A method for evaluating mixed signals in gene sequencing

Abstract

The present invention discloses a method for evaluating mixed gene sequencing signals. By acquiring data, constructing a dictionary, sparsely representing the dictionary, and filtering the results, it is possible to determine whether the sequencing signal of a certain site is a signal of a single fragment (monoclonal) or a mixture of multiple fragment signals (polyclonal).

Description

Technical Field

[0001] The invention relates to a method for evaluating a gene sequencing mixed signal, and belongs to the field of molecular sequencing data processing. Background Art

[0002] Gene sequencing is the process of decrypting gene sequences. The application scope of gene sequencing is very wide, and common disease prediction and diagnosis, virus identification, etc. can all be applied to gene sequencing. In high-throughput DNA sequencing, in principle, in the obtained image data, the signal within a certain area is the result of the amplification reaction of the same DNA molecule (i.e., monoclonal signal), which is also one of the foundations for decoding the DNA sequence through the light signal generated by the reaction. In actual operation, due to process limitations, such as the injection of reaction attachment microspheres and the injection of DNA templates, it is inevitable that multiple DNA templates will enter the same observation area, causing the time series signal observed in this area to be a mixed signal (or simply referred to as a polyclonal signal) generated by the superposition of multiple reaction templates. The DNA sequence obtained by subsequent correction based on this signal does not reflect the true composition of the DNA molecule being tested. Therefore, a method is needed to identify such signals and further separate the original signals.

[0003] Ion torrent has found two indicators that can reflect the characteristics of different types of signals (mixed or non-mixed), PPF (percent positive flows) and SSQ (sum of squares). Different types of signals are identified by screening the values ​​of the two indicators found. PPF refers to the percentage of the number of injection fluids (flows) that produce positive signals. The higher the PPF, the more likely it is a mixed signal. SSQ refers to the sum of the squares of the signal value and the nearest positive number in a flow. The larger the SSQ, the more likely it is a mixed signal. Ion torrent allows users to customize the flow round number interval used to estimate the PPF and SSQ thresholds for reasonable polyclonal screening in the analysis program. However, Ion torrent's method is not suitable for degenerate sequencing. The main reason is that the PPF indicator does not have a good degree of differentiation for different types of signals when performing degenerate sequencing. Therefore, it is necessary to develop a method that can effectively distinguish the types of signals obtained by degenerate sequencing. Summary of the invention

[0004] Ion torrent uses the signal feature combined with the classification method to predict whether the signal is a mixed signal generated by multiple templates, which is an indirect means. Regardless of whether it is non-degenerate sequencing or degenerate sequencing, polyclonal signals are linear combinations of more than two monoclonal signals, and their coefficients are determined by the ratio of each DNA molecular clone (i.e., source) to the total; therefore, the usual characteristics of polyclonal signals in high-throughput sequencing can be represented by the linear combination of multiple source signals, but not by a single source signal. The key to identifying polyclonal signals is to determine whether the representation of a single source signal exists.

[0005] The present invention discloses a method for evaluating mixed gene sequencing signals, which specifically adopts sparse coding technology to realize this idea, and can determine whether the sequencing signal of a certain site is a signal of a single fragment (monoclonal) or a mixture of multiple fragment signals (polyclonal).

[0006] Specifically, the present invention discloses a method for evaluating a gene sequencing mixed signal, which is characterized by comprising the following steps:

[0007] A. Obtaining data: Sequencing obtains sequencing signal intensity results corresponding to the bases of the sequence to be tested, and represents the intensity results as a degenerate polymer sequence or a homopolymer sequence as the sequence to be evaluated; the degenerate polymer sequence is a sequence composed of the number arrangement of degenerate polymers, and the homopolymer sequence is a sequence composed of the number arrangement of monomers constituting the homopolymer;

[0008] B. Constructing a dictionary: determining a reference sequence, representing the reference sequence corresponding to the sequence to be tested as an ideal degenerate polymer sequence or an ideal homopolymer sequence corresponding to the sequencing method, and extracting sub-signals of length k from the ideal degenerate polymer sequence or the ideal homopolymer sequence bit by bit to construct a dictionary;

[0009] C. Dictionary sparsification: extract subsequences from the dictionary and select subsequences that are close to the sequence to be evaluated as the final dictionary;

[0010] D. Sparse representation: Use an optimization algorithm to find a vector that can be left-multiplied by the final dictionary matrix so that the sparsity of the vector and the distance to the first k bits of the sequence to be evaluated are minimized at the same time. The found vector is called a sparse vector;

[0011] E. Filtering results: The sparse vector is analyzed according to the set sparsity threshold and mixedness threshold. When the sparsity is not higher than the sparsity threshold and the mixedness is higher than the mixedness threshold, the sequence to be evaluated corresponding to the sparse vector is determined to be a monoclonal signal; when the sparsity is not higher than the sparsity threshold and the mixedness is not higher than the mixedness threshold, the sequence to be evaluated corresponding to the sparse vector is determined to be a polyclonal signal or a mixed signal; when the sparsity is higher than the sparsity threshold, the sparse representation fails and the signal cannot be determined to be monoclonal or polyclonal.

