight label (WeightTag)" to each sample and outputs structured data as shown in Table 2: Table 2 Structured Data ; 3. Downstream application effect: When performing the "prevalence of Haemophilus influenzae virulence genes" statistics, the system no longer regards these two samples as "1 negative sample", but as "0.227 valid samples" and "0.220 valid samples".

[0061] Technical effects: The “soft retention” strategy of this application avoids the misleading statistical results of junk data (due to its extremely low weight, it has little impact on the overall average) and retains the sample metadata (such as separation time and location), achieving a perfect balance between data quality control and data resource preservation. (1) Maximize data utilization: This invention breaks the traditional binary filtering mode of “either keep or discard”. Through the “reduced weight retention” strategy, it effectively utilizes more than 90% of the draft data in the public database, especially protecting the scarce clinical isolate data. (2) Quantify and correct bias: By introducing the “benchmark-interference” calculation model, it is possible to quantitatively evaluate the impact of fragmentation on gene detection and effectively correct the “false negative” statistical bias caused by poor data quality. (3) Strong robustness of the mathematical model: This invention specifically introduces the min function as a boundary constraint, which effectively prevents the numerical overflow problem of extremely fragmented data and ensures the stability and accuracy of the algorithm under various extreme data conditions.

[0062] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A system for quality correction and optimization of draft data based on the complete genome of Haemophilus influenzae, characterized in that, The system includes: An automated data acquisition module is configured to automatically acquire the target microorganism's dataset and metadata from a public database. The dynamic benchmark building module is configured to construct standard feature vectors based on the target microorganism's dataset and metadata. The interference calculation module is configured to calculate the fragmentation interference coefficient of the dataset to be processed relative to the standard feature vector. ; The adaptive correction and optimization module is configured to adjust the value based on the interference coefficient. The gene detection results are adjusted for confidence level, and a preferred dataset with weighted labels is output; the adaptive adjustment and optimization configuration are based on the fragmentation interference coefficient. The confidence level of gene detection results from Haemophilus influenzae genome draft data is corrected, and a preferred dataset with weight labels is output. The confidence level correction includes confidence level correction for gene deletion and confidence level correction for gene presence. The value of the weight label is matched with the corrected confidence level for gene deletion.

2. The draft data quality correction and optimization system based on the complete genome of Haemophilus influenzae as described in claim 1, characterized in that, The dataset to be processed includes all sketch sequences with an assembly level of "Contig" or "Scaffold"; The metadata includes sequencing platform, submission year, and environmental parameters of the isolation source.

3. The draft data quality correction and optimization system based on the complete genome of Haemophilus influenzae as described in claim 1, characterized in that, The standard feature vector constructed by the dynamic benchmark construction module includes N50 and the contig number.

4. The draft data quality correction and optimization system based on the complete genome of Haemophilus influenzae as described in claim 1, characterized in that, The formula for calculating the fragmentation interference coefficient K by the interference calculation module is as follows: ； in, The ratio of the measured N50 of the sample to the baseline N50 reflects the loss of sequence continuity; min() indicates taking the minimum of the two values ​​within the parentheses when the number of sample fragments ( Exceeding the preset threshold ( When ), the system forces the value of this item to be 1, where ; + =1, It is a continuous weight. It is a fragmented weight.

5. The draft data quality correction and optimization system based on the complete genome benchmark of Haemophilus influenzae according to claim 4, characterized in that, The interference coefficient The range of values ​​for is 0 ≤ K ≤ 1. A higher value indicates a higher degree of data fragmentation.

6. The draft data quality correction and optimization system based on the complete genome of Haemophilus influenzae as described in claim 1, characterized in that, The confidence correction includes corrections for gene deletions and corrections for gene presences.

7. The draft data quality correction and optimization system based on the complete genome of Haemophilus influenzae as described in claim 6, characterized in that, The correction formula for gene deletion is: ； in The adjusted confidence level for gene deletion. This is the preset correction factor.

8. The draft data quality correction and optimization system based on the complete genome benchmark of Haemophilus influenzae according to claim 7, characterized in that, The correction factor The value is 0.

9.

9. The draft data quality correction and optimization system based on the complete genome of Haemophilus influenzae as described in claim 1, characterized in that, The target microorganism is Haemophilus influenzae. When the adaptive correction and optimization module performs gene presence confidence correction, it first identifies the length and sequence edge status of the contig containing the target gene, and then reduces the sequence integrity score of the target gene according to the numerical gradient of the fragmentation interference coefficient K. The sequence integrity score is the core reference indicator for the corrected confidence of gene presence. The module sets a contig length threshold and a sequence truncation judgment standard. If the length of the contig containing the target gene is lower than the threshold and there are base truncation features at the sequence edge, the influence weight of the fragmentation interference coefficient K on the sequence integrity score is increased. After the adaptive correction and optimization module completes the confidence correction of the gene detection results, it generates a unique confidence weight label for each dataset to be processed. The weight label is assigned based on the corrected confidence value of gene loss, directly mapping the credibility of the gene detection results of the dataset to be processed. The adaptive correction and optimization module is equipped with a data filtering and grading unit. This unit grades all datasets to be processed based on the confidence correction results. The grading results include high-confidence datasets, suspected false negative datasets, and low-confidence datasets. Different grades of datasets correspond to different downstream analysis access permissions. The criteria for judging suspected false negative datasets are: the initial gene detection result is negative, and the corrected confidence value for gene deletion is less than 0.

5. The adaptive correction and optimization module labels such datasets with suspected false negative tags and completely preserves their original data and metadata. The adaptive correction and optimization module is equipped with a structured data output unit. This unit integrates the sample number, original N50, fragmentation interference coefficient K, original gene detection results, system correction conclusions, and final weight labels of the datasets to be processed into a standardized structured table, forming the optimized dataset. The adaptive correction and optimization module is equipped with a dynamic weighting unit. This unit uses the final weight label values ​​of each sample in the optimized dataset as effective sample coefficients, applied to downstream statistical analyses such as the Haemophilus influenzae gene prevalence rate, presenting the statistical results in a weighted calculation manner, reducing the misleading effect of low-confidence datasets on the statistical results.

Images