CN105989248B - Data processing method and device for multiple molecular signals - Google Patents

Abstract

The present invention proposes a data processing method for multiple molecular signals. The method calculates the mixing coefficients C(A←B) and C(B←A) between the fluorescent signals of any different molecular cluster A and molecular cluster B, which are used to measure the fluorescent signals of the molecular cluster A and molecular cluster B The severity of mutual mixing, and then, can intervene and reduce the interference of mutual mixing between different molecular clusters, so as to improve the identification accuracy of molecular recognition technology.

Description

Data processing method and device for multiple molecular signals

technical field

The invention relates to the field of data processing of molecular sequencing, in particular to a data processing method and device.

Background technique

Illumina's gene sequence synthesis sequencing technology and the sequencing platform based on this technology are widely used and have been one of the most successful second-generation gene sequencing technologies. It first randomly immobilizes short single-stranded DNA molecules on the surface of the chip, and then forms clusters of single-stranded molecules containing the same sequence through replication. In each round of sequencing, by adding four mononucleotides with different fluorescently labeled reversible terminator groups, the complementary strand of the molecular cluster grows by only one base per round. After that, the surface of the chip is photographed on the laser spectrum of different frequencies. Each channel mainly corresponds to one fluorescence. After the photographing is completed, the terminator group is washed away for the next round of sequencing. In this way, by locating the molecular cluster, then extracting the fluorescent signal of each round of sequencing of the same molecular cluster, and determining the bases identified by each round of sequencing according to the different types of fluorescent signals, and then completing the sequencing of the sequence contained in this molecular cluster . This technology is applied on platforms such as GA, Hiseq and Miseq. For more detailed content of this technology and existing related data processing technology, please refer to the literature Bentley etc., 2008; Li&Speed, 1999; Massingham&Goldman, 2012; Whiteford etc., 2009, etc.

But this technology still has many deficiencies. In addition to spectral cross-color and phase loss, the following problems are also included: First, due to the limitation of the accuracy of the sequencer, the scenes in different photos have displacement and slight stretching from less than one pixel to tens or even hundreds of pixels. At the same time, there are also small non-zero, random light intensity background values at those positions where the molecular clusters do not emit light. What's more troublesome is that since the molecules of the sequence fragments are randomly dropped on the chip, the formed molecular clusters may be relatively close together, and the signals of these relatively close molecular clusters in each photo will be mixed together Mutual influence (as shown in Fig. 1A, Fig. 1B, Fig. 1C and Fig. 2, Fig. 1A is a partial schematic diagram of a picture of a channel of a round of sequencing measured by the prior art, showing the molecular clusters that are closer; Fig. Schematic diagram of partial sequencing rounds of two closely spaced molecular cluster signals after correction of spectral cross-color and phase dephasing. The 13th base of the first molecular cluster was misrecognized; Figure 1C is a schematic diagram of the mixed signals of adjacent molecular clusters; Figure 2 is a schematic diagram of the positioning and signal mixing of two molecular clusters that are closer together When the two molecular clusters are close to each other, the positions of the two molecular clusters determined according to the maximum value of the fluorescence signal will be close to each other, and the signals will be mixed at the same time). As shown in FIG. 2 , there may also be deviations in determining the coordinate positions of molecular clusters that are closer.

Aiming at the problems in the above-mentioned related technologies, no effective solution has been proposed yet.

The following is an explanation of the relevant terms in this field:

Molecular cluster: the English name is Cluster, which refers to a collection of specific molecules in the molecular sequencing process, which contains molecules with the same sequence, and the average distance between these molecules is smaller than the average distance between molecules of different molecular clusters.

Sequencing: The purpose of sequencing is to identify the sequence of molecules within a molecular cluster. The sequence of the molecule refers to the type of molecular building block at a particular position in the molecule. Taking DNA molecular sequencing as an example, the sequence is the type of each base in a specific segment of the DNA molecule.

Fluorescence signal: the English name is fluorescence intensity, which refers to the light intensity obtained by a predetermined measurement method, and the molecular fluorescent markers in the molecular cluster are excited and emitted, also known as fluorescence intensity.

Signal mixing: no English name, refers to the fluorescent signals of any molecular clusters that appear in the fluorescent signals derived from the fluorescent labels of other molecular clusters.

Channel: The English name is channel. When measuring the fluorescent label of a molecular cluster in a certain state, each measurement method is called a channel.

