A method of correcting errors in sequence information generated by nucleic acid sequencing

Abstract

The present invention discloses a method for correcting errors in sequence information generated by nucleic acid sequencing. The method corrects the sequencing signal of a nucleic acid to be tested without adding nucleic acid molecules with known sequences. The method finds a set of parameters such that the corrected signal under the parameters is as close to an integer value as possible.

Description

Technical Field

[0001] This invention relates to a method for correcting sequence information errors generated by nucleic acid sequencing, belonging to the field of gene sequencing. Background Technology

[0002] In high-throughput sequencing, due to incomplete sequencing reactions, side reactions caused by impurities, base mismatches, and other reasons, the length of the newly formed nucleic acid strands gradually becomes inconsistent as the sequencing reaction proceeds, leading to deviations in sequencing signal intensity. This phenomenon is called phase loss. Phase loss is a major factor limiting read length and accuracy in high-throughput sequencing. In practice, to reduce the impact of phase loss, on the one hand, sequencing reagents and reaction conditions can be optimized to reduce the degree of phase loss; on the other hand, signal processing can be performed on the raw sequencing signal to correct the signal disturbances caused by phase loss.

[0003] US Patent 8364417B2 discloses a method for correcting signal distortion caused by phase loss, which is particularly suitable for sequencing technologies based on single-nucleotide addition (SNA). Simply put, this method attempts to correct the raw sequencing signal using a set of parameters. In supervised mode, these parameters are estimated from the sequencing signals of a set of nucleic acid molecules with known sequences; in unsupervised mode, the method attempts to find a set of parameters such that the minimum value of the corrected signal is exactly greater than 0.

[0004] Patent CN107958138B discloses another method for correcting signal distortion caused by phase loss. This method is applicable not only to sequencing technologies based on SNA but also to sequencing technologies based on error-correction code (ECC) sequencing. This method is supervised, requiring the estimation of the degree of phase loss using the sequencing signals of a set of nucleic acid molecules with known sequences, and then correcting the sequencing signal of the DNA to be tested. However, this method requires adding nucleic acid molecules with known sequences to the sequencing process, which is cumbersome. Furthermore, the nucleic acid molecules with known sequences are usually of low complexity, and their estimation results often have strong bias. Therefore, there is a need to develop an unsupervised correction method to correct sequence information errors generated by gene sequencing. Summary of the Invention

[0005] This patent discloses a method for correcting sequence information errors generated during nucleic acid sequencing. This method corrects the sequencing signal of the nucleic acid to be tested without adding additional nucleic acid molecules with known sequences. Simply put, this method attempts to find a set of parameters that, under these parameters, yield a corrected signal as close as possible to integer values.

[0006] A method for correcting sequence information errors generated by nucleic acid sequencing, characterized in that it includes:

[0007] A. Sequencing the nucleic acid to be tested to obtain the raw signal corresponding to the nucleic acid sequence;

[0008] B. Construct a phase mismatch using a set of parameters, and use the phase mismatch to calculate the correction signal of the nucleic acid to be tested;

[0009] C. Optimize the parameters so that the correction signal is as close to an integer as possible;

[0010] D. Output the nucleic acid sequence corresponding to the correction signal;

[0011] The parameters include the lead coefficient and the lag coefficient;

[0012] The phase mismatch refers to the change in sequencing results due to lead and / or lag.

[0013] According to a preferred embodiment, the set of parameters also includes unit signal, offset, attenuation coefficient, etc.

[0014] According to a preferred embodiment, in the sequencing process, the nucleotide substrate molecules added in each sequencing reaction can be one, two, three, or four.

[0015] According to a preferred embodiment, the sequencing refers to a sequencing process with the 3' end open; the types of nucleotides added to the sequencing reaction can be one, two, or three.

[0016] According to a preferred embodiment, the sequencing includes two different sequencing reagents: a first sequencing reagent and a second sequencing reagent; the two sequencing reagents are added cyclically; wherein the first sequencing reagent contains at least two different nucleotide monomers having a detectable label; the second sequencing reagent contains one or more nucleotide monomers having a detectable label, and the nucleotide monomers are different from those present in the first sequencing reagent, and wherein the second sequencing reagent is provided after the first sequencing reagent is provided, and the signal generated by the detectable label is detected after the nucleotide monomers are incorporated into the nucleic acid to be tested.

[0017] According to a preferred embodiment, the signal detected in the sequencing can be an electrical signal, a biofluorescence signal, a chemifluorescence signal, or a combination thereof.

[0018] According to a preferred embodiment, let the correction signal c = (c1, c2, ..., c n The phrase "making the correction signal as close to an integer as possible" refers to making a certain objective function of the correction signal reach its maximum or minimum value. This objective function and its maximum or minimum value include, but are not limited to:

[0019] 1) Make minimize;

[0020] 2) Make minimize;

[0021] 3) Make minimize

[0022] 4) Make max|c i -[c i ]|minimize;

[0023] 5) Make minimize;

[0024] 6) Make minimize;

[0025] 7) Make minimize;

[0026] Among them, [c i ] is c i The rounding here refers to rounding to the nearest integer, either rounding up or rounding down. In some implementations, when c... i When < 0, we can let [c i Take other meaningful values, such as 1. Also, satisfy other validity checks, such as the strength being less than 0 or greater than the maximum reasonable value for measurement.

[0027] According to a preferred embodiment, a portion (not all) of the correction signal is made as close to an integer as possible. This portion can be the beginning, end, middle, odd, or even cycles of the correction signal, or it can be cycles in the correction signal that are greater than or less than a given threshold. The given threshold can be a positive number such as 1, 1.5, 2, or 3.

[0028] According to the preferred implementation, a group of multiple nucleic acid sequencing signals can be corrected together, i.e., group correction.

[0029] According to a preferred embodiment, in the sequencing, the nucleic acid sequence to be tested is linked with bases of a known sequence. During sequencing, the unit signal can be obtained through the signal of the bases of the known sequence.

[0030] According to a preferred embodiment, the unit signal at each sampling point is different.

[0031] According to a preferred embodiment, signal normalization and phase correction can be two sequential steps, for example:

[0032] 1. Using a set of normalization parameters, such as unit signal and offset, normalize the original signal f into a normalized signal s; the normalization parameters in this step will not change in subsequent steps.

[0033] 2. Using a set of phase decoupling parameters, such as leading and lagging, correct the normalized signal s into a corrected signal c; optimize the phase decoupling parameters in this step so that the corrected signal c is as close to an integer as possible.

[0034] According to a preferred embodiment, the aforementioned "making the correction signal sufficiently close to an integer" means that the maximum or minimum value of a certain objective function of the correction signal is less than or greater than a given threshold, for example, the objective function... The minimum value should be less than the threshold 0.5n; otherwise, the sequencing signal quality of the nucleic acid sequence is considered to be too poor.

