A method for improving the quality of long-fragment capillary sequencing data

By adaptively adjusting multi-dimensional peak shape characteristics and channel advantage ratio, the problem of rigid peak boundary recognition and insufficient quality value correction in capillary long fragment sequencing technology is solved, realizing accurate restoration and improved reliability of sequencing data. It is applicable to sequencing data of different lengths and signal intensities, and provides stable and reliable quality assessment, especially in scientific research and clinical testing.

CN121709022BActive Publication Date: 2026-11-17THE FIRST RES INST OF MIN OF PUBLIC SECURITY
View PDF 2 Cites 0 Cited by

Patent Information

Application Number
CN202511848369.2
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2025-12-09
Publication Date
2026-11-17
Estimated Expiration
2045-12-09

AI Technical Summary

Technical Problem

Existing capillary long-fragment sequencing technologies suffer from problems such as rigid peak boundary recognition, limited quality value correction dimensions, and insufficient adaptability to long-fragment sequencing, leading to issues such as signal attenuation and peak overlap, which affect the accuracy and reliability of sequencing data.

Method used

By extracting multiple peak shape features and channel advantage ratios, the original quality value is adaptively adjusted to generate a corrected quality value that better reflects the reliability of base recognition. This process includes data reading and preprocessing, peak shape feature calculation, quality correction factor construction, and result visualization.

Benefits of technology

It achieves accurate restoration and in-depth optimization of sequencing data, improves the usability of sequencing data and the accuracy of analysis results, adapts to sequencing data characteristics of different fragment lengths and signal intensities, and is especially suitable for demanding scientific research and clinical testing scenarios.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN121709022B_ABST
    Figure CN121709022B_ABST
Patent Text Reader

Abstract

This invention discloses a method for quality correction of long capillary sequencing data. Addressing the core shortcomings of existing technologies, such as rigid peak boundary identification, limited quality value correction dimensions, poor adaptability to long-fragment sequencing, and lack of visual verification, this method solves the problem of inaccurate peak shape parameter calculation by dynamically identifying peak boundaries. It optimizes quality value correction by integrating multi-dimensional peak shape features such as half-width at half-maximum (HWHM), symmetry, and interference levels. Furthermore, it combines visualization analysis to achieve traceability of the correction process. This method can resolve the technical pain points of low accuracy, insufficient reliability, and limited effective length in long-fragment sequencing data, meeting the demand for high-quality sequencing data in various scenarios.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of biological detection technology, specifically to a method for improving the quality of capillary long-fragment sequencing data. Background Technology

[0002] Capillary electrophoresis is a highly efficient separation technique based on capillary columns. It separates DNA fragments according to molecular weight using an electric field, and combines this with fluorescence detection to acquire sequencing signals. It is the core separation method in capillary electrophoresis sequencing. Long fragments refer to DNA fragments with a length ≥500 bp (base pairs), which are typically used in gene cloning, chromosome fragment analysis, and detection of large gene mutations. Their sequencing process is susceptible to problems such as signal attenuation and peak overlap.

[0003] Existing capillary long-fragment sequencing technologies suffer from three major drawbacks: First, peak boundary identification methods are rigid, generally employing fixed windows or static thresholds to define peak ranges. This makes it difficult to adapt to the dynamic changes in signal intensity and peak shape distortion in long-fragment sequencing, leading to calculation errors in key peak shape parameters such as half-width at half-maximum (HWHM) and symmetry. Second, quality value correction relies on a single dimension, primarily depending on peak height and simple peak width indicators, neglecting important characteristics such as peak shape symmetry, internal interference, and inter-channel signal competition, including channel dominance ratio. This results in low accuracy in identifying low-quality bases affected by peak overlap or interference signals, and a significant deviation between the corrected quality value and the actual base reliability. Third, there is insufficient adaptability to long-fragment sequencing. When processing sequences longer than 500 bp, existing algorithms struggle to address issues such as peak height decrease, peak width expansion, and adjacent peak overlap caused by signal attenuation. Consequently, the Q30 ratio of long-fragment data is generally lower than that of short-fragment data, limiting the effective sequencing length. Furthermore, current technologies lack visualization support for the quality value correction process, making it difficult for users to intuitively evaluate the correction basis and hindering the traceability and in-depth analysis of experimental results. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention aims to provide a method for quality correction of capillary long-fragment sequencing data. By extracting and fusing multiple key peak shape features and channel dominance ratios, the original quality value is adaptively adjusted to generate a corrected quality value that more accurately reflects the reliability of base recognition.