[0012] According to a preferred embodiment, the sequencing includes multi-base sequencing.

[0013] According to a preferred embodiment, in step A, the first k bits of the degenerate polymer sequence or the homopolymer sequence can be selected as the sequence to be evaluated.

[0014] According to a preferred embodiment, k is preferably in the range of 8-20, more preferably in the range of 10-15.

[0015] According to a preferred embodiment, the sub-signals of length k constitute each column of the dictionary.

[0016] According to a preferred embodiment, step C for determining the final dictionary includes: extracting consecutive n bits (n < k) from each sequence to be evaluated as the subsequence to be evaluated, traversing and comparing the first n bits of the subsequence to be evaluated with the first n bits of each sub-signal in the aforementioned dictionary, and taking the items in the dictionary that are the same as the rounded first n bits of the actual signal as the alternative set; in the alternative set, calculating the distance between the first k bits of the sequence to be evaluated and each item in the dictionary one by one, and taking out the first m items in the set of elements arranged in ascending order of distance as the final dictionary.

[0017] According to a preferred embodiment, the value of n can be 3 or 4 or 5; the value range of m is 20-300, preferably 50-150.

[0018] According to a preferred embodiment, the distance includes but is not limited to Pearson correlation coefficient, Spearman correlation coefficient, average mutual information, Euclidean distance, Hamming distance, Chebyshev distance, Mahalanobis distance, Manhattan distance, Minkowski distance, the maximum or minimum value of the absolute value of the difference between corresponding signals; the optimization algorithms include but are not limited to matching pursuit, orthogonal matching pursuit, weak matching pursuit, threshold method, basis pursuit, IRLS algorithm, Lasso algorithm, weighted support vector machine algorithm.

[0019] According to a preferred embodiment, the value range of the sparsity threshold is 2-10, preferably 2-5, more preferably 2-3; the value range of the mixing threshold is preferably 0.6-1, preferably 0.8-1, more preferably 0.9-1.

[0020] The present invention also discloses an evaluation method for a mixed signal of gene sequencing, which is characterized by including the following steps:

[0021] 1) Sequencing to obtain the results corresponding to the bases of the sequence to be measured;

[0022] 2) Compiling the reference sequence corresponding to the sequence to be measured into the theoretical results corresponding to the sequencing method;

[0023] 3) Divide the theoretical results of the compilation of the reference sequence into possible result clusters of 6-50 bases;

[0024] 4) Compare the sequencing results corresponding to the bases of the sequence to be tested with the possible result clusters in step 3).

[0025] According to a preferred embodiment, the polyclonal signal or the monoclonal signal is determined using any of the methods described above.

[0026] According to a preferred embodiment, the identified polyclonal signal or mixed signal is discarded, the signal that cannot be determined as monoclonal or polyclonal is discarded, and the identified monoclonal signal is separated for subsequent data processing and analysis.

[0027] Beneficial Effects of the Invention

[0028] Compared with the prior art, the method of the present invention has the following advantages:

[0029] 1. The method of the present invention can solve the problem of identifying polyclonal signals in degenerate sequencing, while the methods of the prior art lose their effectiveness in degenerate sequencing.

[0030] 2. Compared with the method of classifying signals according to features in the prior art, the accuracy of the present invention is higher.

[0031] 3. This method has a wider scope of application. In addition to being applicable to the evaluation of sequencing signals of degenerate sequencing, it can also be used to evaluate non-degenerate sequencing signals.

[0032] 4. This method reflects the essential composition of the signal and does not rely on the steps of finding, mining and reconstructing other features of the signal, which usually requires more specialized domain knowledge and experience and consumes considerable energy. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 . A schematic diagram of the process of obtaining the final dictionary using the method of the present invention.

[0034] Figure 2 . Schematic diagram of the process of identifying sequencing signals using sparse representation. DETAILED DESCRIPTION

[0035] The following discussion is intended to enable those skilled in the art to understand and use the disclosed methods, and is provided in the context of specific applications and their requirements. Various modifications to the disclosed embodiments will be apparent to those skilled in the art, and the general principles defined herein may be applied to other embodiments and applications without departing from the spirit and scope of the disclosed technology. Therefore, the disclosed methods are not limited to the embodiments shown, but include the widest scope consistent with the principles and features disclosed herein.

[0036] Unless defined otherwise, all technical and scientific terms used herein have the same meanings as commonly understood by one of ordinary skill in the art.