[0035] According to preferred embodiments, the methods for optimizing parameters include, but are not limited to, grid search, exhaustive search, gradient descent, stochastic gradient descent, Newton's method, Hessian matrix method, interior point method, sequential quadratic programming, active set method, trust region method, Levenberg-Marquardt method, heuristic search, etc., where heuristic search includes, but is not limited to, genetic algorithms, simulated annealing algorithms, ant colony optimization, harmonic optimization, spark optimization, particle swarm optimization, immune optimization, gene expression algorithms, etc. The optimization methods mentioned here are all conventional mathematical concepts.

[0036] According to the preferred embodiment, (since c, s, and f can be reversibly calculated from each other) signal normalization and phase correction can be performed iteratively, for example:

[0037] 1. Using the empirical value p0 of the normalization parameter, normalize the original signal f into a normalized signal s0;

[0038] 2. Using the empirical value q0 of the phase loss parameter, the normalized signal s0 is corrected into the corrected signal c0;

[0039] 3. Attempt to find a set of normalization parameters p1 such that the theoretical original signal f0 calculated from the correction signal c0 according to parameters p1 and q0 is as close as possible to the measured original signal f;

[0040] 4. Attempt to find a set of phase loss parameters q1 such that the corrected signal c1 calculated from the original signal f according to parameters p1 and q1 is as close as possible to an integer;

[0041] 5. Repeat steps 3-4 to obtain a series of corrected signals (c0, c1, c2, ...) and the theoretical original signals (f0, f1, f2, ...) until the corrected signals converge.

[0042] According to the preferred embodiment, (since c, s, and f can be reversibly calculated from each other) signal normalization and phase correction can be performed iteratively, for example:

[0043] 1. Using the empirical value p0 of the normalization parameter, normalize the original signal f into a normalized signal s0;

[0044] 2. Using the empirical value q0 of the phase loss parameter, the normalized signal s0 is corrected into the corrected signal c0;

[0045] 3. Attempt to find a set of phase loss parameters q1 such that the corrected signal c1 calculated from the original signal f according to parameters p0 and q1 is as close as possible to an integer;

[0046] 4. Attempt to find a set of normalization parameters p1 such that the theoretical original signal f0 calculated by the correction signal c1 according to parameters p1 and q1 is as close as possible to the measured original signal f;

[0047] 5. Repeat steps 3-4 to obtain a series of corrected signals (c0, c1, c2, ...) and the theoretical original signals (f0, f1, f2, ...) until the corrected signals converge.

[0048] According to the preferred embodiment, while ensuring that the correction signal is as close to an integer as possible, the parameters are further optimized so that:

[0049] (A) The correlation between the front and rear wheels of the correction signal c is minimal, for example:

[0050] a) Make (c1,c2,…,c n-1 (c2,c3,…,c) and (c2,c3,…,c) n Minimize the correlation coefficient between them; or,

[0051] b) Make (c1,c2,…,c n-2 (c3,c4,…,c) and (c3,c4,…,c) n Minimize the correlation coefficient between them;

[0052] The aforementioned correlation coefficients include, but are not limited to, Pearson correlation coefficient, Spearman correlation coefficient, and cosine distance.

[0053] The opposite of the x-axis, the opposite of the Euclidean distance, the opposite of the Manhattan distance, etc.;

[0054] (B) The distribution of the correction signal should be as close as possible to the reference distribution, for example as follows:

[0055] a) The reference distribution is To minimize the Kullback-Leibler divergence between the distribution of the corrected signal and the reference distribution; or,

[0056] b) The reference distribution is To minimize the Kullback-Leibler divergence between the distribution of the corrected signal and the reference distribution;

[0057] The distribution of the correction signal can be obtained in a variety of ways, such as interval statistics or kernel density estimation.

[0058] According to the preferred embodiment, if a set of parameters cannot be found that makes the correction signal sufficiently close to an integer, the sequencing signal quality of the nucleic acid sequence is considered to be too poor. In this case, any one of the following choices or a combination thereof can be made:

[0059] 1) Do not output the correction results for this nucleic acid sequence;

[0060] 2) Output a result indicating that the correction failed, such as NaN, None, Null, Nil, null value, etc.

[0061] 3) Output the correction results and report relevant information in the log;

[0062] 4) Output the correction result and mark it with a flag indicating failure.

[0063] According to a preferred embodiment, multiple sequencing reactions are executed in parallel, and step AD is performed for each sequencing reaction.

[0064] The present invention also discloses a gene sequencing system comprising: (a) a sequencer that detects multiple signals generated in response to multiple nucleotide substrates introduced during sequencing; and (b) a computer including executable code stored thereon, which obtains a corrected nucleic acid sequence using a method for correcting sequence information errors generated during nucleic acid sequencing as described in any of the preceding embodiments.

[0065] Beneficial effects

[0066] The method of this invention makes the correction signal as close to an integer as possible. The unsupervised mode of US8364417B2 mentioned in the background art makes the minimum value of the correction signal exactly greater than 0. The method of this invention has the following advantages:

[0067] 1. US8364417B2 is only applicable to single nucleotide addition (SNA) sequencing and has limited effectiveness in error-correcting code (ECC) sequencing. This method is effective for both SNA and ECC sequencing.

[0068] 2. CN107958138B is a supervised method, while this method is unsupervised, does not depend on nucleic acid molecules with known sequences, is simple to operate, and has low bias. Attached Figure Description

[0069] Figure 1The raw sequencing signal of a DNA sequence according to one embodiment of the present invention.

[0070] Figure 2 A heatmap of the objective function distribution according to one embodiment of the present invention.

[0071] Figure 3 A correction signal for a DNA sequence according to one embodiment of the present invention.

[0072] Figure 4 The raw sequencing signal of a DNA sequence according to one embodiment of the present invention.

[0073] Figure 5 A heatmap of the objective function distribution according to one embodiment of the present invention.

[0074] Figure 6 A correction signal for a DNA sequence according to one embodiment of the present invention.

[0075] Figure 7 Signal distribution of λ phage genome sequencing according to one embodiment of the present invention. Top: raw fluorescence signal; Middle: normalized signal; Bottom: correction signal.

[0076] Figure 8 Histogram of parameter distribution of each sequencing signal in a λ phage genome sequencing according to one embodiment of the present invention.

[0077] Figure 9 A histogram showing the changes in parameters of each sequencing signal at the final convergence of phase correction and at the first iteration, according to one embodiment of the present invention.

[0078] Figure 10 A histogram showing the distribution of the objective function values ​​according to one embodiment of the present invention. The left figure shows the result of only one parameter estimation; the right figure shows the result of alternating estimation until the correction signal converges.

[0079] Figure 11 A histogram showing the distribution of sequencing error numbers according to one embodiment of the present invention. The left figure shows the result of parameter estimation only once; the right figure shows the result of alternating estimation until the correction signal converges. Detailed Implementation

[0080] To further clarify the present invention, the following specific embodiments are listed. The specific parameters, steps, etc., involved are conventional knowledge in the art. The specific embodiments and examples are not intended to limit the scope of protection of the present invention. Unless otherwise specified, all terms involved in this invention have their conventional meanings in the art. Unless otherwise specified, all gene sequences involved in this invention serve only as simple illustrations of the methods described in this invention and have no special meaning or limitation; simple substitutions do not affect the implementation effect of the present invention. The sequences, parameters, and specific steps involved in this invention can be considered as conventional examples in the art.