[0005] To achieve the above objectives, the present invention adopts the following technical solution:

[0006] A method for quality correction of long-fragment capillary sequencing data includes the following steps:

[0007] S1. Data Reading and Preprocessing: Parse the AB1 format file generated by capillary electrophoresis sequencing, extract the raw signal intensity data of the four fluorescence detection channels, the identified base sequences, the position index of each base in the raw signal, and the corresponding raw quality value. Then, perform format verification and missing value processing on the extracted data, and perform baseline calibration on the signal data.

[0008] S2. Peak shape feature calculation: For each base in the base sequence, the range of the peak is determined in the original signal of the corresponding fluorescence detection channel according to its position index in the original signal. The search is carried out from the apex of the peak to both sides until the signal intensity drops below the background threshold or shows a significant rebound. Based on the determined peak range, the four morphological features of the peak are calculated: full width at half maximum (FWHM), symmetry, SNR (signal-to-noise ratio), and interference level.

[0009] While calculating the peak shape characteristics, the channel dominance ratio is calculated. The signal intensity of the four fluorescence detection channels at the peak position of the target base is extracted, and the ratio of the signal intensity of the fluorescence detection channel corresponding to the target base at the peak position of the target base to the maximum value of the signal intensity of the other three fluorescence detection channels at the peak position of the target base is calculated.

[0010] S3. Constructing the quality correction factor: Assign weights to five features: full width at half maximum (FWHM), symmetry, SNR, channel dominance ratio, and interference level; normalize each feature value to map it to the range of 0 to 1; multiply each normalized feature value by its corresponding weight and sum them to obtain the comprehensive score; subtract 1 from the comprehensive score to obtain the quality correction factor for the corresponding base.

[0011] Each base is calculated based on a quality correction factor, resulting in a corrected quality value, Corrected_Q, as follows:

[0012] Corrected_Q=Original_Q+(Correction_Factor×Scaling_Factor)

[0013] Wherein, Correction_Factor represents the quality correction factor, Original_Q is the original quality value, and Scaling_Factor represents the scaling factor;

[0014] S4. Results Output and Visualization: The corrected quality value is associated with the original base sequence and the position index of the base, and the base sequence file and quality comparison report are output. At the same time, a variety of visualization charts are generated, including a comparison chart of the original quality value and the corrected quality value, a base sequence peak chart with special markings of positions with significant changes in quality value, and an in-depth analysis chart of positions with significant decreases in quality value, which intuitively shows the peak characteristics and the reasons for correction.

[0015] Furthermore, in step S2, the background threshold is dynamically set to the larger of 10% of the peak intensity and the local minimum value.

[0016] Furthermore, in step S2, symmetry is obtained by comparing the distances from the peak apex to the left and right intersections of the full width at half maximum (FWHM), and the degree of interference is determined by detecting the secondary signal strength in the peak core region.

[0017] The present invention also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described method.

[0018] The present invention also provides a computer device, including a processor and a memory, wherein the memory is used to store a computer program; and the processor is used to execute the computer program to implement the above-described method.

[0019] The beneficial effects of this invention are as follows:

[0020] This application, through an innovative design of adaptive peak boundary recognition and multi-dimensional peak shape feature correction, breaks through the limitations of traditional sequencing quality assessment that relies solely on peak height and simple thresholds, achieving accurate restoration and in-depth optimization of sequencing data quality. Its core technological value lies not in pursuing numerical improvement in quality values, but in correcting the misjudgments of sequencing signal quality by traditional algorithms through scientific peak shape feature analysis, making the quality assessment results more consistent with the true reliability of actual sequencing data. Traditional algorithms often fail to fully consider the complex characteristics of peak width, symmetry, and internal interference, leading to some bases with latent quality defects being assigned excessively high quality values, resulting in false quality assessment results of "high numerical value but low reliability." This invention precisely defines the effective range of a peak by dynamically searching for peak boundaries from the peak apex to both sides. It constructs a correction factor by combining multi-dimensional peak shape parameters such as full width at half maximum (FWHM), symmetry, signal-to-noise ratio (SNR), channel dominance ratio (CDR), and interference level. This effectively identifies peak shape anomalies and signal interference that are ignored by traditional algorithms, and reasonably corrects the originally overestimated quality values. The corrected quality values ​​better reflect the actual reliability of base identification, providing a more rigorous quality basis for subsequent applications such as gene sequence analysis and gene mutation detection.