Sequencing wheel: The English name is cycle. When measuring molecular fluorescent markers in different measurement methods, the measurement process of a state is a sequencing wheel.

Spectral cross-color, the English name is laser-crosstalk or spectrum-crosstalk, refers to the phenomenon that the fluorescent label corresponding to a certain type of group causes the fluorescent signal to be non-zero in more than one channel.

Phase dephasing, the English name is phasing, refers to the phenomenon that the fluorescent label corresponding to the group at a specific position causes the fluorescent signal to be non-zero in more than one sequencing round.

Molecular cluster positioning, the English name is template generation, which refers to determining which coordinates in the image have molecular clusters that meet predetermined conditions.

Contents of the invention

Aiming at the problems existing in related technologies, especially that the signals of molecular clusters that are closer together will mix together and affect each other, the present invention proposes a processing method for sequencing data of multiple molecules.

The contents of this method include:

(1) Calculating the mixing coefficient C (A←B) between any molecular cluster A and the fluorescent signal of the molecular cluster B meeting the predetermined conditions, which is used to measure the mixing of the molecular cluster B in the fluorescent signal of the molecular cluster A severity.

(2) Using the calculated confounding coefficients, the fluorescence signals of the molecular clusters are processed.

The significance of the present invention lies in that the data processing method proposed by the present invention effectively measures the severity of interference or mixing between the fluorescent signals of different molecular clusters by calculating the mixing coefficients between the fluorescent signals of different molecular clusters. Furthermore, it is possible to process the signals of molecular clusters that are closer to each other during molecular sequencing, and use the processing results for molecular sequence identification and output related information for sequence identification, so as to greatly improve the identification accuracy of molecular identification technology. The prior art uses the method of image deblurring to reduce the mixing of fluorescent signals of molecular clusters, but the degree of mixing of some fluorescent signals does not conform to the kernel function mode used in the fuzzing method, resulting in a certain degree of mixing still remaining in the fluorescent signals of molecular clusters, affecting Accuracy of sequence identification. The invention effectively makes up for this deficiency in the prior art.

The technical route of the data processing method that the present invention proposes comprises:

(1) Calculate the mixing coefficient C(A←B) between any molecular cluster A and the fluorescent signal of the molecular cluster B that meets the predetermined conditions, and the C(A←B) is used to measure the source of the fluorescent signal of the molecular cluster A The severity of the confounding of the molecular cluster B is the ratio of E(A←B) to E(B←B), wherein the E(A←B) is the fluorescence signal of the molecular cluster A that belongs to The fluorescent signal of the molecular fluorescent label in the molecular cluster B, the E(B←B) is the fluorescent signal of the molecular fluorescent label in the molecular cluster B among the fluorescent signals of the molecular cluster B. The C(A←B) is calculated by the following formula:

C(A←B)＝argmin _c (f(I _A -cI _B )+h(c));

Among them, h(c) is a pre-set monotone non-decreasing function, which is used to control the influence of excessive confounding coefficients on sequence recognition accuracy, and I _A and I _B are molecular cluster A and molecular cluster B in the pre-specified sequencing round and the fluorescent signal of the sequencing channel, Used to measure the severity of confounding in the input fluorescent signal. where n is the number of sequencing rounds, for the number of sequencing rounds j, r _j is a preset function, w _j is a scalar calculated from the fluorescence signals of all molecular clusters in the j-round sequencing or a preset constant . High aliasing in the input signal makes the value of f(I) larger, so the calculated aliasing coefficient makes the fluorescence signal of molecular cluster A corrected for signal aliasing, and the degree of aliasing decreases.

When calculating argmin _c (f(I _A -cI _B )+h(c)), it is completed by using the quantile method to find the zero point of the derivative function of f(I _A -cI _B )+h(c).

(2) Process the fluorescent signal of the molecular cluster according to the mixing coefficient, so as to complete the identification of the sequence of the molecules in the molecular cluster and the calculation of information related to the sequence identification.