[0081] Unless otherwise defined, all technical terms, symbols, and other technical and scientific terms used herein are intended to have the same meaning as commonly understood by one of ordinary skill in the art to which the claimed subject matter pertains. In some instances, for clarity and / or convenience of reference, terms with their commonly understood meanings are defined herein, and such definitions are incorporated herein, but this should not necessarily be construed as indicating a material difference from the meanings commonly understood in the art. Many of the techniques and procedures described or referenced herein are known to those skilled in the art and are commonly used when employing conventional methods.

[0082] The sequencing reaction referred to in this invention refers to the nucleic acid synthesis reaction catalyzed by nucleic acid polymerase. The nucleotides (or deoxynucleotides) used in the sequencing reaction can be natural or chemically modified. The 3' end of the chemically modified nucleotides (or deoxynucleotides) can be open or closed. Examples of sequencing methods using nucleotides (or deoxynucleotides) with open 3' ends include pyrosequencing, semiconductor sequencing, and fluorescence sequencing; examples of sequencing methods using nucleotides (or deoxynucleotides) with closed 3' ends include reversible cycle termination and dideoxynucleotide termination. "High-throughput sequencing" or "next-generation sequencing" includes sequencing methods that determine many (typically thousands to billions) nucleic acid sequences in an inherently parallel manner, i.e., where the DNA template is prepared for batch processing rather than sequencing one at a time, and preferably for parallel readout of many sequences, or for ultra-high-throughput serial processing that can be parallelized itself. Such methods include, but are not limited to, pyrosequencing (e.g., as commercialized by 454 Life Sciences, Inc., Branford, CT); ligation sequencing (e.g., as in SOLiD); and sequencing by linking (e.g., as in SOLiD). TM Technology, Life Technologies, Inc., Carlsbad, CA (commercialized); sequencing through the use of modified nucleotide synthesis (such as in TruSeq) TM and HiSeq TMThe technology is commercialized by Illumina, Inc., San Diego, CA; in HeliScope TM Commercialized in China by Helicos Biosciences Corporation, Cambridge, MA; and in PacBio RS by Pacific Biosciences of California, Inc., Menlo Park, CA, using ion detection technologies such as Ion Torrent TM Sequencing technologies include: sequencing of DNA nanospheres (Complete Genomics, Inc., Mountain View, CA); sequencing technologies based on nanopores (e.g., developed by Oxford Nanopore Technologies, LTD, Oxford, UK); and highly parallelized sequencing methods, for example.

[0083] On the one hand, this invention relates to methods for reading and / or correcting sequence information from raw or unprocessed signals from second-generation sequencing technologies (e.g., for gene or genome sequencing). On the other hand, this invention considers the many factors that cause deviations between raw signals and actual sequence information during nucleic acid sequencing to achieve comprehensive correction of the detected sequence information, thereby reading accurate DNA sequences from the raw sequencing signals. On the other hand, the methods disclosed herein do not affect the normal process of the sequencing reaction. On the other hand, this invention relates to the processing of both monochrome and multicolor sequencing signals.

[0084] In some respects, this paper employs DNA sequencing methods. In some embodiments, the method includes immobilizing the DNA to be tested on a solid surface, hybridizing it with one or more sequencing primers, and / or sequentially performing sequencing reactions and detecting the signals released by the reactions. Each reaction includes the following steps: adding a reaction solution containing nucleotides, enzymes, and other reagents necessary for the reaction to a reactor (e.g., a chip) to initiate a specific biochemical reaction; detecting the signals released by the reaction; and / or cleaning the reactor. The added nucleotides can be natural deoxynucleotides or nucleotides with chemically modified groups, but in one respect, they should have a hydroxyl group at their 3' end. The number of nucleotide types added in each reaction can be one, two, or three, but not four (referring to ACGT or ACGU). The union of the nucleotide types added in two adjacent reactions includes all four nucleotide types. For example, if A and G are added in the first reaction, C and T will be added in the second reaction. In another example, if ACG is added in the first reaction, T will be added in the second reaction. In some implementations, the added nucleotide can be a nucleotide with a chemically modified group, whose 3' end can be reversibly blocked by the modified group. The number of nucleotide types added in each sequencing reaction can be four (referring to ACGT or ACGU), that is, only one nucleotide can be extended at a time. After the reversibly blocking group is removed, the sequencing substrate is added for the next extension.

[0085] The sequencing reaction for the nucleic acid to be tested can be SNA sequencing, ECC sequencing, or fuzzy sequencing. Specifically, SNA sequencing refers to the sequencing method used by companies such as Ion Torrent and 454. This sequencing method includes four different sequencing reagents: a first sequencing reagent, a second sequencing reagent, a third sequencing reagent, and a fourth sequencing reagent; these four reagents are added cyclically. Each sequencing reagent contains a nucleotide monomer with a detectable label. The second sequencing reagent is provided after the first sequencing reagent; the nucleotide monomer is incorporated into the nucleic acid to be tested, and the signal generated by the detectable label is detected. The third sequencing reagent is provided after the second sequencing reagent; the nucleotide monomer is incorporated into the nucleic acid to be tested, and the signal generated by the detectable label is detected. The fourth sequencing reagent is provided after the third sequencing reagent; the nucleotide monomer is incorporated into the nucleic acid to be tested, and the signal generated by the detectable label is detected. ECC sequencing includes various available sequencing workflows, such as multi-round 2+2 sequencing, multi-round 1+3 sequencing, or multi-round... 3x4 sequencing, etc.; specifically, 2+2 sequencing refers to sequencing that includes two different sequencing reagents: a first sequencing reagent and a second sequencing reagent; the two sequencing reagents are added cyclically; wherein the first sequencing reagent contains two different nucleotide monomers with detectable labels; the second sequencing reagent contains two different nucleotide monomers with detectable labels, and the nucleotide monomers are different from those present in the first sequencing reagent, and wherein the second sequencing reagent is provided after the first sequencing reagent is provided, and the nucleotide monomers are incorporated into the nucleic acid to be tested, and the signal generated by the detectable label is detected, for example, a fluorescent signal; in 2+2 sequencing, the combination of fluorescently labeled nucleotides can be used to obtain a fluorescent signal value associated with the target DNA sequence. Possible combinations are as follows: M / K mode: dA4P and dC4P are presented in odd-numbered rounds, and dG4P and dT4P are presented in even-numbered rounds; or vice versa; R / Y mode: dA4P and dG4P are presented in odd-numbered rounds, and dC4P and dT4P are presented in even-numbered rounds; or vice versa; and W / S mode: dA4P and dT4P are presented in odd-numbered rounds, and dC4P and dG4P are presented in even-numbered rounds; or vice versa. Sequencing the nucleic acid to be tested using one of these modes (e.g., M / K mode) for multiple cycles can be called a round.1+3 sequencing, similar to 2+2 sequencing, refers to sequencing that includes two different sequencing reagents: a first sequencing reagent and a second sequencing reagent; these reagents are added cyclically; the first sequencing reagent contains three different nucleotide monomers with detectable labels; the second sequencing reagent contains one nucleotide monomer with a detectable label, and this nucleotide monomer is different from the nucleotide monomer present in the first sequencing reagent; the second sequencing reagent is provided after the first sequencing reagent, and the nucleotide monomer is incorporated into the nucleic acid to be tested, followed by detection of the signal generated by the detectable label, such as a fluorescence signal. 3x4 sequencing refers to sequencing that includes four different sequencing reagents: a first sequencing reagent, a second sequencing reagent, a third sequencing reagent, a fourth sequencing reagent, a fifth sequencing reagent, a sixth sequencing reagent, a seventh sequencing reagent, a seventh sequencing reagent, a thief ... The sequencing reagents consist of a second, third, and fourth round; these four sequencing reagents are added cyclically; each sequencing reagent contains three different nucleotide monomers with detectable markers; the second sequencing reagent is provided after the first sequencing reagent, and the signal generated by the detectable marker is detected after incorporating the nucleotide monomers into the nucleic acid to be tested; the third sequencing reagent is provided after the second sequencing reagent, and the signal generated by the detectable marker is detected after incorporating the nucleotide monomers into the nucleic acid to be tested; the fourth sequencing reagent is provided after the third sequencing reagent, and the signal generated by the detectable marker is detected after incorporating the nucleotide monomers into the nucleic acid to be tested; for example, the four sequencing reagents are B, D, H, and V. Fuzzy sequencing is a single round of 2+2 sequencing, a single round of 1+3 sequencing, or a single round of 3x4 sequencing.

