Nucleic acid optimization detection method and system based on nonlinear amplification modeling

Abstract

The present application relates to the technical field of nucleic acid detection, and discloses a nucleic acid optimization detection method and system based on nonlinear amplification modeling, which comprises the following steps: obtaining an initial fluorescence time sequence of a sample to be detected, generating an initial signal matrix and constructing an amplification kinetics feature set, inputting the amplification kinetics feature set into a multi-target regression model, obtaining a nonlinear deviation correction factor and generating a correction coefficient vector, adjusting a detection threshold according to the correction coefficient vector, constructing an optimized detection parameter set, inputting the optimized detection parameter set into a preset nucleic acid detection model, and outputting a nucleic acid abundance result set including a quantitative signal and a decision value, so as to improve the consistency of detection results under the conditions of large nonlinear deviation of nucleic acid amplification and weak low-abundance target signal.

Description

Technical Field

[0001] This invention relates to the field of nucleic acid detection technology, and more specifically, to a nucleic acid optimization detection method and system based on nonlinear amplification modeling. Background Technology

[0002] Nucleic acid testing is widely used in laboratory and field testing. It usually relies on temperature-controlled amplification, fluorescence monitoring and signal acquisition. Limited by sample quality, instrument size and detection delay, existing methods rely on amplification curve fitting, signal threshold determination, periodic data smoothing and multi-channel selection, and are suitable for conditions where the sample is intact, the signal intensity is sufficient and the environment is stable.

[0003] In actual testing, there are two types of technical factors in the nucleic acid amplification process: nonlinear dynamic changes and weakening of low-abundance target signals. These factors affect amplification curve fitting, signal determination, and data consistency, resulting in reduced repeatability and comparability of test results. Therefore, the technical problem that needs to be solved is: how to ensure the consistency of test results under the condition of large nonlinear deviation in nucleic acid amplification and weakening of low-abundance target signals.

[0004] In view of this, the present invention proposes a nucleic acid optimization detection method and system based on nonlinear amplification modeling to solve the above problems. Summary of the Invention

[0005] To overcome the aforementioned shortcomings of existing technologies, this invention provides a nucleic acid optimization detection method and system based on nonlinear amplification modeling.

[0006] To achieve the above objectives, the present invention provides the following technical solution: Firstly, it provides a nucleic acid optimization detection method based on nonlinear amplification modeling, including: Obtain the initial fluorescence time series of the sample to be tested, generate an initial signal matrix based on the initial fluorescence time series, and construct an amplification kinetics feature set based on the initial signal matrix; The amplification kinetics feature set is input into a multi-objective regression model to obtain the nonlinear bias correction factor for each sample, and a correction coefficient vector is generated based on the nonlinear bias correction factor. The detection threshold is adjusted based on the correction coefficient vector to construct an optimized detection parameter set, which is used for the enhancement processing of low-abundance nucleic acid signals. The optimized set of detection parameters is input into the preset nucleic acid detection model, and the set of nucleic acid abundance results is output. The set of nucleic acid abundance results includes the quantized signal and judgment value after nonlinear deviation correction for each sample.

[0007] Secondly, a nucleic acid optimization detection system based on nonlinear amplification modeling is provided, which is used to implement the aforementioned nucleic acid optimization detection method based on nonlinear amplification modeling, including: Feature construction module: used to obtain the initial fluorescence time series of the sample to be tested, generate an initial signal matrix based on the initial fluorescence time series, and construct an amplification kinetic feature set based on the initial signal matrix; Correction generation module: used to input the amplification kinetics feature set into the multi-objective regression model, obtain the nonlinear bias correction factor for each sample, and generate a correction coefficient vector based on the nonlinear bias correction factor; Parameter optimization module: used to adjust the detection threshold according to the correction coefficient vector, thereby constructing an optimized detection parameter set, which is used for the enhancement processing of low-abundance nucleic acid signals; The results output module is used to input the optimized detection parameter set into the preset nucleic acid detection model and output a set of nucleic acid abundance results, which includes the quantized signal and judgment value after nonlinear deviation correction for each sample.

[0008] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention acquires the initial fluorescence time series of the sample to be tested, generates an initial signal matrix based on the initial fluorescence time series, and then constructs an amplification kinetic feature set based on the initial signal matrix. First, the nonlinear changes during nucleic acid amplification are incorporated into a unified characterization. Then, the amplification kinetic feature set is input into a multi-objective regression model to obtain the nonlinear deviation correction factor for each sample and generate a correction coefficient vector, enabling different samples to form an adjustable deviation correction basis during the amplification stage. Next, the detection threshold is adjusted based on the correction coefficient vector to construct an optimized detection parameter set for low-abundance nucleic acid signal enhancement processing, ensuring that weak signal samples obtain a detection aperture matching their amplification deviation during the judgment process. Finally, the optimized detection parameter set is input into a preset nucleic acid detection model, outputting a set of nucleic acid abundance results including the quantized signal after nonlinear deviation correction and the judgment value. This improves the consistency of detection results under conditions of large nonlinear deviation in nucleic acid amplification and weakened low-abundance target signals. Attached Figure Description

[0009] Figure 1 This is a schematic diagram of the nucleic acid optimization detection method based on nonlinear amplification modeling in this invention; Figure 2 This is a schematic diagram of the nucleic acid optimization detection system based on nonlinear amplification modeling in this invention. Figure 3 This is a flowchart illustrating the method for generating amplified state codes in this invention. Figure 4 This is a schematic flowchart of the method for generating a gated reaction rate table in this invention. Detailed Implementation

[0010] To make the objectives, technical solutions, and advantages of the present invention clearer, the present invention will be further described in detail below with reference to specific embodiments and the accompanying drawings. In the following detailed description, many specific details are set forth to provide a thorough understanding of the exemplary embodiments described. However, it will be apparent to those skilled in the art that the described embodiments may be practiced without some or all of these specific details. In other exemplary embodiments, well-known structures have not been described in detail to avoid unnecessarily obscuring the concepts of this disclosure. It should be understood that the specific embodiments described herein are merely illustrative of the present invention and are not intended to limit the present invention. Furthermore, the various aspects described in the embodiments may be combined arbitrarily without conflict.