[0037] During gene sequencing, generally, there is a DNA fragment on each micropit (or each site), and after amplification, a cluster of the fragment is formed. Then during the second-generation sequencing, after chemical reaction, the site will undergo the same reaction and extend the same number of bases. In fact, when sequencing is performed, there may be more than one DNA fragment on a site, which we can call polyclonal. It is difficult to judge whether a signal is a polyclonal signal or a monoclonal signal only from the perspective of the number of sequencing signals. The existence of polyclonal signals generally affects the acquisition of sequencing signals and makes sequencing signals difficult to interpret. Therefore, signals determined to be polyclonal can generally be used as discarded data. The present invention provides a method for judging polyclonal signals, which aims to distinguish mixed signals generated due to experimental process limitations by representing the signal after phase correction during DNA sequencing in a specific space, so as to improve the accuracy of sequencing. The present invention is applicable to a sequencing method for determining a DNA sequence by detecting the signal multiplier, and the types of signals include but are not limited to fluorescent signals, electrical signals, and chemical signals. Under the condition of meeting the above conditions, the present invention is not limited to degenerate sequencing, and can also be used for non-degenerate sequencing.

[0038] Specifically, the first aspect of the present invention discloses a method for evaluating a gene sequencing mixed signal, characterized in that it comprises the following steps:

[0039] A. Obtaining data: Sequencing obtains sequencing signal intensity results corresponding to the bases of the sequence to be tested, and represents the intensity results as a degenerate polymer sequence or a homopolymer sequence as the sequence to be evaluated; the degenerate polymer sequence is a sequence composed of the number arrangement of degenerate polymers, and the homopolymer sequence is a sequence composed of the number arrangement of monomers constituting the homopolymer;

[0040] B. Constructing a dictionary: determining a reference sequence, representing the reference sequence corresponding to the sequence to be tested as an ideal degenerate polymer sequence or an ideal homopolymer sequence corresponding to the sequencing method, and extracting sub-signals of length k from the ideal degenerate polymer sequence or the ideal homopolymer sequence bit by bit to construct a dictionary;

[0041] C. Dictionary sparsification: extract subsequences from the dictionary and select subsequences that are close to the sequence to be evaluated as the final dictionary;

[0042] D. Sparse representation: Use an optimization algorithm to find a vector that can be left-multiplied by the final dictionary matrix so that the sparsity of the vector and the distance to the first k bits of the sequence to be evaluated are minimized at the same time. The found vector is called a sparse vector;

[0043] E. Filtering results: The sparse vector is analyzed according to the set sparsity threshold and mixedness threshold. When the sparsity is not higher than the sparsity threshold and the mixedness is higher than the mixedness threshold, the sequence to be evaluated corresponding to the sparse vector is determined to be a monoclonal signal; when the sparsity is not higher than the sparsity threshold and the mixedness is not higher than the mixedness threshold, the sequence to be evaluated corresponding to the sparse vector is determined to be a polyclonal signal or a mixed signal; when the sparsity is higher than the sparsity threshold, the sparse representation fails and the signal cannot be determined to be monoclonal or polyclonal.

[0044] There are many DNA sequencing methods. According to the number of sample molecules involved in each sequencing reaction, it can be divided into multi-molecule sequencing and single-molecule sequencing; according to the type of sequencing reaction, it can be divided into synthesis sequencing, ligation sequencing, excision sequencing, etc.; according to the number of nucleotides that can be measured in each round of reaction, it can be divided into single-base sequencing and multi-base sequencing, and the multi-base sequencing refers to that one or more nucleotides can be measured in each round of sequencing reaction. The common Illumina sequencing method belongs to single-base sequencing, which uses a 3-end blocking method, and each sequencing extends one base, so its signal is 0 and 1. It should be noted that the signals 0 and 1 here are relatively comparative values. It can be known that when sequencing is actually performed, the reaction is first performed, and then the reaction signal is obtained using a camera device such as CCD. However, due to the high-throughput sequencing, the range of each data point or each cluster of sequences to be tested is very small during the reaction, generally between 0.1-3 microns, and such a small chemical reaction is difficult to be accurately measured. Therefore, the adjusted CCD obtains an image, for example, in which the brightness of the bright spot, or the reaction point, is 1000, and the brightness of the non-reaction point is 500 (a simple example, not an actual experiment). Then the data point with a brightness close to 1000 can be defined as 1, and the data point with a brightness closer to 500 can be defined as 0.

[0045] In the present invention, the sequencing includes multi-base sequencing. Degenerate sequencing is a type of multi-base sequencing, which is a new form of sequencing-by-synthesis technology. It is a gene sequencing method that is not closed at the 3' end. It simultaneously injects a degenerate substrate composed of at least two base substrates in a sequencing reaction, and the number of bases extended in each reaction may be multiple, such as 1, 2, 3, 4, 5, 6, etc. When the multi-base sequencing method is compared with the sequencing method that extends one base each time, the subsequent data processing is obviously different. In the process of sequencing, multiple bases are extended each time, then corresponding to the previous Illumina example, illustratively, the signal intensity may be 1000, 2000, 3000, 4000, 5000, 6000, etc. It can be seen that the more finely the sequencing signal needs to be distinguished, the greater the difficulty. The sequencing process is affected by various factors, such as incomplete reaction, insufficient reagent entry, unstable light source intensity, uneven light source intensity, the influence of the reaction itself, etc., which may lead to the reaction being ahead or behind. In the later stage of the reaction, these accumulated factors will become more and more serious. It is even more difficult to define the signal value 1. In addition, for excision sequencing, an indefinite number of bases are excised in each sequencing reaction during the sequencing process, and the sequencing signal can also be evaluated using the method described in the present invention.