[0021] Meanwhile, this invention achieves comprehensive optimization of uncertain base identification. By jointly determining the channel dominance ratio and peak interference level, it can accurately screen out uncertain bases caused by signal overlap and channel competition in traditional methods, completely eliminating ambiguities in quality assessment. This allows users to clearly distinguish between reliable bases and potentially risky bases, significantly improving the usability of sequencing data and the accuracy of analysis results. Furthermore, this invention presents key information such as the quality value correction process, peak shape characteristic parameters, and reasons for quality adjustments in intuitive charts through visualized output. This breaks down the black-box barrier of traditional quality correction algorithms, allowing users to directly trace the core basis for each base quality value adjustment and understand the correlation between peak shape characteristics and quality assessment. This not only improves the credibility and interpretability of experimental results but also provides important reference for optimizing sequencing experiments and debugging instrument performance.

[0022] The technical solution of this invention has broad applicability, applicable to both conventional length sequencing fragments and long fragment sequencing data. It can accurately adapt to the characteristics of sequencing data with different fragment lengths and signal intensities, effectively addressing common problems in long fragment sequencing such as signal attenuation, peak distortion, and overlapping of adjacent peaks. It provides a stable and reliable quality assessment solution for various gene sequencing application scenarios, especially suitable for scientific research and clinical testing scenarios with stringent requirements for data reliability, and promotes the development of sequencing data quality assessment towards a more accurate, realistic, and transparent direction. Attached Figure Description

[0023] Figure 1 This is a flowchart illustrating the overall process of the method in Embodiment 1 of the present invention.

[0024] Figure 2 This refers to the sequence map file in Embodiment 2 of the present invention;

[0025] Figure 3 This is a schematic diagram of the original quality value results of the sequence in Embodiment 2 of the present invention;

[0026] Figure 4 This is a schematic diagram of the quality value results after sequence correction in Embodiment 2 of the present invention;

[0027] Figure 5 This is the sequence map of positions 650-700 in Embodiment 3 of the present invention;

[0028] Figure 6 This is a schematic diagram of the original mass value results at positions 650-700 in Embodiment 3 of the present invention;

[0029] Figure 7 This is a detailed analysis diagram of the high-quality peak at sequence position 101 in Embodiment 3 of the present invention;

[0030] Figure 8This is a schematic diagram of the corrected quality value results at positions 650-700 in Embodiment 3 of the present invention;

[0031] Figure 9 This is the overall fluorescence intensity spectrum of the sequence in Example 4 of the present invention;

[0032] Figure 10 This is a detailed analysis of the peak at sequence position 476 in Embodiment 4 of the present invention, along with a comparison of its quality values.

[0033] Figure 11 This is the original mass value spectrum of the region at position 650-700 in Embodiment 4 of the present invention;

[0034] Figure 12 This is the mass value spectrum after correction for position 650-700 in Embodiment 4 of the present invention. Detailed Implementation

[0035] The present invention will be further described below with reference to the accompanying drawings. It should be noted that this embodiment is based on the present technical solution and provides detailed implementation methods and specific operation processes, but the protection scope of the present invention is not limited to this embodiment.

[0036] Example 1

[0037] This embodiment provides a method for quality correction of long-fragment capillary sequencing data, such as... Figure 1 As shown, it includes the following steps:

[0038] S1. Data Reading and Preprocessing: Parse the AB1 format file generated by capillary electrophoresis sequencing, extract the raw signal intensity data of the four fluorescence detection channels, the identified base sequences, the position index of each base in the raw signal, and the corresponding raw quality value. Then, perform format verification and missing value processing on the extracted data to ensure data integrity, and perform baseline calibration on the signal data to remove the overall offset caused by the systematic error of the genetic analysis instrument.