[0011] The user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data must comply with relevant regulations.

[0012] Example 1 Figure 1 This disclosure illustrates a nucleic acid optimization detection method based on nonlinear amplification modeling, provided in at least one embodiment, comprising: S10: Obtain the initial fluorescence time series of the sample to be tested, generate an initial signal matrix based on the initial fluorescence time series, and construct an amplification kinetics feature set based on the initial signal matrix; In this embodiment, the initial fluorescence time series refers to the fluorescence intensity records collected by the sample to be tested in multiple consecutive cycles during the initial stage of nucleic acid amplification. The initial signal matrix refers to a two-dimensional data structure arranged according to the sample number and cycle number. The amplification kinetics feature set refers to a data set formed by the cycle fluorescence table, fluorescence growth ratio table, growth ratio difference table, and amplification kinetics fragment. The amplification kinetics feature set is used to characterize the signal change relationship of the sample to be tested between amplification initiation, amplification acceleration, and amplification plateau.

[0013] Understandably, the process involves first organizing the continuously acquired initial fluorescence time series into an initial signal matrix, and then constructing an amplification kinetics feature set from this initial signal matrix. The goal is to convert the single-point fluorescence records from the initial stage of nucleic acid amplification into a data structure with cycle sequence and sample attribution. This allows subsequent processing to identify amplification stage boundaries from the changes between adjacent cycles. Compared to directly fitting the amplification curve as a whole, this method preserves the subtle growth relationships of low-abundance nucleic acid signals in the early stages of amplification. The initial fluorescence time series records the fluorescence intensity of each sample in each cycle during the early stages of nucleic acid amplification. By organizing these continuously acquired single-point fluorescence records into a matrix structure and arranging them by sample and cycle number, the fluorescence changes in each cycle can be continuously analyzed, thereby identifying the effective growth stage boundaries during amplification. "Continuous" means that the number of cycles in which the fluorescence signal shows a stable upward or downward trend in adjacent cycles is usually two or three or more, avoiding misjudgments caused by single-point fluctuations or occasional noise. This processing method aligns with international standards such as MIQE (Minimum Information for Publication of Quantitative Real-Time PCR). The principles emphasized by Experiments are consistent with those of "reliability of amplification curves, continuity of initial fluorescence data, and effective amplification judgment".

[0014] For example, the samples to be tested include sample A and sample B. The fluorescence intensities of sample A in the FAM channel from cycle 1 to cycle 6 are 10, 12, 17, 29, 47, and 60, respectively. The fluorescence intensities of sample B in the FAM channel from cycle 1 to cycle 6 are 9, 10, 13, 20, 31, and 39, respectively. The initial signal matrix is ​​represented as [[10,12,17,29,47,60],[9,10,13,20,31,39]]. The rows of this initial signal matrix correspond to the sample number, and the columns correspond to the cycle number. Subsequent cycle fluorescence tables, fluorescence growth ratio tables, and growth ratio difference tables are all generated based on this initial signal matrix.

[0015] A set of amplification kinetic features is constructed based on the initial signal matrix, including: The initial signal matrix is ​​rearranged according to the sample number and cycle number to obtain the cyclic fluorescence table, which records the fluorescence intensity of the same sample in consecutive cycles. Based on the fluorescence intensity relationship between adjacent cycles in the cyclic fluorescence table, a fluorescence growth ratio table is obtained, in which the fluorescence growth ratio table records the fluorescence intensity ratio between adjacent cycles; Based on the growth change relationship between adjacent cycles in the fluorescence growth ratio table, a growth ratio difference table is obtained, in which the growth ratio difference table records the difference between adjacent fluorescence growth ratios; The amplification kinetics fragment is obtained by extracting the cyclic fluorescence table based on the sign transition position in the growth ratio difference table. Based on the amplification kinetics fragment, fluorescence growth ratio table, and growth ratio difference table, the set of amplification kinetics characteristics is obtained.

[0016] In this embodiment, the initial signal matrix is ​​rearranged according to the sample number and cycle number to obtain the cyclic fluorescence table. The cyclic fluorescence table arranges the fluorescence intensity of the same sample in consecutive cycles as a searchable record. Taking the above initial signal matrix as an example, the cyclic fluorescence table is shown in Table 1 below: Table 1: Cyclic Fluorescence Table In this embodiment, a fluorescence growth ratio table is obtained based on the fluorescence intensity relationship between adjacent cycles in the cyclic fluorescence table, and a growth ratio difference table is obtained based on the growth change relationship between adjacent cycles in the fluorescence growth ratio table. The fluorescence growth ratio table reflects the amplification relationship between adjacent cycles, and the growth ratio difference table reflects the direction of change of the amplification relationship itself. Taking sample A as an example, the fluorescence growth ratio table corresponding to sample A is 1.20, 1.42, 1.71, 1.62, 1.28, and the corresponding growth ratio difference table is 0.22, 0.29, -0.09, -0.34. This growth ratio difference table is used to identify the position where the amplification growth changes from enhancement to weakening.