Wherein, processing the fluorescent signal of the molecular cluster includes correcting signal mixing in the fluorescent signal of the molecular cluster, and the correction method includes:

The matrix _II formed by the fluorescent signals of the molecular clusters without signal confounding is calculated by the following formula:

C · I _I = I _O ;

Wherein in the matrix _II , the elements of each row correspond to the fluorescence signals of a molecular cluster, and the elements of each column correspond to the fluorescence signals of all molecular clusters of one channel in a sequencing round; The matrix formed by the confounding coefficient; The I _O is the matrix formed by the molecular cluster fluorescence signals that need to be corrected. In the matrix I _O , the elements in each row correspond to the fluorescence signals of a molecular cluster, and the elements in each column Fluorescent signal corresponding to all molecular clusters in one channel in one sequencing round.

Processing the fluorescent signal of the molecular cluster further includes performing subsequent processing on the fluorescent signal of the molecular cluster corrected for signal confusion, so as to complete the recognition of the molecular sequence.

(3) In order to calculate the mixing coefficient between molecular clusters more easily, the method uses a predetermined method to process the input data before calculating the mixing coefficient between the fluorescent signals of different molecular clusters, and the predetermined method includes at least one of the following:

Correcting spectral cross-color, correcting phase loss, and preprocessing raw image data to generate the molecular cluster fluorescence signal.

When preprocessing the raw image data to generate the fluorescent signal of the molecular cluster, the method includes the following steps:

Background light removal, normalization, generation of alignment templates, cluster localization and extraction of cluster fluorescence signals.

Wherein, the step of generating an alignment template includes:

aligning images of channels in which spectral bleed-through is present, and correcting spectral bleed-through of the aligned images;

Comparing the luminances of pixels at the same position in each of the images that have undergone spectral color cross-correction, and retaining the value with the highest luminance in the same position to generate an alignment template.

In the step of generating an alignment template, the method for aligning different images (or images with the alignment template) includes:

Select a predetermined coordinate range and a predetermined number of areas in the two images to be aligned, and perform a displacement operation on the selected area of one of the images;

For the area of the predetermined coordinate range of the two images, search for the displacement of the whole-point coordinates of the area in one of the images, and use the displacement coordinate corresponding to the maximum correlation between the area and the area in the other image as the initial point , locate displacements by BFGS or other algorithms for solving unconstrained optimization problems.

The molecular cluster localization step comprises:

A positioning operation is performed on the image corrected for spectral color crosstalk, and the positioning operation includes:

Find the bright spot in the image that has been corrected for spectral cross-color, and fit a parabola in two directions respectively through the fluorescence signals of the target bright spot and multiple bright spots around the target bright spot, and calculate the symmetry axis of the parabola to determine the coordinates of the target bright spot;

The coordinates of the molecular clusters corresponding to each bright spot are calculated by the mean value of the coordinates of bright spots without neighbors, wherein the bright spots without neighbors are bright spots within a unit pixel containing bright spots, and are around the unit pixel containing bright spots There are no bright spots of the same channel and the same sequencing round in the range of two unit pixels except the bright spots contained in itself.

According to another aspect of the present invention, a data processing device is provided.

The unit includes:

Calculation Confusion Coefficient module, used to calculate the confounding coefficient between the fluorescent signals of different molecular clusters. Wherein, the mixing coefficient C (A←B) between any molecular cluster A and the fluorescent signal of the molecular cluster B meeting the predetermined conditions is used to measure the severity of the mixing of the molecular cluster B to the fluorescent signal of the molecular cluster A degree.

The device may also include a processing module, which is used to process the fluorescent signal of the molecular cluster through the mixing coefficient, so as to complete the recognition of the molecular sequence.

The device may also include a preprocessing module, which is used to process the input data in a predetermined manner before calculating the mixing coefficient between the fluorescent signals of different molecular clusters.

Wherein, the module of calculating the mixing coefficient is further used to calculate the following mixing coefficient: for any molecular cluster A and the molecular cluster B meeting the predetermined conditions, the mixing coefficient C (A←B) is E(A←B) and E(B ←B), wherein, the E(A←B) is the fluorescent signal of the fluorescent label derived from the molecular cluster B in the fluorescent signal of the molecular cluster A, and the E(B←B) is the molecular The fluorescent signal of cluster B is derived from the fluorescent signal of the molecular fluorescent label in the molecular cluster B.

Calculating the mixing coefficient module is further used to calculate the C(A←B) by the following formula:

C(A←B)＝argmin _c (f(I _A -cI _B )+h(c));

Among them, h(c) is a pre-set monotone non-decreasing function, I _A and I _B are the fluorescent signals of molecular cluster A and molecular cluster B in the pre-specified sequencing round and sequencing channel, where n is the number of sequencing rounds, for the number of sequencing rounds j, r _j is a preset function, w _j is a scalar calculated from the fluorescence signals of all molecular clusters in the j-round sequencing or a preset constant , where j≥1.