[0039] S2. Peak Shape Feature Calculation: For each base in the base sequence, based on its position index in the original signal (corresponding field in the AB1 format file, enabling key field retrieval), the peak range is determined in the original signal of the corresponding fluorescence detection channel. The search proceeds outwards from the peak apex until the signal intensity drops below the background threshold or shows a significant rebound (the background threshold is dynamically set to the larger of 10% of the peak apex intensity and the local minimum). Based on the determined peak range, four morphological features are calculated: peak half-width at half-maximum (HWHM), symmetry, signal-to-noise ratio (SNR), and interference level. HWHM reflects the sharpness of the peak; symmetry is obtained by comparing the distances from the peak apex to the left and right intersections of the HWHM; and interference level is determined by detecting the secondary signal intensity in the core region of the peak.

[0040] While calculating the peak shape characteristics, the channel dominance ratio is calculated. The signal intensity of the four fluorescence detection channels at the peak position of the target base is extracted, and the ratio of the signal intensity of the fluorescence detection channel corresponding to the target base at the peak position of the target base to the maximum value of the signal intensity of the other three fluorescence detection channels at the peak position of the target base is calculated to quantify the specificity of the base signal.

[0041] S3. Constructing the Quality Correction Factor: Weights are assigned to five features: full width at half maximum (FWHM), symmetry, surface area ratio (SNR), channel advantage ratio, and interference level. These weights reflect the importance of each feature to the quality value. Each feature value is normalized using exponential functions and hyperbolic tangent functions to map it to the interval between 0 and 1, achieving adaptive penalty or reward. The normalized feature values ​​are multiplied by their corresponding weights and summed to obtain a comprehensive score. Subtracting 1 from the comprehensive score gives the quality correction factor for the corresponding base. This quality correction factor ranges from -1 to 1; a positive value indicates that the original quality value of the corresponding base is underestimated, while a negative value indicates that the original quality value of the corresponding base is overestimated.

[0042] Each base is calculated based on a quality correction factor, resulting in a corrected quality value, Corrected_Q, as follows:

[0043] Corrected_Q=Original_Q+(Correction_Factor×Scaling_Factor)

[0044] Wherein, Correction_Factor represents the quality correction factor, Original_Q is the original quality value, and Scaling_Factor represents the scaling factor. The scaling factor is an empirical constant used to control the correction magnitude, ensuring that the corrected quality value is within a reasonable range of 0-60.

[0045] S4. Results Output and Visualization: The corrected quality values ​​are correlated with the original base sequences and base position indices, and the base sequence files and quality comparison reports are output in FASTA, TXT, and other formats. Various visualization charts can also be generated, including a comparison chart of the original and corrected quality values, a peak chart of base sequences with specially marked positions of significant quality value changes, and an in-depth analysis chart of positions where the quality value drops sharply, intuitively displaying peak characteristics and the reasons for correction.

[0046] Example 2

[0047] This embodiment aims to provide an application example of using the method described in Example 1 to accurately correct the quality of short data segments and identify and correct hidden low-quality bases.

[0048] For short-fragment sequencing data of routine PCR products, traditional sequencing quality assessment algorithms rely solely on signal intensity and simple noise ratio scoring. This can easily misclassify bases with internal peak interference or extremely low signal-to-noise ratios but acceptable signal intensity as high-quality bases. The core objective of this embodiment is to verify the ability of the method in Example 1 to identify hidden low-quality bases in short-fragment data. Through multi-dimensional peak shape feature analysis, it corrects the scoring bias of traditional algorithms, outputting quality values ​​that more closely reflect the reliability of true base identification, thus providing data support for subsequent precise experiments such as molecular cloning verification.

[0049] Using default core parameters, including a peak search window of 20 bp, a signal enhancement factor of 1.5, and an enhancement region initial ratio of 0.8, the fluorescence signals of the A, C, G, and T channels, base sequences, the position index of each base in the original signal, and the original quality values ​​were extracted from the AB1 format file. After format verification and baseline calibration, four peak shape characteristics—full width at half maximum (FWHM), symmetry, signal-to-noise ratio (SNR), and interference level—were calculated for each base's peak position. Simultaneously, the channel dominance ratio was calculated to quantify the signal specificity. The weights were then fixed as shown in Table 1.

[0050] Table 1

[0051]

[0052] The five eigenvalues ​​are weighted and summed according to their respective weights in Table 1 to generate a quality correction factor for each base, ultimately yielding the corrected quality value for each base.