[0017] It should be noted that the amplification kinetics fragment is obtained by extracting the cycle fluorescence table based on the sign inflection points in the growth ratio difference table. These sign inflection points are not directly collected data, but rather stage boundaries obtained from the changes in the difference between adjacent fluorescence growth ratios. Taking sample A as an example, the growth ratio difference table shows a sign inflection point between 0.29 and -0.09. Combining this with the records from cycles 2 to 5 extracted from the cycle fluorescence table, the amplification kinetics fragment is shown below: {(A,2,12),(A,3,17),(A,4,29),(A,5,47)}.

[0018] Based on the sign transition positions in the growth ratio difference table, the cyclic fluorescence table is extracted to obtain the amplification kinetics fragment, including: Based on the sign change relationship between adjacent items in the growth ratio difference table, a candidate inflection point table is obtained. Based on the candidate inflection point table, the fluorescence growth ratio table is retrieved to obtain candidate growth ratio fragments. Based on the continuous rising and falling segments in the candidate growth ratio fragments, the amplification start position and amplification plateau position are obtained. The amplification start position corresponds to the starting cycle number of the continuous rising segment, and the amplification plateau position corresponds to the ending cycle number of the continuous falling segment. The amplification kinetics fragment is obtained by extracting cyclic fluorescence tables based on the amplification initiation and plateau positions.

[0019] In this embodiment, "continuous" means that in the initial fluorescence sampling cycle, adjacent fluorescence intensities show an increasing or decreasing trend for at least two or three consecutive cycles. This ensures that the identified amplification stage boundary is a stable and genuine amplification signal, rather than a misjudgment caused by single-point fluctuations or occasional noise. By organizing the continuously acquired initial fluorescence time series into an initial signal matrix, and then constructing a cycle fluorescence table, a fluorescence growth ratio table, and a growth ratio difference table, an amplification kinetics fragment can be formed, thereby accurately locating the amplification initiation position and the amplification plateau position.

[0020] In this embodiment, a candidate inflection point table is obtained based on the sign change relationship between adjacent items in the growth ratio difference table. Then, a candidate growth ratio fragment is obtained by searching the fluorescence growth ratio table based on the candidate inflection point table. This process locates the candidate boundary of amplification changes using the growth ratio difference table and verifies whether an amplification growth segment exists near the candidate boundary using the fluorescence growth ratio table. Taking sample A as an example, the candidate inflection point table is represented as {(A,4)}, and the candidate growth ratio fragment is represented as: {(A,2,1.42),(A,3,1.71),(A,4,1.62),(A,5,1.28)}.

[0021] In this embodiment, the amplification start position and amplification plateau position are obtained based on the continuous rising segment and continuous falling segment in the candidate growth ratio fragment. The continuous rising segment corresponds to the interval in which the amplification reaction gradually enters the effective amplification range, and the continuous falling segment corresponds to the interval in which the amplification reaction approaches the plateau stage. Taking the above-mentioned candidate growth ratio fragment as an example, 1.42 to 1.71 forms a continuous rising segment, and 1.71 to 1.28 forms a continuous falling segment. The amplification start position is the second cycle, and the amplification plateau position is the fifth cycle.

[0022] In this embodiment, the cyclic fluorescence table is extracted based on the amplification start position and the amplification plateau position to obtain the amplification kinetics fragment. Continuing to use the data of the above sample A, the amplification start position is the 2nd cycle, the amplification plateau position is the 5th cycle, and the corresponding amplification kinetics fragment in the cyclic fluorescence table is {(A,2,12),(A,3,17),(A,4,29),(A,5,47)}. This amplification kinetics fragment is used to subsequently construct the amplification kinetics feature set and to focus subsequent model processing on the effective stage related to amplification deviation.

[0023] S20: Input the set of amplification kinetics features into the multi-objective regression model to obtain the nonlinear deviation correction factor for each sample, and generate a correction coefficient vector based on the nonlinear deviation correction factor; In this embodiment, after the amplification kinetics feature set is input into the multi-objective regression model, the multi-objective regression model outputs an amplification efficiency estimate and a background drift estimate based on the field information of the same amplification process. The amplification efficiency estimate is used to characterize the effective amplification state of the sample amplification reaction, and the background drift estimate is used to characterize the signal offset caused by baseline fluctuations during fluorescence acquisition. The nonlinear deviation correction factor is obtained by combining the amplification efficiency estimate, the background drift estimate, and the reaction rate field table. The correction coefficient vector is arranged according to the sample number and fluorescence channel number to form the nonlinear deviation correction factor corresponding to each sample.

[0024] In this embodiment, the multi-objective regression model is a neural network structure containing a shared coding layer and multiple output branches. The shared coding layer is used to map the reaction rate field table in the amplification kinetics feature set to a unified amplification state code, ensuring that the dynamic information of the amplification initiation stage and the amplification acceleration stage can be represented together. Subsequently, different output branches generate amplification efficiency estimates and background drift estimates respectively. The two are combined with the reaction rate field table to calculate the nonlinear bias correction factor for each sample.

[0025] Understandably, the aforementioned set of amplification kinetics features already retains amplification kinetics fragments, fluorescence growth ratio tables, and growth ratio difference tables. Therefore, the multi-objective regression model does not judge the sample status solely based on a single fluorescence intensity, but simultaneously utilizes the amplification stage position, growth change relationship, and background drift relationship for bias correction. Continuing with the data from sample A, the amplification kinetics fragments are {(A,2,12),(A,3,17),(A,4,29),(A,5,47)}, with corresponding fluorescence growth ratios of 1.42, 1.71, and 1.62, and corresponding growth ratio differences of 0.29 and -0.09. This data is used to form the subsequent reaction rate field table.