[0046] During actual sequencing, as the sequencing reaction proceeds, the DNA molecules of the same cluster may extend to different lengths, and this situation accumulates as the sequencing length increases. This will cause certain difficulties in interpreting the sequencing signal and destroy the euploidy of the sequencing signal. For example, when a DNA fragment is sequenced, the sequencing signal obtained at the beginning is relatively close to an integer by comparison. At this time, the sequencing reaction is relatively sufficient and there are not many heterogeneous reactions. When the sequencing reaches 50 bases (a simple example, not an actual experiment), the sequencing signal result may become not close to an integer. Signals that are not close to integers will affect the accuracy of judging polyclonal signals.

[0047] For step A of the present method, in the process of sequencing, a common multi-base sequencing method is first used to sequence a gene sequence, and a sequencing signal corresponding to the gene sequence is obtained by reading data by a sequencer. This sequencing signal is generally converted from the intensity value or current value of the image. The process of sequencing signal processing does not belong to the specific scope of the present invention. The applicant's previous patents CN201510944878.5, or CN202010061629.2, or CN201610899880.X, etc., describe in detail the method for obtaining a more accurate sequencing signal or the processing method of the sequencing signal. Here, the sequencing signal processing process is simply described: after sequencing to obtain the original sequencing data, first determine whether the data is compliant. For compliant data, extract the sequencing signal intensity result, subtract background noise and normalize it. Generally, the result is an integer or floating-point vector arranged according to the acquisition time, for example, the sequencing signal of a single base is X, then when the sequencing signal obtained is 2X, it can be known that this sequencing may extend 2 bases. Taking 2+2 degenerate sequencing as an example for multi-base sequencing, two nucleotide substrates are introduced into each round of sequencing reaction. For all odd-numbered rounds, two nucleotide substrates are introduced, while for even-numbered rounds, another two nucleotide substrates are introduced. For example, A / C is introduced into the odd rounds, and G / T is introduced into the even rounds. This method can ensure that base extension will occur in each round of reaction, and there will be no empty reaction. Therefore, the reaction speed is faster. The sequence obtained by the degenerate sequencing reaction is not an accurate base sequence, but a degenerate base sequence. For example, MMKMKKKMMMKKKM…, where M represents A / C and K represents G / T. For degenerate sequencing, the sequencing signal intensity result is expressed as a degenerate polymer sequence. A degenerate polymer sequence refers to a sequence composed of the number of degenerate polymers. For example, for the sequence MMKMKKKMMMKKKM, its degenerate polymer sequence is (2,1,1,3,3,3,1). In addition, the polybase sequencing method also includes sequencing methods such as Ion torrent, in which only one nucleotide substrate is introduced into each round of sequencing reaction, and the number of nucleotides that are extended can be one or more (not considering the case where no extension occurs), which is a homopolymer, that is, a polymer formed by polymerization of only one monomer. A homopolymer sequence refers to a sequence composed of the number of monomers that constitute the homopolymer. For example, if 4 A's are extended, the homopolymer is AAAA, and the homopolymer sequence is 4. In particular, when the number of extended nucleotides is 1, it cannot be called a homopolymer in a strict sense, but the present invention represents it as 1. The corresponding relationship between base sequence, degenerate base sequence, degenerate polymer sequence (taking MK sequencing as an example) and homopolymer sequence can be seen in the example shown in Table 1, that is, the degenerate polymer sequence is (4,5,3,3,5), and the homopolymer sequence is (3,1,3,2,2,1,3,3,2).

[0048] Table 1. Various sequence representations

[0049] Base sequence A A A C T T T G G C C A T T T A A A C C Degenerate base sequence M M M M K K K K K M M M K K K M M M M M Degenerate polymer sequence 4 5 3 3 5 Homopolymer sequence 3 1 3 2 2 1 3 3 2

[0050] In the present invention, the sequencing signal used is a normalized sequencing signal (referred to as a normalized signal), that is, a ratio corresponding to a single-base sequencing signal.

[0051] According to a preferred embodiment, when the length of the normalized signal of the sequence to be tested is relatively long, it is not necessary to use all the normalized signals as the calculation of monoclonal and polyclonal, and only a part of them needs to be selected as the input of the method. Empirically, the first k continuous normalized signals that are closer to the integer value can be selected as input, that is, the first k positions of the degenerate polymer sequence or homopolymer sequence are selected as the sequence to be evaluated; the value of k is related to the size of the reference sequence and the number of sequencing cycles, and the value of k is preferably 8-20, and more preferably 10-15. Different selection methods have little effect on the final distinction between monoclonal and polyclonal. Polyclonal signals have clearer characteristics, and no matter which position is selected, they are a mixture of multiple sequencing signals. The selected positions are different, but the results obtained are the same.