[0053] Corrected_Q=Original_Q+Correction_Factor×25

[0054] The results show that the original average quality value was 46, and after correction, the average quality value decreased to 33.4, a reduction of 12.6. The Q30 ratio decreased from 78% to 63.9%, a reduction of 14.1%, and the Q20 ratio decreased from 89.1% to 72.5%, a reduction of 16.6%. The number of low-quality bases increased from 38 to 139, an increase of 101, while the number of uncertain bases identified decreased from 405 to 0, achieving a 100% improvement rate. This demonstrates that the method in Example 1 successfully identified low-quality bases ignored by traditional algorithms. The top 5 bases with the most significant quality drops (positions 246, 262, 269, 271, and 279) all exhibit infinite half-width at half-maximum (FWHM = inf), interference level = 1, and extremely low SNR (0.5-3.5). This indicates that the peak shape of these bases cannot be accurately defined, and there is a problem of serious internal interference or the signal being submerged by noise. Traditional algorithms give a high score of 62 points simply because the signal strength is high. However, this embodiment uses multi-dimensional feature fusion to reasonably correct it to about 45 points, which is more in line with the actual recognition reliability.

[0055] The visualization results further validated the rationality of the correction.

[0056] Figure 2 This is the sequence map file for this embodiment. The fluorescence signals of the four channels A (green), C (blue), G (black), and T (red) are clearly presented in the figure, with the fluorescence intensity reaching up to 5000-6000. Figure 3 The image shows the original mass value in this embodiment. Figure 4 The image shows a comparison of the corrected quality values ​​in this embodiment. The horizontal axis represents the position of the spectrum, and the vertical axis represents the quality value. The threshold lines Q30 (green dashed line) and Q20 (yellow dashed line) are clearly marked in the image. In the original quality value histogram, areas at positions 246, 262, 269, 271, and 279 are all green, perfectly matching the data in the sequence results. After correction, the histograms at these positions turn yellow, corresponding to the corrected Q=45 result, accurately reflecting the true risk of base identification.

[0057] Example 3

[0058] This embodiment aims to use the method described in Example 1 to perform targeted correction of peak diffusion and signal attenuation before optimizing long-segment data for biological experiments.

[0059] This embodiment aims to verify the ability of the method in Example 1 to assess the quality of long-fragment sequencing samples that have not undergone experimental condition optimization. The sample is an unoptimized long-fragment sample in AB1 format, which contains long-fragment sequencing data of 809 bp in length. The experimental conditions such as PCR amplification temperature and sequencing reaction system have not been optimized, and it is a typical representative of samples with defects in the original signal peak shape in long-fragment sequencing.

[0060] This embodiment employs dedicated analysis parameters adapted for long segments. The peak search window is set to 25 to match the peak shape diffusion characteristics of long segments. The initial signal enhancement ratio is 0.6, and the enhancement factor is 1.8 to attempt to compensate for the signal attenuation at the end. The quality correction weight configuration is consistent with Table 1 to ensure accurate multi-dimensional evaluation of peak shape characteristics.

[0061] First, the AB1 format file is read, and the four-channel raw fluorescence signal, base position, base position index, and raw mass value are extracted. This can be seen from the regional sequence map (e.g., ...). Figure 5 As shown in the figure, the spectrum covers the entire area, and the peak fluorescence intensity can reach 3500. The signal in the target area of ​​650-700 shows obvious diffusion characteristics, and the interweaving of the four-channel signals leads to insufficient separation.

[0062] Combining the original mass value spectrum at positions 650-700, the original mass values ​​in this region are distributed in the range of 0-60. Although some positions are close to the Q30 reference line, the overall fluctuations are large, showing a trend of mass instability. Figure 6 As shown.

[0063] The signal was then adaptively enhanced, and the generated sequence spectrum further verified the peak shape defect in the 500+ region. The signal peak at positions 650-700 showed significant broadening, and the peak outlines corresponding to some bases were blurred, contrasting sharply with the high-quality peaks. For example... Figure 7 As shown, the fluorescence intensity spectrum of the high-quality peak at position 101, with a sharp peak shape, a full width at half maximum (FWHM) of only 2, a symmetry of 0.67, a signal-to-noise ratio of 0.7, an original mass value of 62, and a corrected mass value of 57, with minimal mass attenuation, is a typical normal peak shape sample.