[0026] The amplification kinetics feature set is input into a multi-objective regression model to obtain the nonlinear bias correction factor for each sample, including: Based on the amplification kinetic fragments in the amplification kinetics feature set, a fragment feature table is obtained. Based on the fragment feature table, fluorescence growth ratio table, growth ratio difference table, and amplification initiation position, a reaction rate field table is obtained. Input the reaction rate field table into the shared coding layer of the multi-objective regression model to obtain the amplified state code; Amplification efficiency estimates and background drift estimates are generated based on the amplification state coding, where the amplification efficiency estimates and background drift estimates are derived from different output branches of the multi-objective regression model. Based on the amplification efficiency estimate, background drift estimate, and reaction rate field table, the nonlinear bias correction factor for each sample is obtained.

[0027] In this embodiment, the reaction rate field table provides dynamic characteristics such as fluorescence intensity, fluorescence growth ratio, and growth ratio difference for each sample in each cycle after the amplification initiation position. It is the main input of the multi-objective regression model. The amplification efficiency estimate is calculated from the amplification state representation encoded by the shared coding layer, reflecting the effective amplification degree of the sample in each cycle. The background drift estimate is also generated based on the amplification state encoding, but focuses on capturing the baseline drift or background noise impact on the fluorescence signal in the early stage of amplification. The nonlinear deviation correction factor combines the amplification efficiency estimate, background drift estimate, and reaction rate field table to form a correction coefficient that integrates the actual amplification gain and background drift information. This coefficient is used to adjust the detection threshold and construct an optimized detection parameter set.

[0028] In this embodiment, the fragment feature table is derived from the amplification kinetic fragment in the amplification kinetic feature set. The fragment feature table records the sample number, cycle number, fluorescence intensity and channel number within the amplification kinetic fragment. The reaction rate field table introduces the fluorescence growth ratio table, growth ratio difference table and amplification start position on the basis of the fragment feature table, forming a model input field that can describe the signal growth state after amplification is initiated.

[0029] For example, continuing with the data from sample A, the reaction rate field table is shown below: Table 2: Reaction Rate Field Table In this embodiment, the shared coding layer of the multi-objective regression model encodes the reaction rate field table to obtain the amplification state code. The shared coding layer maps the fields of the amplification initiation segment and the amplification acceleration segment to the same state expression, so that the amplification efficiency estimate and the background drift estimate can be generated based on the same amplification state code, avoiding the source of inconsistent bias in low abundance signals when the amplification efficiency judgment and the background drift judgment are modeled separately.

[0030] Furthermore, the amplification efficiency estimate and the background drift estimate are derived from different output branches of the multi-objective regression model. The amplification efficiency estimate reflects the amplification efficiency corresponding to the growth relationship in the reaction rate field table, while the background drift estimate reflects the offset introduced by the fluorescence background change in the same reaction rate field table. Based on the amplification efficiency estimate, the background drift estimate, and the reaction rate field table, a nonlinear bias correction factor is obtained for each sample, so that the correction result is simultaneously constrained by the amplification state and the background drift state.

[0031] like Figure 3 As shown, inputting the reaction rate field table into the shared coding layer of the multi-objective regression model yields the amplified state coding, including: Based on the amplification start position in the reaction rate field table, the gating position column is obtained. Based on the gating position column, the reaction rate field table is extracted to obtain the gated reaction rate table. By inputting the gated reaction rate table into the shared coding layer of the multi-objective regression model, the initial amplification state coding is obtained, which includes the amplification initiation segment coding and the amplification acceleration segment coding. The initial amplification status code is corrected according to the cyclic numbering order to obtain the amplification status code.

[0032] In this embodiment, the gated position column is formed by the amplification start position in the reaction rate field table. The gated position column is used to determine the cycle number of each sample entering the effective amplification stage. The gated reaction rate table reads the field content after the amplification start position from the reaction rate field table according to the gated position column, so that the shared coding layer mainly processes the reaction rate field related to effective amplification.

[0033] It is understandable that low-abundance nucleic acid targets are prone to signal weakening and background drift during the early stages of amplification. If the complete reaction rate field table is directly input into the shared coding layer, the background fields before the amplification start position will participate in the amplification state coding. In this embodiment, the reaction rate field table is truncated by the gated position column, so that the model input is concentrated in the field range after the amplification start position, thereby reducing the interference of the background fields on the amplification state coding.

[0034] For example, continuing with the data from sample A, the amplification start position in the reaction rate field table is cycle 2, and the gating position column is recorded as {(A,2)}. After reading the reaction rate field table according to the gating position column, the gated reaction rate table retains the fragment fluorescence intensities of 12, 17, 29, and 47, fragment fluorescence growth ratios of 1.42, 1.71, and 1.62, and fragment growth ratio differences of 0.29 and -0.09, corresponding to cycles 2 to 5.

[0035] Furthermore, after the gated reaction rate table is input into the shared coding layer of the multi-objective regression model, the initial amplification state code is obtained. The initial amplification state code includes the amplification initiation segment code and the amplification acceleration segment code. Subsequently, the initial amplification state code is corrected according to the cycle numbering order to obtain the amplification state code. This correction ensures that the amplification state code maintains the cycle sequence and avoids the appearance of field arrangements in the coding results that do not match the amplification reaction time order.