[0052] For the reference sequence corresponding to the sequence to be tested, it is a relatively easy process to determine the reference sequence. For example, when the sequence to be tested is RNA from a certain tissue of a mouse, the reference sequence is the transcriptome sequence of the mouse; when the sequence to be tested is DNA from a certain tissue of a human, the reference sequence is the human genome sequence. After determining the reference sequence, it is necessary to represent the reference sequence corresponding to the sequence to be tested as an ideal degenerate polymer sequence or an ideal homopolymer sequence corresponding to the sequencing method. That is to say, when the sequence to be tested is subjected to degenerate sequencing, the reference sequence is represented as a degenerate polymer sequence corresponding to the degenerate sequencing method; when the sequence to be tested is subjected to ordinary multi-base sequencing (i.e., only one nucleotide substrate is added in each round of reaction), the reference sequence is represented as the corresponding homopolymer sequence.

[0053] Sparse dictionary learning

[0054] The official name of dictionary learning and sparse representation in academia should be sparse dictionary learning. The algorithm theory includes two stages: dictionary construction stage (Dictionary Generate) and sample representation stage (Sparsecoding with a precomputed dictionary). Sparse representation is essentially a signal representation method, which selects as few basic signals as possible from the original signal, and expresses most or all of the original signal through linear combination of these basic signals. Through sparse representation, a more concise representation of the signal can be obtained, so that it is easier to obtain the information contained in the signal, and it is more convenient to further process the signal. The essence of identifying mixed signals is to calculate whether the signal can be approximately represented by a column vector in the dictionary. The method of the present invention specifically adopts sparse coding technology to realize this idea, thereby completing the identification of polyclonal signals.

[0055] Dictionary construction

[0056] The linear space used to display the sequence to be evaluated is called a "dictionary". The space is a linear algebra concept, represented by an integer matrix, which can be understood as the set of all potential signal types. Constructing a reasonable dictionary for the sequence to be evaluated is the key to this method. Figure 1 The figure shows a schematic diagram of the process of constructing the final dictionary, which is a sparse dictionary. The specific construction process is described in detail below.

[0057] In the present method, the dictionary set is to represent the reference sequence corresponding to the sequence to be tested as an ideal degenerate polymer sequence or an ideal homopolymer sequence corresponding to the sequencing method, and to extract a set of sub-signals of length k from the ideal degenerate polymer sequence or the ideal homopolymer sequence bit by bit, and the sub-signals of length k constitute each column of the dictionary. The value of k is related to the size of the reference sequence and the number of sequencing cycles, and the value of k is preferably 8-20, and more preferably 10-15. In order to facilitate subsequent inquiries, these strings of length k, more generally referred to as k-mer, are stored in a certain data structure to facilitate subsequent inquiries. The data structure includes, but is not limited to, a binary tree, a red-black tree, an inverted index, and the like. Considering the increase in the amount of calculation caused by the subsequent matrix calculation as the genome increases, a certain screening method is adopted in the "dictionary" construction process to continuously select subsets as potential spaces that are truly used for signal representation. The following describes this screening process in detail, i.e., the dictionary sparsification process.

[0058] For each sequence to be evaluated, extract consecutive n bits (n < k) as the subsequence to be evaluated, and traverse and compare the first n bits of the subsequence to be evaluated with the first n bits of each sub-signal in the aforementioned dictionary. The items in the dictionary that are the same as the rounded first n bits of the actual signal are used as the alternative set; if there are no items with exactly the same first n bits in the dictionary, directly proceed to the next step. n can take values of 3, or 4, or 5.

[0059] In the alternative set, calculate the distance between the first k bits of the sequence to be evaluated and each dictionary item one by one. The distance includes but is not limited to Pearson correlation coefficient, Spearman correlation coefficient, average mutual information, Euclidean distance, Hamming distance, Chebyshev distance, Mahalanobis distance, Manhattan distance, Minkowski distance, the maximum or minimum value of the absolute value of the difference between corresponding signals. Take out the first m set elements arranged in ascending order of the distance as the final dictionary. The selection of m should consider the completeness of possible signals and the balance of computational complexity. Depending on the situation of the reference sequence, the value range of m is 20 - 300, preferably 50 - 150.

[0060] Sparse Representation

[0061] Sparse representation is also known as sparse coding. A vector containing mostly zero elements can be called a sparse vector. For a linear equation, x = D * α, when D is an underdetermined matrix, D is called a dictionary, and x is the signal we are interested in. Sparse representation means representing the relationship between x and D with a vector α that is as sparse as possible. For a vector α with a fixed dimension, the more zero elements there are in the vector α, the sparser it is. Mathematically speaking, the sparse representation algorithm is to solve a bi-objective optimization problem that simultaneously minimizes the sparsity and the distance from D * α to x, as Figure 2 shown. The present invention needs to use an appropriate optimization algorithm to find a vector that is as sparse as possible, such that the result of multiplying the aforementioned final dictionary matrix by this vector is as close as possible to the first k bits of the sequence to be evaluated. The found vector is called a sparse vector. The optimization algorithms include but are not limited to matching pursuit, orthogonal matching pursuit, weak matching pursuit, threshold method, basis pursuit, IRLS algorithm, Lasso algorithm, weighted support vector machine algorithm, etc.