[0086] Based on the IUPAC nucleic acid symbols, the degenerate bases in this invention are represented by the letters in Table 1. For example, M represents A and / or C bases.

[0087] Table 1. Letters representing degenerate bases

[0088] letter The bases represented M AC K GT R AG Y CT W AT S CG B CGT D AGT H ACT V ACG

[0089] In this invention, a sequencing reaction round, or sequencing cycle, refers to the process of providing a sequencing reagent, incorporating a nucleotide with a detectable label into the nucleic acid to be tested, and then detecting the signal generated by the detectable label.

[0090] In the aforementioned sample introduction process, if more than one type of nucleotide (or deoxynucleotide) substrate molecule is added in a sequencing reaction, these different types of nucleotide molecules can release signals of the same or different types. If two types of nucleotides are added in a reaction, these two types of nucleotides can release signals of the same or different types; if three types of nucleotides are added in a reaction, these three types of nucleotides can release signals of the same or different types; optionally, two types release the same signal, while the other releases a different signal. The type of signal in this article refers to the form of the signal (e.g., electrical signal, bioluminescent signal, chemiluminescent signal, etc.), or the color of the optical signal (e.g., green fluorescence signal, red fluorescence signal, etc.), or a combination thereof. For simplicity, a signal in which all nucleotides in a reaction release the same type of signal is called a monochromatic signal; a signal in which all nucleotides in a reaction release more than one type of signal is called a polychromatic signal. The term "color" here is for simplicity only; the type of signal is not limited to different colors of optical signals (e.g., wavelength).

[0091] In high-throughput DNA sequencing, under ideal conditions, the intensity of the raw signal released in each sequencing reaction is directly proportional to the number of bases incorporated into the nascent DNA strand. However, in reality, this ratio does not always exist due to various reasons. For example, firstly, the intensity of the raw signal generally decreases due to fluid corrosion, DNA template hydrolysis, and / or base mismatches. Secondly, due to incomplete sequencing reactions, side reactions (e.g., unwanted reactions), and / or base mismatches, the length of the nascent DNA strand gradually becomes desynchronized as the sequencing reaction progresses (e.g., the length of the nascent DNA strand is inconsistent due to phase loss). The desynchronized length of the nascent DNA strand then leads to a deviation between the raw signal intensity and the actual target DNA sequence. Thirdly, due to spontaneous hydrolysis of nucleotides and / or background fluorescence from the sequencing chip or substrate, the overall intensity of the raw signal will be high. All these factors make it difficult to directly read the target DNA sequence from the intensity of the raw sequencing signal based on the ideal ratio between the two.

[0092] To more clearly describe the content of this invention, the terms used in this invention do not have any extra-standard meanings, and their meanings are listed again below:

[0093] 1. Raw sequencing signal: The signal directly detected by the sequencer, characterized by unknown unit signal and offset, and extremely large differences between different DNA clusters to be tested.

[0094] 2. Normalized signal: A signal derived from the original sequencing signal through some transformation method, characterized by a unit signal of 1 and an offset of 0.

[0095] It is worth noting that the terms used in this invention are common terms in the field and do not exceed their conventional meaning in the field.

[0096] In multibase sequencing, when DNA is extended by n bases, two factors can cause the original sequencing signal to be less than exactly n:

[0097] 1. Unit signal: The increase in signal detected by the sequencer when DNA extends by one base, which is related to the number of DNA molecules involved in the extension reaction, camera exposure time, excitation light intensity, camera photosensitivity, etc.

[0098] 2. Offset: The reference signal detected by the sequencer when there is no base extension, which is related to factors such as chip material and spontaneous hydrolysis of sequencing reaction substrate.

[0099] In this invention, the normalization of the raw sequencing signal refers to the process of eliminating the influence of unit signal and offset on sequencing. Specifically, the normalization of the raw sequencing signal refers to transforming the raw sequencing signal into a normalized signal through some means, wherein the unit signal of the normalized signal is 1 and the offset is 0. The normalization of the raw sequencing signal is a necessary step in the sequencer's base readout process. It can be performed independently after the sequencer acquires the raw sequencing signal and before base readout, or it can be integrated into a base readout method. Here, base readout refers to the process of converting a sequencing signal (raw fluorescence signal or normalized signal) into a DNA sequence, which may or may not include a phase correction step.

[0100] 1. Phase Delay: High-throughput sequencing aims to obtain the sequence information of a target nucleic acid by performing a series of enzymatic reactions and detecting the signals released during these reactions. Ideally, if some nascent DNA strands have extended to the nth base, and the nucleotides added to the current enzymatic reaction precisely pair with and are complementary to the (n+1)th and (n+m)th bases of the target DNA template, the nascent DNA strands in the enzymatic reaction will extend to the (n+m)th base. If the nascent DNA strands in the enzymatic reaction have actually extended beyond the (n+m)th base, then a "lead" has occurred in the nascent DNA strands of that enzymatic reaction. If the nascent DNA strands in the enzymatic reaction have not actually extended to the (n+m)th base, then a "lag" has occurred in the nascent DNA strands of that enzymatic reaction. "Lead" and "lag" are collectively referred to as phase delay. It should be noted that when the nascent DNA strand extends to the nth base, multiple "leads" and "lags" may occur in any possible order. Phase delay is a major factor limiting read length and accuracy in high-throughput sequencing.

[0101] 2. Phase correction: In algorithms, the process of fitting and recorrecting phase-delayed sequencing signals according to a sequencing response model is called "phase correction".