[0036] like Figure 4 As shown, the gated reaction rate table is obtained by extracting the reaction rate field table based on the gating position column, including: Based on the sample number in the reaction rate field table matched with the gating position column, a sample gating mapping table is obtained, which records the correspondence between the sample number and the amplification start position; Based on the sample gating mapping table, locate the starting row of the field in the reaction rate field table to obtain the starting row table of the fields; The reaction rate field table is read from the field start row table to obtain the gated reaction rate table.

[0037] In this embodiment, the sample gating mapping table establishes a correspondence between the gating position column and the sample number in the reaction rate field table. The field start row table locates the field reading starting point of each sample in the reaction rate field table according to the sample gating mapping table. The gating reaction rate table is obtained by reading the reaction rate field table from the field start row table, thereby ensuring that different samples enter the model for processing according to their respective amplification starting positions.

[0038] For example, if the gating position column is {(A,2),(B,3)}, it means that the amplification start position of sample A is the 2nd cycle and the amplification start position of sample B is the 3rd cycle. The sample gating mapping table records that sample number A corresponds to the 2nd cycle and sample number B corresponds to the 3rd cycle accordingly. The field start row table further records that sample A starts reading from the field corresponding to the 2nd cycle and sample B starts reading from the field corresponding to the 3rd cycle.

[0039] It is understandable that the amplification start position of different samples to be tested may be different. If the reaction rate field table is truncated according to a fixed cycle number, more background fields will be mixed into the low abundance samples that start amplification later. In this embodiment, the field reading start point of each sample is located by the sample gating mapping table and the field start row table respectively, so that the gating reaction rate table retains the field content corresponding to the amplification start state of each sample.

[0040] Furthermore, the gated reaction rate table serves as the input to the shared coding layer, bridging the data conversion relationship between the reaction rate field table and the amplification status coding. This process enables subsequent amplification efficiency estimates and background drift estimates to be generated based on the amplification start position of each sample, avoiding additional coding bias introduced between different samples due to different amplification start times.

[0041] S30: Adjust the detection threshold according to the correction coefficient vector to construct an optimized detection parameter set, which is used for the enhancement processing of low-abundance nucleic acid signals; In this embodiment, the correction coefficient vector inherits the aforementioned nonlinear deviation correction factor. The correction coefficient vector organizes the correction coefficients corresponding to each sample according to the sample number and channel number. The detection threshold is the channel judgment benchmark used by the preset nucleic acid detection model to judge the fluorescence signal. The optimized detection parameter set combines the corrected detection threshold and the correction coefficient to ensure that the low abundance nucleic acid signal obtains parameter conditions that match the amplification deviation before entering the preset nucleic acid detection model.

[0042] It is understandable that when the fluorescence intensity of low-abundance nucleic acid signals is close to the background signal, directly using a fixed detection threshold can easily cause the real amplification signal to be covered by background fluctuations. In this embodiment, the detection threshold is adjusted according to the correction coefficient vector. Instead of lowering the judgment criteria of all channels individually, the sample amplification deviation is correlated with the fluorescence channel, so that the corrected detection threshold of different channels can correspond to the deviation state of each sample during the amplification process.

[0043] For example, continuing to use the data of sample A, which corresponds to the FAM channel, the correction coefficient corresponding to sample A in the correction coefficient vector is 0.86, the basic detection threshold of the FAM channel is 35, and the corrected detection threshold can be formed by the basic detection threshold and the correction coefficient. Thus, the corrected detection threshold of sample A in the FAM channel is 30.10, and this corrected detection threshold is entered into the subsequent optimized detection parameter set.

[0044] The detection threshold is adjusted based on the correction coefficient vector to construct an optimized detection parameter set, including: The channel correction mapping table is obtained by matching the channel number of the sample to be detected with the correction coefficient vector; The basic detection threshold of each fluorescence channel is corrected according to the channel calibration mapping table to obtain the channel threshold table, which records the corrected detection threshold corresponding to each fluorescence channel. Construct a parameter candidate table based on the channel threshold table and the correction coefficient vector; The optimized detection parameter set is obtained by selecting the corrected detection threshold corresponding to low-abundance nucleic acid signals from the parameter candidate table.

[0045] In this embodiment, the channel correction mapping table is used to record the correspondence between the correction coefficients in the correction coefficient vector and the channel number of the sample to be detected, the channel threshold table is used to record the corrected detection threshold of each fluorescence channel, the parameter candidate table is used to record the corrected detection threshold and correction coefficient of the sample to be detected in the corresponding fluorescence channel, and the optimized detection parameter set is formed by the parameter records corresponding to the low abundance nucleic acid signal.

[0046] It is understandable that fluorescent channels usually correspond to different nucleic acid targets. The threshold adjustment of the channel where the low abundance nucleic acid signal is located cannot be processed without the channel number. In this embodiment, the channel number of the sample to be detected is first matched with the correction coefficient vector to obtain the channel correction mapping table. Then, the basic detection threshold is corrected based on the channel correction mapping table so that the correction coefficient can be accurately applied to the corresponding fluorescent channel.

[0047] For example, the correction coefficient vector includes a correction coefficient of 0.86 for the FAM channel corresponding to sample A and a correction coefficient of 0.93 for the HEX channel corresponding to sample B. The channel correction mapping table can record (A, FAM, 0.86) and (B, HEX, 0.93). If the basic detection threshold of the FAM channel is 35 and the basic detection threshold of the HEX channel is 33, then the channel threshold table can record a corrected detection threshold of 30.10 for the FAM channel and a corrected detection threshold of 30.69 for the HEX channel.