[0062] For example, when the degenerate polymer sequence of a sequence to be tested is (2, 1, 2, 1, 1, 2, 1, 1, 2, 2), the first k signals can be selected as the sequence to be evaluated (when k is 8), and the sequence to be evaluated is (2, 1, 2, 1, 1, 2, 1, 1). During actual sequencing, the normalized signal may not be composed entirely of integers, and the first part that is closer to the integer, or the part that is closer to the integer, can be selected. Simply, select the first sequencing normalized signal as input. The first 5-50 continuous normalized signals can be selected as input; more preferably, the first 6-45 signals are selected, and more preferably 7-40 signals, or 8, 9, 10, 11, 12, 13, 14, 15 signals are selected. The front ones do not need to be precisely defined, and can be selected from the first sequencing signal, or from the 2nd, 3rd, or 4th sequencing signals.

[0063] For example, the normalized signals of a gene sequence to be tested are: 2.3, 2.2, 2.5, 3, 1, 1, 2, 1, 3, 3.1, 3.5, 3.6, 1.3, 1.2…; when selecting, the first 9 signals, for example, or the 8 signals starting from the third one can be selected. In theory, the selected position does not affect the final result; there is a clear difference between monoclonal signals and polyclonal signals, and polyclonal signals are a mixture of two or more monoclonal signals. However, since the actual measured signal will be affected by factors such as dephasing and attenuation, in general, the later the signal is collected, the worse the quality is, and the more likely it is to deviate from an integer; therefore, as a preferred embodiment, the front sub-signal should be selected as the signal to be evaluated as much as possible. In the present invention, this law is described in a mathematical way, and no additional meaning is introduced. This method has a relatively high selection accuracy for polyclonal and monoclonal. Of course, when the signal is extremely poor and the sequencing accuracy is, for example, less than 90%, the accuracy may decrease, but it is understandable that this is not due to the accuracy of the method of the present invention.

[0064] Filter results

[0065] A sparse vector α is a vector composed of non-zero elements and zero elements, which contains many zero or close-to-zero elements. Figure 2As shown, for the optimized α obtained by sparse representation, the first step is to check the sparsity, that is, the number of non-zero elements. Only the α vector whose sparsity is not higher than the sparsity threshold is the focus of this method. The value range of the sparsity threshold is 2-10, preferably 2-5, and more preferably 2-3; other situations are collectively referred to as abnormalities. The abnormal situation indicates that the sparse representation algorithm fails to generate a vector with lower sparsity as designed, indicating that the original signal is very different from the basic signal in the dictionary, which may be a low-quality sequence, or the sequence is not in the reference sequence, and the method cannot be used to determine the monoclonal or polyclonal signal. In the case currently observed, it is very rare for the sparsity to be higher than the sparsity threshold.

[0066] Preferably, all elements of the sparse vector are normalized, wherein the largest element in the sparse vector is the part we are concerned about, which represents the degree of proximity between x and a column vector in D, and is referred to as the degree of mixing. When the degree of mixing of a sparse vector is close to the number 1, it means that the sequencing signal is close to an ideal normalized signal of a reference sequence, and the sequencing signal is a monoclonal signal. On the contrary, when the degree of mixing of a sparse vector is not close to the number 1, or there are multiple elements that are not much different, the sequencing signal is judged to be a polyclonal signal or a mixed signal. Accordingly, a threshold value can be set for the value of the degree of mixing in the sparse vector. If it is greater than this threshold value, it is judged to be a monoclonal signal; if it is less than this threshold value, it is judged to be a polyclonal signal, or a mixed signal; this threshold value is called the degree of mixing threshold. Empirically, when the mixed degree threshold is within the range of 0.6-1, the sequencing signal is considered to be a monoclonal signal; preferably, when the mixed degree threshold is within the range of 0.8-1, the sequencing signal is considered to be a monoclonal signal; more preferably, when the mixed degree threshold is within the range of 0.9-1, the sequencing signal is considered to be a monoclonal signal; more preferably, when the mixed degree threshold is within the range of 0.95-1, the sequencing signal is considered to be a monoclonal signal; otherwise, it is a polyclonal signal or a mixed signal.

[0067] Normally, after the polyclonal signal is distinguished by the method of the present invention, the identified polyclonal signal or mixed signal is discarded, and the signal that cannot be determined as monoclonal or polyclonal is discarded, because the mixed information in the polyclonal signal is usually unavailable, and the identified monoclonal signal is separated for subsequent data processing and analysis, such as for base recognition.

[0068] Preferably, mathematical symbols, plus or minus signs can be used to represent different chemical reactions in polybase sequencing. For example, in 2+2 sequencing, the sequence of KM injections is sequenced to obtain a degenerate polymer sequence of 2, 3, 1, 1, 3, 3, 2, 1. Then the K signal can be represented as a positive value, and the M signal can be represented as a negative value, that is, 2, -3, 1, -1, 3, -3, 2, -1.