[0102] In some implementations, signal normalization and phase correction can be two sequential steps, for example:

[0103] 1. Using a set of normalization parameters, such as unit signal and offset, normalize the original signal f into a normalized signal s; the normalization parameters in this step will not change in subsequent steps.

[0104] 2. Using a set of phase decoupling parameters, such as lead coefficient and lag coefficient, correct the normalized signal s to the corrected signal c; optimize the phase decoupling parameters in this step so that the corrected signal c is as close to an integer as possible.

[0105] signal correction

[0106] The method of the present invention includes constructing a phase mismatch using a set of parameters, and using the phase mismatch to calculate the correction signal of the nucleic acid to be tested; then optimizing the above parameters so that the correction signal is as close as possible to an integer.

[0107] Parameter estimation involves a set of parameters that describe the relevant properties of the sequencing response, such as the lead factor and lag factor. In some implementations, the set of parameters also includes unit signal, attenuation factor, offset, etc.

[0108] In some implementations, the parameters described above can be determined by enumerating a set of empirical values. For example, the lead coefficient can be selected from 0, 0.1%, 0.2%, 0.3%, 0.4%, 0.5%, ..., 1.0%, and the lag coefficient can be selected from 0, 0.1%, 0.2%, 0.3%, 0.4%, 0.5%, ..., 1.0%. All combinations of the above lead and lag coefficients are enumerated as a set of parameters for constructing the phase mismatch.

[0109] In preferred embodiments, the methods for optimizing parameters include, but are not limited to: grid search, exhaustive search, gradient descent, stochastic gradient descent, Newton's method, Hessian matrix method, interior point method, sequential quadratic programming, active set method, trust region method, Levenberg-Marquardt method, heuristic search, etc., where heuristic search includes, but is not limited to, genetic algorithms, simulated annealing algorithms, ant colony optimization, harmonic optimization, spark optimization, particle swarm optimization, immune optimization, gene expression algorithms, etc. All optimization methods mentioned here are conventional mathematical concepts.

[0110] The process of inferring the sequence information of the DNA to be tested based on the obtained parameters and the actual raw sequencing signal of the DNA with an unknown sequence is called signal correction. The basic process of signal correction can be roughly regarded as the inverse process of parameter estimation.

[0111] First, based on the parameters obtained from the aforementioned process, the transformation function is used. The inverse function transforms the actual raw sequencing signal f into a dephase signal s. Treating s as a zero-order dephase signal s0, a transformation matrix T1 is constructed based on s0 and relevant parameters. The generalized inverse matrix of T1 is then used to transform s0 into a first-order dephase signal s1. Next, a transformation matrix T2 is constructed based on s1 and relevant parameters, and the generalized inverse matrix of T2 is used to transform s1 into a second-order dephase signal s2. This process continues, calculating a series of dephase signals s0, s1, s2, ... . If two adjacent dephase signals s0 are found during the calculation... i and s i+1 If they are equal, stop the calculation and return s. i As a result of signal correction.

[0112] The generalized inverse matrix mentioned above can also be replaced by Tikhonov regularization.

[0113] If the sequencing data is a monochromatic signal, calculate directly using the method described above. If the sequencing data is a multicolor signal, separate each type of signal and calculate the data separately using the method described above.

[0114] The above utilizes transformation functions The process of transforming f into s using the inverse function of T, and the process of transforming si into si+1 using the generalized inverse matrix of T, will be detailed below.

[0115] Construction method of transformation matrix T

[0116] On one hand, the construction of the transformation matrix T depends on a sequencing-related signal X and phase-loss parameters. In parameter estimation, signal a is a quasi-ideal signal h (e.g., a normalized signal); in signal correction, signal x is the phase-loss signal s of various orders. i To improve correction accuracy, signal x can be lengthened by adding several 1s after it; in preferred embodiments, typically 1-100 1s are added. In specific embodiments, 5-10 1s are added. On one hand, the phase decoupling parameters include the lead coefficient ε and the lag coefficient λ.

[0117] On one hand, the construction of the transformation matrix T also includes the construction of the secondary matrix D. On the other hand, assuming the signal x has m values ​​and the sequencing reaction is actually performed n times, then both the transformation matrix T and the auxiliary matrix D have n rows and m columns. For example, in the first row of the auxiliary matrix D, only the first column has an element of 1, and all other elements are 0.

[0118] On one hand, the method includes using the k-th row of the auxiliary matrix D to calculate the k-th row of the transformation matrix T. For the first element of the k-th row of the transformation matrix T:

[0119] 1. If k is odd, then the lag phenomenon should be considered, and the element should be designated as (1-λ)D. 1i ;

[0120] 2. If k is even, then the element is set to 0.

[0121] For the i-th element in the k-th row of the transformation matrix T (excluding the 1st element):

[0122] 1. If k and i have the same parity, then lag should be considered, and the element should be designated as (1-λ)D. ki ;

[0123] 2. If k and i have different parity, then the primary lead phenomenon should be considered, and the element should be designated as ε(1-λ)D. k,i-1 ;

[0124] 3. If the (i-1)th element of signal x is less than 2, then the secondary lead phenomenon should be considered. Based on the calculation results of steps 1 and 2 above, this element should be supplemented by the (i-1)th element T in the same row of the transformation matrix T. k,i-1 .

[0125] On the one hand, this paper discloses a method for correcting sequence information errors generated by nucleic acid sequencing, characterized by comprising:

[0126] A. Sequencing the nucleic acid to be tested to obtain the raw signal corresponding to the nucleic acid sequence;

[0127] B. Construct a phase mismatch using a set of parameters, and use the phase mismatch to calculate the correction signal of the nucleic acid to be tested;

[0128] C. Optimize the parameters so that the correction signal is as close to an integer as possible;

[0129] D. Output the nucleic acid sequence corresponding to the correction signal;

[0130] The parameters include the lead factor and the lag factor; the phase mismatch refers to the change in sequencing results caused by the lead and / or lag.

[0131] According to a preferred embodiment, a set of parameters also includes unit signal, offset, attenuation coefficient, etc.

[0132] According to a preferred embodiment, the optimized parameters are output simultaneously with the correction signal; that is, the unsupervised phase correction includes parameter estimation. In this embodiment, each set of parameters corresponds to a correction signal. The correction signal closest to an integer is found, and the parameters corresponding to that correction signal are also output. These parameters can reflect the sequencing quality and indicate whether the sequencing was abnormal.

[0133] According to a preferred embodiment, in the sequencing process, the nucleotide substrate molecules added in each sequencing reaction can be one, two, three, or four.

[0134] According to a preferred embodiment, the sequencing refers to a sequencing process with the 3' end open; the types of nucleotides added to the sequencing reaction can be one, two, or three.

[0135] According to a preferred embodiment, the signal detected in the sequencing can be an electrical signal, a biofluorescence signal, a chemifluorescence signal, or a combination thereof.