[0048] Furthermore, the parameter candidate table combines the channel threshold table and the correction coefficient vector into the dimension of the sample to be detected, so that each sample corresponds to a corrected detection threshold and a correction coefficient. Then, the corrected detection threshold corresponding to low abundance nucleic acid signals is selected from the parameter candidate table to obtain the optimized detection parameter set. This process avoids introducing unnecessary parameter adjustments for non-low abundance nucleic acid signals.

[0049] Based on the parameter candidate table, the corrected detection threshold corresponding to low-abundance nucleic acid signals is selected to obtain the optimized detection parameter set, including: Based on the correction coefficients and corrected detection thresholds in the parameter candidate table, a low abundance candidate parameter table is obtained, which records the parameter records corresponding to low abundance nucleic acid signals. Based on the low abundance candidate parameter table and the channel threshold table, the low abundance channel parameter table is obtained. The low abundance channel parameter table records the corrected detection threshold of the fluorescence channel where the low abundance nucleic acid signal is located. Based on the low abundance channel parameter table and the channel correction mapping table, an optimized detection parameter set is obtained.

[0050] In this embodiment, the low abundance candidate parameter table is formed by the correction coefficients and corrected detection thresholds in the parameter candidate table, the low abundance channel parameter table is obtained by matching the low abundance candidate parameter table with the channel threshold table, and the optimized detection parameter set is obtained by combining the low abundance channel parameter table and the channel correction mapping table, so that the optimized detection parameter set simultaneously includes the corrected detection threshold and the corresponding correction coefficient of the fluorescence channel where the low abundance nucleic acid signal is located.

[0051] Understandably, the low-abundance candidate parameter table is not directly selected from the original fluorescence signal, but is generated step by step through the correction coefficient vector, channel correction mapping table and channel threshold table. The low-abundance channel parameter table further binds the low-abundance nucleic acid signal to the fluorescence channel it belongs to, ensuring that the optimized detection parameter set will not incorrectly apply the corrected detection threshold of one channel to other channels.

[0052] For example, if the parameter candidate table contains (A,FAM,0.86,30.10) and (B,HEX,0.93,30.69), and the low abundance nucleic acid signal corresponds to the FAM channel of sample A, then the low abundance candidate parameter table retains (A,FAM,0.86,30.10), the low abundance channel parameter table records the corrected detection threshold of 30.10 for the FAM channel, and the optimized detection parameter set records sample A, the FAM channel, the correction coefficient 0.86, and the corrected detection threshold 30.10.

[0053] Compared to the existing technology that determines low-abundance targets solely based on fluorescence signal intensity, this embodiment first maps low-abundance nucleic acid signals to a parameter candidate table, and then combines a channel threshold table and a channel correction mapping table to generate an optimized detection parameter set. This allows the detection parameters to inherit the aforementioned amplification deviation correction results and maintain the correspondence between the fluorescence channel and the sample to be detected.

[0054] S40: Input the optimized detection parameter set into the preset nucleic acid detection model and output a set of nucleic acid abundance results, which includes the quantized signal and judgment value after nonlinear deviation correction for each sample.

[0055] In this embodiment, the optimized detection parameter set follows the aforementioned correction coefficient vector and corrected detection threshold. The preset nucleic acid detection model refers to the detection model with configured sample input fields and fluorescence channel fields. The parameterized nucleic acid detection model is the model obtained by configuring the channel detection parameters of the preset nucleic acid detection model according to the optimized detection parameter set. The nucleic acid abundance result set records the quantification signal and judgment value of each sample under the corresponding fluorescence channel.

[0056] It is understandable that the original fluorescence signal is not used directly to output the detection result. Instead, the optimized detection parameter set is first input into the preset nucleic acid detection model so that the channel detection parameters of the preset nucleic acid detection model are matched with the correction state of the low abundance nucleic acid signal. Then, the fluorescence signal of the sample to be detected is processed so that the aforementioned amplification kinetics feature set, nonlinear deviation correction factor and correction detection threshold can be transmitted to the detection result output stage.

[0057] For example, continuing to use the data of sample A, the detection parameter set is optimized to record sample A, FAM channel, correction coefficient 0.86 and correction detection threshold 30.10. After the preset nucleic acid detection model configures the detection parameters of FAM channel according to the optimized detection parameter set, the fluorescence signal of sample A is input into the parameterized nucleic acid detection model to obtain the quantized signal and judgment value corresponding to FAM channel.

[0058] The optimized detection parameter set is input into the preset nucleic acid detection model, and the output is a set of nucleic acid abundance results, including: The parameterized nucleic acid detection model is obtained by configuring the channel detection parameters of the preset nucleic acid detection model based on the optimized detection parameter set; The fluorescence signal of the sample to be tested is input into the parameterized nucleic acid detection model to obtain the channel quantification signal table; A sample decision table is generated based on the channel quantization signal table and the optimized detection parameter set. Based on the channel quantization signal table and the sample decision table, a set of nucleic acid abundance results is obtained.

[0059] In this embodiment, the channel detection parameters of the preset nucleic acid detection model are configured according to the optimized detection parameter set to obtain the parameterized nucleic acid detection model. The channel detection parameters include the corrected detection threshold and the correction coefficient. The parameterized nucleic acid detection model retains the input structure of the preset nucleic acid detection model, and at the same time writes the corrected detection threshold and correction coefficient corresponding to different fluorescence channels into the corresponding channels, so that the subsequent fluorescence signals can be quantized according to the channel parameters.