[0069] In the present invention, the determination method of dictionary and sparse representation in mathematical determination is selected. This method can express the proximity standard of a vector or a set of data to a reference value. In the present invention, the method used does not add any additional restrictions or determinations. Furthermore, in the mathematical algorithm, the actual meaning of each number or calculation is the calculation of the degree of proximity between the actual sequencing signal and the reference sequence. Through a given threshold, it can be determined whether the sequencing signal belongs to a monoclonal signal or a polyclonal signal. Therefore, the mathematical method used is completely one of the specific implementation methods of the claimed technical solution of the present invention.

[0070] D is a matrix composed of ideal normalized signals of possible fragments of the reference sequence. The gene sequence of the reference sequence is known. What it means is that each column is a possible ideal normalized signal. And, when selected, its signal is continuous and has the same dimension as x. The dimension of x is also the number of its sequencing signals. For example, a reference genome of a known sequence contains a base length of 500bp. When the dimension of x is 10; then D is a matrix composed of column vectors composed of 10 signals selected continuously from the ideal normalized signals of the reference genome. Its meaning is to form a matrix with all possible 10 signal situations. Of course, in actual calculations, due to knowing more or less information, there will be simpler ways to simplify the matrix, but the principle is the same.

[0071] Although the matrix calculation method is used in this patent, its actual meaning is still the comparison between the sequencing results and the ideal sequencing results of the reference sequence. The selection of a simple mathematical method does not introduce additional physical meaning or limitation.

[0072] α is a sparse vector, and the calculation method of sparse representation is common knowledge in mathematics. Appropriate distance selection can transform the process of solving α into a convex quadratic optimization problem. There are many solvers for this type of problem, such as interior point methods, fixed-point continuation, in-crowd algorithm, etc. The calculation method of sparse representation is no longer described in detail in the present invention, and this part of the content does not belong to the focus of the present invention.

[0073] The second aspect of the present invention discloses a method for evaluating a gene sequencing mixed signal, characterized in that it comprises the following steps:

[0074] 1) Sequencing to obtain the results corresponding to the bases of the sequence to be tested;

[0075] 2) compiling the reference sequence corresponding to the sequence to be tested into a theoretical result corresponding to the sequencing method;

[0076] 3) dividing the compiled theoretical results of the reference sequence into possible result clusters of 6-50 bases; 4) comparing the results corresponding to the bases of the sequence to be tested obtained by sequencing with the possible result clusters of step 3).

[0077] According to a preferred embodiment, the polyclonal signal or the monoclonal signal is determined using any of the methods described above.

[0078] Each of the features discussed in the first aspect of the present invention's specific implementation section is equally applicable to the specific implementation of the second aspect of the present invention. As shown above, some other features are not repeated here and should be considered to be repeated by reference. Those of ordinary skill in the art will understand how the features identified in these specific implementations can be easily combined with the basic feature groups identified in other specific implementations.

[0079] Example 1

[0080] Perform WS 2+2 degenerate sequencing on λ phage. Select k=10, m=256, and try to distinguish monoclonal and polyclonal signals using sparse representation.

[0081] A signal that is considered monoclonal is:

[0082] 2 -2 2 -2 2 -2 2 -2 2 -2

[0083] The dictionary representing the mono-clonal signal (showing the first 10 columns closest to the signal) is

[0084]

[0085] The sparse representation of the monoclonal signal is as follows. It is actually a column vector. To make it clearer, the transpose of α is shown as follows:

[0086]

[0087] A signal that is considered polyclonal is

[0088] 2 -1 3 -2 2 -1 2 -3 2 -2

[0089] A dictionary of poly-clonal signals (showing the top 10 columns closest to the signal)

[0090]

[0091] Sparse representation of a polyclonal signal

[0092]

[0093] Example 2

[0094] Identification of polyclonal signals in sequencing of strand-specific reverse transcription libraries of tobacco mosaic virus (TMV)

[0095] The genome of tobacco mosaic virus is a single-stranded RNA with a length of 6395. When the cDNA is degenerately sequenced with the KM injection sequence, the repeated 0 sequencing signals are merged to directly obtain a matrix with 1509 columns and 12 rows:

[0096]

[0097] As shown in the figure, each column is a possible sequencing signal in the reference genome, positive numbers are K signals, and negative numbers are M signals.

[0098] Normalizing the matrix by column gives us the dictionary D:

[0099]

[0100] For the following four sets of sequencing signals:

[0101] 1. Ideal monoclonal signal without noise m :

[0102] 4 -2 1 -1 3 -2 1 -1 6 -1 3 -2

[0103] 2. Measured monoclonal signal s′ with noise m :

[0104] 3.653 -2.095 1.221 -1.178 3.190 -1.637 1.429 -0.879 5.978 -1.2642.479 -2.024

[0105] 3. Ideal polyclonal signal without noise p :

[0106] 2.5 -2.5 2.5 -1.5 2 -1.5 2 -1.5 3.5 -1 4.5 -1.5

[0107] 4. Measured polyclonal signal s′ with noisep :

[0108] 2.444 -2.523 2.529 -1.435 1.957 -1.477 1.991 -1.422 3.472 -0.9994.675 -1.546

[0109] Use the lasso method to optimize the following function:

[0110]

[0111] And set the mixing threshold to 0.6,

[0112] For x=s m , The mixing degree was 0.889, which was determined to be monoclonal.