[0136] According to a preferred embodiment, the sequencing includes two different sequencing reagents: a first sequencing reagent and a second sequencing reagent; the two sequencing reagents are added cyclically; wherein the first sequencing reagent contains at least two different nucleotide monomers having a detectable label; the second sequencing reagent contains one or more nucleotide monomers having a detectable label, and the nucleotide monomers are different from those present in the first sequencing reagent, and wherein the second sequencing reagent is provided after the first sequencing reagent is provided, and the signal generated by the detectable label is detected after the nucleotide monomers are incorporated into the nucleic acid to be tested.

[0137] According to a preferred embodiment, let the correction signal c = (c1, c2, ..., c n "Making the correction signal as close to an integer as possible" means making a certain objective function of the correction signal reach its maximum or minimum value. This objective function and its maximum or minimum value include, but are not limited to:

[0138] 1) Make minimize;

[0139] 2) Make minimize;

[0140] 3) Make minimize

[0141] 4) Make max|c i -[c i ]|minimize;

[0142] 5) Make minimize;

[0143] 6) Make minimize;

[0144] 7) Make minimize;

[0145] The above formulas only need to satisfy one of them. Where, [c i ] is c i The rounding here refers to rounding to the nearest integer, either rounding up or rounding down. In some implementations, when c... i When < 0, we can let [c i Take other meaningful values, such as 1. Also, satisfy other validity checks, such as the strength being less than 0 or greater than the maximum reasonable value for measurement.

[0146] According to a preferred embodiment, a portion (not all) of the correction signal is made as close to an integer as possible. This portion can be the beginning, end, middle, odd, or even cycles of the correction signal, or it can be cycles in the correction signal that are greater than or less than a given threshold. The given threshold can be a positive number such as 1, 1.5, 2, or 3.

[0147] According to the preferred implementation, a group of multiple nucleic acid sequencing signals can be corrected together, i.e., group correction.

[0148] According to the preferred embodiment, signal normalization may be a necessary step. Specific implementation methods for signal normalization can be found in patents CN107958138B and CN202010089595.8.

[0149] According to a preferred embodiment, in the sequencing process, the nucleic acid sequence to be tested is linked with bases of a known sequence. During sequencing, a unit signal can be obtained through the signal of these known sequence bases. In this invention, there are multiple ways to obtain the unit signal information. The nucleic acid to be tested is linked with bases of a known sequence, and during sequencing, a unit signal can be obtained through the signal of these known sequence bases. When performing high-throughput sequencing, the unit signal at each sampling point or reaction chamber can be different, which does not affect the sequencing reaction itself. For specific methods of obtaining the unit signal, please refer to patent CN202010089595.8, the content of which can be incorporated herein by reference.

[0150] According to the preferred embodiment, the unit signal of each sampling point is different, which does not affect the sequencing reaction itself. A sampling point, or reaction chamber, is a tiny area on the surface of a gene sequencing chip and is the basic unit where the sequencing reaction occurs. For example, a sampling point can be a micro-pit or micro-well on the chip surface. Different sampling points are separated from each other, and the sequencing reactions of different nucleic acid molecules can occur relatively independently within them.

[0151] According to a preferred embodiment, signal normalization and phase correction can be two sequential steps, for example:

[0152] 1. Using a set of normalization parameters, such as unit signal and offset, normalize the original signal f into a normalized signal s; the normalization parameters in this step will not change in subsequent steps.

[0153] 2. Using a set of phase decoupling parameters, such as leading and lagging, correct the normalized signal s into a corrected signal c; optimize the phase decoupling parameters in this step so that the corrected signal c is as close to an integer as possible.

[0154] According to a preferred embodiment, the aforementioned "making the correction signal sufficiently close to an integer" means that the maximum or minimum value of a certain objective function of the correction signal is less than or greater than a given threshold, for example, the objective function... The minimum value should be less than the threshold 0.5n; otherwise, the sequencing signal quality of the nucleic acid sequence is considered to be too poor.

[0155] According to preferred embodiments, the methods for optimizing parameters include, but are not limited to, grid search, exhaustive search, gradient descent, stochastic gradient descent, Newton's method, Hessian matrix method, interior point method, sequential quadratic programming, active set method, trust region method, Levenberg-Marquardt method, heuristic search, etc., where heuristic search includes, but is not limited to, genetic algorithms, simulated annealing algorithms, ant colony optimization, harmonic optimization, spark optimization, particle swarm optimization, immune optimization, gene expression algorithms, etc. The optimization methods mentioned here are all conventional mathematical concepts.

[0156] In some preferred embodiments, the method in patent EP19181402 can be used to correct the signal and make the corrected signal as close as possible to an integer.

[0157] According to the preferred embodiment, based on "making the correction signal as close to an integer as possible", the parameters are further optimized so that:

[0158] (A) The correlation between the front and rear wheels of the correction signal c is minimal, for example:

[0159] a) Make (c1,c2,…,c n-1 (c2,c3,…,c) and (c2,c3,…,c) n Minimize the correlation coefficient between them; or,

[0160] b) Make (c1,c2,…,c n-2 (c3,c4,…,c) and (c3,c4,…,c) n Minimize the correlation coefficient between them;

[0161] The aforementioned correlation coefficients include, but are not limited to, Pearson correlation coefficient, Spearman correlation coefficient, the negative of cosine distance, the negative of Euclidean distance, and the negative of Manhattan distance;

[0162] (B) The distribution of the correction signal should be as close as possible to the reference distribution, for example as follows:

[0163] a) The reference distribution is To minimize the Kullback-Leibler divergence between the distribution of the corrected signal and the reference distribution; or,

[0164] b) The reference distribution is To minimize the Kullback-Leibler divergence between the distribution of the corrected signal and the reference distribution;

[0165] The distribution of the correction signal can be obtained in a variety of ways, such as interval statistics or kernel density estimation.

[0166] According to the preferred embodiment, if a set of parameters cannot be found that makes the correction signal sufficiently close to an integer, the sequencing signal quality of the nucleic acid sequence is considered to be too poor. In this case, any one of the following choices or a combination thereof can be made:

[0167] 1) Do not output the correction results for this nucleic acid sequence;

[0168] 2) Output a result indicating that the correction failed, such as NaN, None, Null, Nil, null value, etc.

[0169] 3) Output the correction results and report relevant information in the log;

[0170] 4) Output the correction result and mark it with a flag indicating failure.

[0171] According to a preferred embodiment, multiple sequencing reactions are executed in parallel, wherein step AD is performed for each sequencing reaction. High-throughput sequencing reactions typically process at least about 10 sequences simultaneously. 3 10 4 10 5 10 6 10 7 108 Or 10 9 Sequencing reactions can be performed on one or more sequences.

[0172] In some implementations, population correction can be performed, meaning that a population of multiple nucleic acid sequencing signals can be corrected together.

[0173] Specifically, suppose that in a certain sequencing run, n rounds of sequencing signals were obtained for m nucleic acid sequences:

[0174]

[0175] A phase mismatch is constructed using a set of parameters, and the correction signals for the above sequencing signals are calculated using the phase mismatch:

[0176]

[0177] We attempt to find a set of parameters that minimizes the following objective function:

[0178]

[0179] Alternatively, you can construct a similar objective function by referencing the seven objective functions listed below:

[0180] 1) Make minimize;

[0181] 2) Make minimize;

[0182] 3) Make minimize

[0183] 4) Make max|c i -[c i ]|minimize;

[0184] 5) Make minimize;

[0185] 6) Make minimize;

[0186] 7) Make minimize;

[0187] Among them, [c i ] is c i The rounding here refers to rounding up or down.