[0060] In this embodiment, the fluorescence signal of the sample to be tested is input into the parameterized nucleic acid detection model to obtain a channel quantization signal table. The channel quantization signal table is used to record the quantization signal of the sample to be tested in each fluorescence channel. Continuing to use the data of sample A, the parameterized nucleic acid detection model has been configured with a FAM channel correction detection threshold of 30.10 and a correction coefficient of 0.86. The channel quantization signal table after inputting sample A can record sample A, FAM channel, and quantization signal of 32.40.

[0061] Furthermore, a sample determination table is generated based on the channel quantization signal table and the optimized detection parameter set. A nucleic acid abundance result set is then obtained based on the channel quantization signal table and the sample determination table. The sample determination table records the determination relationship between the channel quantization signal and the corrected detection threshold. The nucleic acid abundance result set organizes the quantization signal and determination value into the same sample record. Continuing with the data from sample A, after the channel quantization signal 32.40 corresponds to the corrected detection threshold 30.10, the sample determination table can record sample A, FAM channel, determination value positive, and the nucleic acid abundance result set can record sample A, FAM channel, quantization signal 32.40, determination value positive.

[0062] Compared to the existing technology that only performs threshold judgment at the amplification curve, this embodiment optimizes the detection parameter set and configures it into the preset nucleic acid detection model. Then, the parameterized nucleic acid detection model generates a channel quantification signal table and a sample judgment table, so that the final output set of nucleic acid abundance results maintains a data transfer relationship with the set of amplification kinetic characteristics and the nonlinear deviation correction factor, which is beneficial to improving the comparability of detection outputs of different samples.

[0063] Example 2 Please see Figure 2As shown, based on the same inventive concept, this embodiment discloses a nucleic acid optimization detection system based on nonlinear amplification modeling. For details not covered in this embodiment, please refer to the relevant sections of Embodiment 1. The system includes: Feature construction module: used to obtain the initial fluorescence time series of the sample to be tested, generate an initial signal matrix based on the initial fluorescence time series, and construct an amplification kinetic feature set based on the initial signal matrix; Correction generation module: used to input the amplification kinetics feature set into the multi-objective regression model, obtain the nonlinear bias correction factor for each sample, and generate a correction coefficient vector based on the nonlinear bias correction factor; Parameter optimization module: used to adjust the detection threshold according to the correction coefficient vector, thereby constructing an optimized detection parameter set, which is used for the enhancement processing of low-abundance nucleic acid signals; The results output module is used to input the optimized detection parameter set into the preset nucleic acid detection model and output a set of nucleic acid abundance results, which includes the quantized signal and judgment value after nonlinear deviation correction for each sample.

[0064] The detailed description above, in conjunction with the accompanying drawings, describes examples but does not represent all examples that can be implemented or fall within the scope of the claims. The terms “example” and “exemplary” are used in this specification to mean “serving as an example, instance or illustration” and do not mean “superior to or better than other examples”.

[0065] Throughout this specification, the phrase "an embodiment" or "an embodiment" means that a particular feature, structure, or characteristic described in connection with that embodiment is included in at least one embodiment of the invention. Therefore, the use of these phrases may refer to more than one embodiment. Furthermore, the described features, structures, or characteristics may be combined in any suitable manner in one or more embodiments.

[0066] It should also be noted that these embodiments may be described as processes depicted as flowcharts, structural diagrams, or block diagrams. Although a flowchart may describe the operations as sequential processes, many of these operations can be performed in parallel or concurrently, and the order of these operations may be rearranged.

Claims

1. A method for optimized detection of nucleic acids based on modeling of non-linear amplification, characterized in that, include: Obtain the initial fluorescence time series of the sample to be tested, generate an initial signal matrix based on the initial fluorescence time series, and construct an amplification kinetics feature set based on the initial signal matrix; The amplification kinetics feature set is input into a multi-objective regression model to obtain the nonlinear bias correction factor for each sample, and a correction coefficient vector is generated based on the nonlinear bias correction factor. The detection threshold is adjusted based on the correction coefficient vector to construct an optimized detection parameter set, which is used for the enhancement processing of low-abundance nucleic acid signals. The optimized set of detection parameters is input into the preset nucleic acid detection model, and the set of nucleic acid abundance results is output. The set of nucleic acid abundance results includes the quantized signal and judgment value after nonlinear deviation correction for each sample.

2. The method for detecting nucleic acid optimization based on modeling of non-linear amplification according to claim 1, wherein, A set of amplification kinetic features is constructed based on the initial signal matrix, including: The initial signal matrix is ​​rearranged according to the sample number and cycle number to obtain the cyclic fluorescence table, which records the fluorescence intensity of the same sample in consecutive cycles. Based on the fluorescence intensity relationship between adjacent cycles in the cyclic fluorescence table, a fluorescence growth ratio table is obtained, in which the fluorescence growth ratio table records the fluorescence intensity ratio between adjacent cycles; Based on the growth change relationship between adjacent cycles in the fluorescence growth ratio table, a growth ratio difference table is obtained, in which the growth ratio difference table records the difference between adjacent fluorescence growth ratios; The amplification kinetics fragment is obtained by extracting the cyclic fluorescence table based on the sign transition position in the growth ratio difference table. Based on the amplification kinetics fragment, fluorescence growth ratio table, and growth ratio difference table, the set of amplification kinetics characteristics is obtained.