[0113] For x = s′ m , The mixing degree was 0.651, which was determined to be monoclonal.

[0114] For x=s p ,

[0115]

[0116] The mixing degree was 0.367, which was determined to be polyclonal.

[0117] For x = s′ p ,

[0118]

[0119] The mixing degree was 0.366, which was determined to be polyclonal.

[0120] The above is only a preferred specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be easily thought of by a person skilled in the art within the technical scope disclosed by the present invention should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention shall be based on the protection scope of the claims.

Claims

1. A method for evaluating a mixed signal of gene sequencing, characterized in that: The following steps are involved: A. Obtaining data: Sequencing to obtain the sequencing signal intensity results corresponding to the bases of the sequence to be tested, and expressing the intensity results as a degenerate polymer sequence or a homopolymer sequence as the sequence to be evaluated; the degenerate polymer sequence is a sequence composed of the number arrangement of degenerate polymers, and the homopolymer sequence is a sequence composed of the number arrangement of monomers constituting the homopolymer; B. Constructing a dictionary: determining a reference sequence, representing the reference sequence corresponding to the sequence to be tested as an ideal degenerate polymer sequence or an ideal homopolymer sequence corresponding to the sequencing method, and extracting sub-signals of length k from the ideal degenerate polymer sequence or the ideal homopolymer sequence bit by bit to construct a dictionary; C. Dictionary sparsification: extract subsequences from the dictionary and select subsequences that are close to the sequence to be evaluated as the final dictionary; D. Sparse representation: Use an optimization algorithm to find a vector that can be left-multiplied by the final dictionary matrix so that the sparsity of the vector and the distance to the first k bits of the sequence to be evaluated are minimized at the same time. The found vector is called a sparse vector; E. Filtering results: The sparse vector is analyzed according to the set sparsity threshold and mixedness threshold. When the sparsity is not higher than the sparsity threshold and the mixedness is higher than the mixedness threshold, the sequence to be evaluated corresponding to the sparse vector is determined to be a monoclonal signal; when the sparsity is not higher than the sparsity threshold and the mixedness is not higher than the mixedness threshold, the sequence to be evaluated corresponding to the sparse vector is determined to be a polyclonal signal or a mixed signal; when the sparsity is higher than the sparsity threshold, the sparse representation fails and the signal cannot be determined to be monoclonal or polyclonal.

2. The method according to claim 1, characterized in that The sequencing includes polybase sequencing.

3. The method according to claim 1, characterized in that In step A, the first k positions of the degenerate polymer sequence or homopolymer sequence are selected as the sequence to be evaluated.

4. The method according to claim 1, characterized in that The value of k is 8-20.

5. The method according to claim 1, characterized in that The value of k is 10-15.

6. The method according to claim 1, characterized in that The sub-signals of length k constitute each column of the dictionary.

7. The method according to claim 1, characterized in that Step C determines the final dictionary, including extracting continuous n bits from each sequence to be evaluated as a subsequence to be evaluated, traversing and comparing the subsequence to be evaluated with the first n bits of each sub-signal in the aforementioned dictionary, and taking the items in the dictionary that are the same as the first n bits of the actual signal as a candidate set, wherein n<k; in the candidate set, calculating the distance between the first k bits of the sequence to be evaluated and the dictionary items one by one, and taking out the first m items of the set elements arranged in ascending order of distance as the final dictionary.

8. The method according to claim 7, characterized in that The value of n is 3, 4 or 5; the value range of m is 20-300.

9. The method according to claim 7, characterized in that: The value of n is 3, 4 or 5; the value range of m is 50-150.

10. The method according to claim 1, characterized in that The distance includes Pearson correlation coefficient, Spearman correlation coefficient, average mutual information, Euclidean distance, Hamming distance, Chebyshev distance, Mahalanobis distance, Manhattan distance, Minkowski distance, and the maximum or minimum value of the absolute value of the corresponding signal difference; the optimization algorithm includes matching pursuit, orthogonal matching pursuit, weak matching pursuit, threshold method, basis pursuit, IRLS algorithm, Lasso algorithm, and weighted support vector machine algorithm.

11. The method according to claim 1, characterized in that: The sparsity threshold value ranges from 2 to 10; the mixing threshold value ranges from 0.6 to 1.

12. A method for evaluating a mixed signal of gene sequencing, characterized in that: The following steps are involved: 1) Sequencing to obtain the results corresponding to the bases of the sequence to be tested; 2) compiling the reference sequence corresponding to the sequence to be tested into a theoretical result corresponding to the sequencing method; 3) Divide the theoretical results of the compilation of the reference sequence into possible result clusters of 6-50 bases; 4) comparing the sequencing results corresponding to the bases of the sequence to be tested with the possible result clusters in step 3); The method of claim 1 is used to determine a polyclonal signal or a monoclonal signal.

Images