[0188] In some implementations, when c i When < 0, we can let [c iTake other meaningful values, such as 1. Also, satisfy other validity checks, such as the strength being less than 0 or greater than the maximum reasonable value for measurement.

[0189] In the aforementioned population-based calibration, to avoid overfitting, the sequencing signals in the population can have the same parameters. These parameters include unit signal, offset, and attenuation coefficient for normalization, as well as lead and lag coefficients for phase misalignment correction. In non-population-based calibration, each sequencing signal is calibrated individually, so the parameters used for calibration are generally different for different sequencing signals. However, in population-based calibration, several parameters can be selected, and all sequencing signals in the population use the same parameters for calibration, which greatly reduces computation and speeds up analysis. Preferably, all sequencing signals in the population use the same attenuation coefficient, lead coefficient, and lag coefficient for calibration. Furthermore, it is preferable that all sequencing signals in the population use the same offset for calibration.

[0190] In some implementations, group correction can be performed first, and then non-group correction can be performed based on the results of the group correction.

[0191] On the other hand, the present invention discloses a gene sequencing system, comprising: (a) a sequencer that detects multiple signals generated in response to multiple nucleotide substrates introduced during sequencing; and (b) a computer including executable code stored thereon, which obtains a corrected nucleic acid sequence using the aforementioned method for correcting sequence information errors generated by nucleic acid sequencing. The method for correcting sequence information errors generated by nucleic acid sequencing includes...

[0192] A. Sequencing the nucleic acid to be tested to obtain the raw signal corresponding to the nucleic acid sequence;

[0193] B. Construct a phase mismatch using a set of parameters, and use the phase mismatch to calculate the correction signal of the nucleic acid to be tested;

[0194] C. Optimize the parameters so that the correction signal is as close to an integer as possible;

[0195] D. Output the nucleic acid sequence corresponding to the correction signal;

[0196] The parameters include the lead factor and the lag factor; the phase mismatch refers to the change in sequencing results caused by the lead and / or lag.

[0197] Each of the features discussed in the specific embodiments of the method for correcting sequence information errors generated by nucleic acid sequencing of the present invention is equally applicable to specific embodiments of gene sequencing systems. As shown above, all other features are not repeated here and should be considered as repetitions by reference. Those skilled in the art will understand how easily the features identified in these embodiments can be combined with basic feature sets identified in other embodiments.

[0198] Example 1

[0199] 1. Synthesize a DNA sequence through chemical synthesis:

[0200] CAAGGGGCATTATTACGGTGGCGAGTTATATGAAGGAAAGTGTCTCGGCC

[0201] 2. The DNA sequence was prepared into a library and sequenced using the SNA sequencing method, which involves sequentially adding C, A, G, and T for 60 rounds. Since the first round only extends the C base, the theoretical signal intensity is 1. Therefore, the sequencing signal was normalized by dividing the sequencing signals from all rounds by the first round signal, resulting in the normalized signal s. Figure 1 As shown, the original sequencing signal is affected by phase loss factors and deviates significantly from the ideal signal.

[0202] 3. Enumerate a series of leading parameters: 0, 0.2%, 0.4%, ..., 3.0%.

[0203] 4. Enumerate a series of lag parameters: 0, 0.1%, 0.2%, ..., 1.5%.

[0204] 5. Enumerate all combinations of the above leading and lagging parameters, for a total of 256 combinations.

[0205] 6. For the lead and lag parameters in each combination, the original sequencing signal is corrected using the method described in patent US8364417B2 to obtain the correction signal c corresponding to that parameter combination.

[0206] 7. Take the first 50 cycles of the correction signal c and calculate... As the objective function value corresponding to this parameter combination, [c i ] is c i Rounding. A heatmap is plotted showing the objective function values ​​for all parameter combinations, as shown below. Figure 2 As shown.

[0207] 8. In Figure 2 Find the minimum of the objective function ( Figure 2(Based on the position marked by the pentagram in the middle), the estimated leading factor is 1.8% and the lagging factor is 0.7%.

[0208] 9. Using the estimated lead and lag coefficients described above, correct the original sequencing signal to obtain the final corrected signal, such as... Figure 3 As shown, the corrected signal is quite close to the ideal signal, and the difference between the two is caused by the noise of the sequencing system.

[0209] Example 2

[0210] 1. Synthesize a DNA sequence through chemical synthesis:

[0211] ACTATGCGACTGGACGGCCTATTTACCGAGAGCTGTTCGAAGGCTGGTTGAATACATGGCAGAAGATTGAGGTGTCCTAAACTTACGCGGCCATAACACCTTAGCC

[0212] 2. The DNA sequence was prepared into a library and sequenced using a 2+2 sequencing method, with a combination of M / K, meaning that A and C were added in odd-numbered rounds, and G and T were added in even-numbered rounds, for 50 rounds of sequencing. The original sequencing signal was normalized using the signal normalization method described in patent CN202010089595.8, resulting in the normalized signal s. Figure 4 As shown, the original sequencing signal is affected by phase loss factors and deviates significantly from the ideal signal.

[0213] 3. Enumerate a series of leading parameters: 0, 0.1%, 0.2%, ..., 2.0%.

[0214] 4. Enumerate a series of lag parameters: 0, 0.1%, 0.2%, ..., 2.0%.

[0215] 5. Enumerate all combinations of the above leading and lagging parameters, totaling 441 combinations.

[0216] 6. For the lead and lag parameters in each combination, the original sequencing signal is corrected using the method described in patent CN107958138B. The normalized signal s is rounded to the nearest integer and treated as the assumed ideal signal h. A flux matrix T is constructed, and the generalized inverse matrix of the flux matrix T is multiplied by the normalized signal s (i.e., the method described in patent CN107958138B) to obtain the correction signal c corresponding to the parameter combination.

[0217] 7. Take the first 30 cycles of the correction signal c and calculate... As the objective function value corresponding to this parameter combination, [c i ] is c iRounding. A heatmap is plotted showing the objective function values ​​for all parameter combinations, as shown below. Figure 5 As shown.

[0218] 8. In Figure 5 Find the minimum of the objective function ( Figure 5 (Based on the position marked by the pentagram in the middle), the estimated leading factor is 1.2% and the lagging factor is 0.5%.

[0219] 9. Using the estimated lead and lag coefficients described above, correct the original sequencing signal to obtain the final corrected signal, such as... Figure 6 As shown, the corrected signal is quite close to the ideal signal, and the difference between the two is caused by the noise of the sequencing system.