[0064] Simultaneously generated corrected quality value spectrum at positions 650-700, as shown below. Figure 8 As shown, the corrected quality values ​​in this region are still concentrated in the 0-60 range, but the overall distribution is more gradual than the original quality values. The quality values ​​in some locations have dropped significantly, which intuitively reflects the real impact of peak-shaped defects on quality.

[0065] The core results of this embodiment clearly demonstrate the inherent quality defects of unoptimized long fragment samples. The original average quality value of the samples was 44.6. After peak shape correction in this embodiment, the average quality value decreased to 30.4, the Q30 ratio decreased from 78% to 62.1%, and the number of low-quality bases (<10) increased from 71 to 148. This decrease in values ​​is not due to method error, but rather a precise correction of the artificially inflated original quality value, especially evident in the terminal region of long fragments. The average quality value at the terminal decreased from the original 22.7 to 10.6 after correction, and the Q30 ratio at the terminal dropped sharply from 36.5% to 8%, indicating that the peak shape defects at the terminal of long fragments in unoptimized samples were so severe that quality could not be improved through correction.

[0066] Although this embodiment reduced the number of uncertain base recognitions from 209 to 0, the quality of the corrected long fragment remained at a low level. This result objectively reflects the inherent quality defects of long fragment sequencing data under unoptimized experimental conditions, and also confirms the core value of the method in Example 1—not to blindly improve the quality value, but to restore the true quality level of the data by accurately analyzing peak shape characteristics, providing a clear reference for defect areas for subsequent biological experiment optimization.

[0067] Example 4

[0068] This embodiment aims to compare the quality of long fragment data before and after optimization of biological experiments using the method of Example 1, so as to achieve a quantitative evaluation of the sequencing experiment effect.

[0069] To verify the quality correction effect of the peak shape identification algorithm optimized for long-fragment sequencing data, this embodiment takes long-fragment sequencing data optimized for biological experiments as the research object. By quantitatively comparing the core quality indicators, terminal region quality performance and key base position quality changes before and after correction, the method of Example 1 is systematically evaluated to improve the accuracy and reliability of sequencing data, providing technical support for the subsequent analysis of long-fragment sequencing data.

[0070] Read long-fragment sequencing data from the sample, along with accompanying maps including overall sequence fluorescence intensity maps (e.g.) Figure 9 (as shown), detailed peak analysis and quality value comparison (e.g.) Figure 10 As shown), the original quality value spectrum of the region at position 650-700 (as shown) Figure 11 (as shown) and the mass value spectrum after correction at position 650-700 (as shown) Figure 12 (As shown). After processing the raw sequencing data according to the method in Example 1, quality indicators before and after correction were extracted for quantitative comparative analysis.

[0071] A quantitative comparison of core quality indicators before and after correction shows that the average quality value decreased from the original 48 to the corrected 32.5, the Q30 proportion decreased from 83.5% to 68.7%, the Q20 proportion decreased from 89.1% to 78.3%, the number of low-quality bases increased from 67 to 140, while the number of uncertain bases identified decreased from the original 317 to 0, resulting in an improvement rate. Figure 10 As can be seen, some peak shape parameters in the original data were abnormal. Although some overall indicators decreased after correction, uncertain bases were completely eliminated, improving the uniqueness of sequence identification. Figure 9 The overall fluorescence intensity map of the sequence shown intuitively displays the fluorescence intensity distribution across the entire region, providing visual support for overall quality assessment.

[0072] Quality analysis of the terminal regions of long fragments showed that the average terminal quality value decreased from the original 26 to the corrected 12.9, and the terminal Q30 proportion decreased from 43% to 16%. Figure 11 The original quality value spectrum shown is as follows: Figure 12 As shown in the corrected quality value spectrum, this region is a typical example of quality decay in long fragments. Although the quality value decreased after correction, the stability of peak shape recognition was improved through special evaluation at the end of the algorithm and signal enhancement technology. The distribution trend of fluorescence intensity in the regional sequence spectrum also confirms this conclusion.