3. The method of claim 2, wherein the modeling of the non-linear amplification is based on a model of the non-linear amplification. Based on the sign transition positions in the growth ratio difference table, the cyclic fluorescence table is extracted to obtain the amplification kinetics fragment, including: Based on the sign change relationship between adjacent items in the growth ratio difference table, a candidate inflection point table is obtained. Based on the candidate inflection point table, the fluorescence growth ratio table is retrieved to obtain candidate growth ratio fragments. Based on the continuous rising and falling segments in the candidate growth ratio fragments, the amplification start position and amplification plateau position are obtained. The amplification start position corresponds to the starting cycle number of the continuous rising segment, and the amplification plateau position corresponds to the ending cycle number of the continuous falling segment. The amplification kinetics fragment is obtained by extracting cyclic fluorescence tables based on the amplification initiation and plateau positions.

4. The nucleic acid optimization detection method based on nonlinear amplification modeling according to claim 3, characterized in that, The amplification kinetics feature set is input into a multi-objective regression model to obtain the nonlinear bias correction factor for each sample, including: Based on the amplification kinetic fragments in the amplification kinetics feature set, a fragment feature table is obtained. Based on the fragment feature table, fluorescence growth ratio table, growth ratio difference table, and amplification initiation position, a reaction rate field table is obtained. Input the reaction rate field table into the shared coding layer of the multi-objective regression model to obtain the amplified state code; Amplification efficiency estimates and background drift estimates are generated based on the amplification state coding, where the amplification efficiency estimates and background drift estimates are derived from different output branches of the multi-objective regression model. Based on the amplification efficiency estimate, background drift estimate, and reaction rate field table, the nonlinear bias correction factor for each sample is obtained.

5. The method of claim 4, wherein the modeling of the non-linear amplification is based on a model of the non-linear amplification. 5 Inputting the reaction rate field table into the shared coding layer of the multi-objective regression model yields the amplified state coding, including: Based on the amplification start position in the reaction rate field table, the gating position column is obtained. Based on the gating position column, the reaction rate field table is extracted to obtain the gated reaction rate table. By inputting the gated reaction rate table into the shared coding layer of the multi-objective regression model, the initial amplification state coding is obtained, which includes the amplification initiation segment coding and the amplification acceleration segment coding. The initial amplification status code is corrected according to the cyclic numbering order to obtain the amplification status code.

6. The nucleic acid optimization detection method based on nonlinear amplification modeling according to claim 5, characterized in that, Based on the gating position column, the reaction rate field table is extracted to obtain the gating reaction rate table, which includes: Based on the sample number in the reaction rate field table matched with the gating position column, a sample gating mapping table is obtained, which records the correspondence between the sample number and the amplification start position; Based on the sample gating mapping table, locate the starting row of the field in the reaction rate field table to obtain the starting row table of the fields; The reaction rate field table is read from the field start row table to obtain the gated reaction rate table.

7. The nucleic acid optimization detection method based on nonlinear amplification modeling according to claim 1, characterized in that, The detection threshold is adjusted based on the correction coefficient vector to construct an optimized detection parameter set, including: The channel correction mapping table is obtained by matching the channel number of the sample to be detected with the correction coefficient vector; The basic detection threshold of each fluorescence channel is corrected according to the channel calibration mapping table to obtain the channel threshold table, which records the corrected detection threshold corresponding to each fluorescence channel. Construct a parameter candidate table based on the channel threshold table and the correction coefficient vector; The optimized detection parameter set is obtained by selecting the corrected detection threshold corresponding to low-abundance nucleic acid signals from the parameter candidate table.

8. The nucleic acid optimization detection method based on nonlinear amplification modeling according to claim 7, characterized in that, Based on the parameter candidate table, the corrected detection threshold corresponding to low-abundance nucleic acid signals is selected to obtain the optimized detection parameter set, including: Based on the correction coefficients and corrected detection thresholds in the parameter candidate table, a low abundance candidate parameter table is obtained, which records the parameter records corresponding to low abundance nucleic acid signals. Based on the low abundance candidate parameter table and the channel threshold table, the low abundance channel parameter table is obtained. The low abundance channel parameter table records the corrected detection threshold of the fluorescence channel where the low abundance nucleic acid signal is located. Based on the low abundance channel parameter table and the channel correction mapping table, an optimized detection parameter set is obtained.

9. The nucleic acid optimization detection method based on nonlinear amplification modeling according to claim 1, characterized in that, The optimized detection parameter set is input into the preset nucleic acid detection model, and the output is a set of nucleic acid abundance results, including: The parameterized nucleic acid detection model is obtained by configuring the channel detection parameters of the preset nucleic acid detection model based on the optimized detection parameter set; The fluorescence signal of the sample to be tested is input into the parameterized nucleic acid detection model to obtain the channel quantification signal table; A sample decision table is generated based on the channel quantization signal table and the optimized detection parameter set. Based on the channel quantization signal table and the sample decision table, a set of nucleic acid abundance results is obtained.

10. A nucleic acid optimization detection system based on nonlinear amplification modeling, used to implement the nucleic acid optimization detection method based on nonlinear amplification modeling as described in any one of claims 1-9, characterized in that, include: Feature construction module: used to obtain the initial fluorescence time series of the sample to be tested, generate an initial signal matrix based on the initial fluorescence time series, and construct an amplification kinetic feature set based on the initial signal matrix; Correction generation module: used to input the amplification kinetics feature set into the multi-objective regression model, obtain the nonlinear bias correction factor for each sample, and generate a correction coefficient vector based on the nonlinear bias correction factor; Parameter optimization module: used to adjust the detection threshold according to the correction coefficient vector, thereby constructing an optimized detection parameter set, which is used for the enhancement processing of low-abundance nucleic acid signals; The results output module is used to input the optimized detection parameter set into the preset nucleic acid detection model and output a set of nucleic acid abundance results, which includes the quantized signal and judgment value after nonlinear deviation correction for each sample.

Images