[0220] Example 3

[0221] 1. Purified λ phage genomic DNA samples were purchased from New England Biolabs. Sequencing libraries were constructed using a 2+2 sequencing method (M / K combination), where substrates A and C were added in odd-numbered rounds, and substrates G and T were added in even-numbered rounds, for a total of 50 rounds of sequencing. To facilitate signal normalization, the sequencing library included a calibration sequence AGCCGTATGACCTGA after the sequencing primer binding region. The raw sequencing signals could be normalized using the signal normalization method described in patent CN202010986791.5. The calibration process for each sequencing signal is detailed in Example 1.

[0222] 2. After sequencing, a large number of raw fluorescent signals of DNA sequences were obtained, and their distribution is shown in the figure. Figure 7 As shown in the figure above, the intensity distributions of its normalized and corrected signals are as follows. Figure 7 The middle and lower figures are shown. Because different DNA sequences extend at different bases and have different unit signals in each sequencing round, the original signal intensity distribution is broad and single-peaked. After normalization, the differences in unit signals among all DNA sequences are eliminated, resulting in a multi-peak distribution, reflecting the differences in the number of bases extended in each sequencing round. However, as the sequencing reaction progresses, the peaks in the normalized signal intensity distribution become increasingly broad and gradually merge, reflecting the signal disorder caused by phase loss. After correction, the corrected signal also exhibits a multi-peak distribution, and these peaks are narrower than the normalized signal peaks, not widening or merging with the progress of the sequencing reaction, reflecting that the effect of phase loss is greatly reduced.

[0223] 3. During the calibration process, the histogram of the estimated parameters (unit signal, odd-numbered wheel offset, even-numbered wheel offset, lead, lag) is shown below. Figure 8 As shown.

[0224] 4. Using a degenerate alignment method (see patent CN202010525168.X), the correction signal was encoded into a degenerate sequence and aligned to the reference genome of λ phage using alignment software bwa. The alignment rate was 80.43%, the genome coverage was 99.99%, the insertion error rate was 3.42%, the deletion error rate was 3.54%, and the total error rate was 6.96%.

[0225] Example 4

[0226] 1. Purified λ phage genomic DNA samples were purchased from New England Biolabs, and a sequencing library was constructed. Sixty rounds of MK sequencing were performed on the library. To facilitate signal normalization, a calibration sequence AGCCGTATGACCTGA was added after the sequencing primer binding region for normalization.

[0227] 2. For each sequence, try to find a set of phase decoupling parameters that make the normalized signal as close to an integer as possible, and use the estimated optimal parameters to correct the signal.

[0228] 3. For each sequence, using the phase-corrected signal as a known sequence, attempt to find a set of normalization parameters that make the fluorescence signal calculated based on the normalization parameters, the previously estimated phase-correction parameters, and the known sequence as close as possible to the measured signal.

[0229] 4. For each sequence, normalize the signal using the latest estimated normalization parameter, find the phase decoupling parameter again to make the normalized signal as close to an integer as possible, and correct the signal using the estimated optimal parameter.

[0230] 5. Repeat steps 3 and 4 until the correction signals before and after re-estimating the phase loss parameters are equal, and output the correction signal at this point as the sequencing result.

[0231] 6. Figure 9 This demonstrates how the distribution of each parameter changes when the iteration finally converges compared to when only one estimate is performed.

[0232] 7. Figure 10 The results show that the objective function of the alternating estimation method is optimized to be smaller, which means that the final output signal is closer to an integer.

[0233] 8. Figure 11 The results show that the alternating estimation method has fewer average errors per sequence and a higher proportion of perfectly corrected signals.

[0234] 9. Using a degenerate alignment method (see patent CN202010525168.X for the alignment method), the correction signal was encoded into a degenerate sequence and aligned to the reference genome of λ phage using the alignment software bwa. The alignment rate was 84.39%, the genome coverage was 99.99%, the insertion error rate was 2.58%, the deletion error rate was 3.13%, and the total error rate was 5.7%.

[0235] The above description is merely a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for correcting sequence information errors generated during nucleic acid sequencing, characterized in that, include: A. Sequencing the nucleic acid to be tested to obtain the raw signal corresponding to the nucleic acid sequence; B. Construct a phase mismatch using a set of parameters, and use the phase mismatch to calculate the correction signal of the nucleic acid to be tested; C. Optimize the parameters so that all or part of the correction signal is close to an integer; D. Output the nucleic acid sequence corresponding to the correction signal; The parameters include the lead coefficient and the lag coefficient; The phase mismatch refers to the change in sequencing results due to lead and / or lag.

2. The method according to claim 1, characterized in that, The set of parameters also includes unit signal, offset, and attenuation coefficient.

3. The method according to claim 1 or 2, characterized in that, In the sequencing process, one, two, three, or four nucleotide substrate molecules are added in each sequencing reaction.

4. The method according to claim 1 or 2, characterized in that, The sequencing process refers to a sequencing process with the 3' end open; the sequencing reaction involves adding one, two, or three types of nucleotides.

5. The method according to claim 4, characterized in that, The sequencing process includes two different sequencing reagents: a first sequencing reagent and a second sequencing reagent; the two sequencing reagents are added cyclically; wherein the first sequencing reagent contains at least two different nucleotide monomers having a detectable label; the second sequencing reagent contains one or more nucleotide monomers having a detectable label, and the nucleotide monomers are different from those present in the first sequencing reagent, and wherein the second sequencing reagent is provided after the first sequencing reagent is provided, and the signal generated by the detectable label is detected after the nucleotide monomers are incorporated into the nucleic acid to be tested.

6. The method according to claim 1 or 2, characterized in that, Set the correction signal The phrase "making all or part of the correction signal approach an integer" refers to making a certain objective function of the correction signal reach its maximum or minimum value. This objective function and its maximum or minimum value include: 1) Make minimize; 2) Make minimize; 3) Make minimize( ); 4) Make minimize; 5) Make minimize; 6) Make minimize; 7) Make minimize; in, yes The rounding refers to rounding up or down.

7. The method according to claim 1 or 2, characterized in that, The original signals from multiple nucleic acid sequences are calibrated together.

8. The method according to claim 1 or 2, characterized in that, Based on making the correction signal close to an integer, the parameters are further optimized so that: (A). The correlation between the front and rear wheels of the correction signal c is minimal, including: a) make and Minimize the correlation coefficient between them; or, b) make and Minimize the correlation coefficient between them; The correlation coefficients include Pearson correlation coefficient, Spearman correlation coefficient, the negative of cosine distance, the negative of Euclidean distance, and the negative of Manhattan distance; (B). The distribution of the correction signal c should be as close as possible to the reference distribution, including: a) The reference distribution is This minimizes the Kullback-Leibler divergence between the distribution of the corrected signal and the reference distribution; or, b) The reference distribution is This minimizes the Kullback-Leibler divergence between the distribution of the corrected signal and the reference distribution.

9. The method according to claim 1, characterized in that, Multiple sequencing reactions are executed in parallel, and step AD is performed for each sequencing reaction.

10. A gene sequencing system, comprising: (a) A sequencer that detects multiple signals generated in response to multiple nucleotide substrates introduced during sequencing; (b) A computer storing executable code that, when executed by the computer, causes the computer to obtain a corrected nucleic acid sequence using the method of any one of claims 1-9.

Images