[0073] Screening of base positions with significant quality value drops revealed five key positions with a quality value decrease of 22: position 62 (base G, original Q49 → corrected Q27), position 68 (base G, original Q54 → corrected Q32), position 185 (base C, original Q59 → corrected Q37), position 441 (base C, original Q62 → corrected Q40), and position 603 (base T, original Q58 → corrected Q36). Combined with peak shape parameter analysis in the detailed peak analysis diagram, the quality value decrease at these positions is mainly related to a larger peak half-width (FWHM) (19.8-22), poorer symmetry (0.4-0.11), lower signal-to-noise ratio (0.1-2.2), and channel interference (1). The above examples demonstrate that the method can eliminate uncertain bases and optimize the consistency of peak shape identification.

[0074] Using the method described in Example 1 to perform quality correction on long-fragment sequencing data, although some overall quality indicators and end-region quality indicators decreased, the number of uncertain base recognitions was completely eliminated. Furthermore, through multi-dimensional peak shape evaluation and end-specific optimization, the recognition stability and reliability of long-fragment sequencing data were improved. This provides an effective technical solution for the quality correction of long-fragment sequencing data, supporting the accuracy of subsequent sequence analysis, and is particularly suitable for sequencing applications sensitive to uncertain bases.

[0075] For those skilled in the art, various corresponding changes and modifications can be made based on the above technical solutions and concepts, and all such changes and modifications should be included within the protection scope of the claims of this invention.

Claims

1. A method for quality correction of capillary long fragment sequencing data, the method comprising: Includes the following steps: ​ S1. Data Reading and Preprocessing: Parse the AB1 format file generated by capillary electrophoresis sequencing, extract the raw signal intensity data of the four fluorescence detection channels, the identified base sequences, the position index of each base in the raw signal, and the corresponding raw quality value. Then, perform format verification and missing value processing on the extracted data, and perform baseline calibration on the signal data. S2. Peak shape feature calculation: For each base in the base sequence, the range of the peak is determined in the original signal of the corresponding fluorescence detection channel according to its position index in the original signal. The search is carried out from the apex of the peak to both sides until the signal intensity drops below the background threshold or shows a significant rebound. Based on the determined peak range, the four morphological features of the peak are calculated: full width at half maximum (FWHM), symmetry, SNR (signal-to-noise ratio), and interference level. Symmetry is obtained by comparing the distances from the peak apex to the left and right intersection points of the full width at half maximum (FWHM), and the degree of interference is determined by detecting the secondary signal strength in the core region of the peak. While calculating the peak shape characteristics, the channel dominance ratio is calculated. The signal intensity of the four fluorescence detection channels at the peak position of the target base is extracted, and the ratio of the signal intensity of the fluorescence detection channel corresponding to the target base at the peak position of the target base to the maximum value of the signal intensity of the other three fluorescence detection channels at the peak position of the target base is calculated. S3. Construct quality correction factors: assign weights to five features: full width at half maximum (FWHM), symmetry, SNR, channel advantage ratio, and interference level. Normalize each feature value to map it to the interval between 0 and 1; multiply each normalized feature value by its corresponding weight and sum them to obtain the comprehensive score; subtract 1 from the comprehensive score to obtain the quality correction factor of the corresponding base. Each base is calculated based on a quality correction factor, resulting in a corrected quality value, Corrected_Q, as follows: Corrected_Q=Original_Q+(Correction_Factor×Scaling_Factor) Wherein, Correction_Factor represents the quality correction factor, Original_Q is the original quality value, and Scaling_Factor represents the scaling factor; S4. Results Output and Visualization: The corrected quality value is associated with the original base sequence and the position index of the base, and the base sequence file and quality comparison report are output. At the same time, a variety of visualization charts are generated, including a comparison chart of the original quality value and the corrected quality value, a base sequence peak chart with special markings of positions with significant changes in quality value, and an in-depth analysis chart of positions with significant decreases in quality value, which intuitively shows the peak characteristics and the reasons for correction.

2. The method of claim 1, wherein, In step S2, the background threshold is dynamically set to the larger of 10% of the peak intensity and the local minimum value.

3. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the method described in any one of claims 1-2.

4. A computer device, comprising: It includes a processor and a memory, the memory being used to store a computer program; the processor being used to execute the computer program to implement the method of any one of claims 1-2.

Citation Information

Patent Citations

  • Electrophoresis mobility offset correction method and system

    CN117743783A

  • Capillary electrophoresis mobility correction method for Sanger sequencing

    CN118136119A