The processing module may further include a correction unit, which is used to correct the signal mixing in the fluorescence signal of the molecular cluster, and the correction method includes:

The matrix composed of the fluorescence signals without signal confounding of the molecular clusters is calculated by the following formula:

C · I _I = I _O ;

Wherein in the matrix _II , the elements of each row correspond to the fluorescence signals of a molecular cluster, and the elements of each column correspond to the fluorescence signals of all molecular clusters of one channel in a sequencing round; The matrix formed by the confounding coefficient; The I _O is the matrix formed by the molecular cluster fluorescence signals that need to be corrected. In the matrix I _O , the elements in each row correspond to the fluorescence signals of a molecular cluster, and the elements in each column Fluorescent signal corresponding to all molecular clusters in one channel in one sequencing round.

The processing module may further include a downstream processing unit for performing subsequent processing on the fluorescent signal of the molecular cluster corrected for signal confusion, so as to complete the recognition of the molecular sequence.

Wherein, the preprocessing module includes an image processing unit and a preprocessing unit. The image processing unit is used to process the image to generate molecular cluster fluorescence signals when the input data is an image obtained by sequencing. The preprocessing unit is used to analyze the molecular cluster fluorescent signals. Processing is performed to meet the conditions required to calculate the confounding coefficients.

Wherein, the image processing unit is further used to use the method of the present invention to perform the following operations on the images obtained by sequencing to generate molecular cluster fluorescence signals: remove background light, normalize, generate alignment templates, molecular cluster positioning and extract molecular cluster fluorescence Signal.

Wherein, the image processing unit includes a correction subunit and a positioning subunit:

The proofreading unit is used to correct the spectral cross-color of the image corresponding to the channel with spectral cross-color;

The locating subunit is used to perform a molecular cluster locating operation on the image corrected for spectral cross-color.

Description of drawings

In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following will briefly introduce the drawings that are used in the embodiments. Apparently, the drawings in the following description only conform to some embodiments of the present invention, and those skilled in the art can also obtain the corresponding drawings of other embodiments according to these drawings without creative work. .

Fig. 1A is a partial schematic diagram of a picture of one channel of one round of sequencing measured by the prior art;

Figure 1B is a schematic diagram of partial sequencing rounds of two close molecular cluster signals after correction of spectral cross-color and phase dephasing. In this figure, the second molecular cluster generates adjacent molecular clusters for the first molecular cluster signal Mixed, and caused the 13th base of the first molecular cluster to be misidentified;

Figure 1C is a schematic diagram of signal mixing between three molecular clusters;

Figure 2 is a schematic diagram of the impact of closer molecular clusters on molecular cluster positioning;

3 is a schematic flow diagram of a data processing method according to an embodiment of the present invention;

FIG. 4 is a schematic diagram of a step flow of a data processing method according to an embodiment of the present invention;

Fig. 5 is a schematic diagram of data processing results according to an embodiment of the present invention;

Fig. 6 is a schematic structural diagram of a data processing device according to an embodiment of the present invention.

Detailed ways

The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only some, not all, embodiments of the present invention. All other embodiments obtained by persons of ordinary skill in the art based on the embodiments of the present invention belong to the protection scope of the present invention.

In the process of realizing the present invention, the inventors found that in the existing technical solutions of molecular sequencing, the fluorescent signals of the molecular clusters provided by the sequencing instrument are partly carried out (this data is stored in a file with an extension of CIF or uncompressed TXT file). The file in this format mainly includes the fluorescent signal on each channel of each molecular cluster and each round of sequencing. Since the data provided by the sequencing instrument has discarded the molecular clusters that are too close to be mixed seriously, the current method does not have a good way to deal with the mixed signals, but uses a robust method to try to reduce the number of small clusters. The impact of partial confounding.

According to an embodiment of the present invention, a data processing method is provided, which is mainly applied to molecular sequencing. The method calculates the confounding coefficient between any molecular cluster and another molecular cluster meeting predetermined conditions, and applies the calculated confounding coefficient to the identification of molecular sequences, so as to overcome the influence of signal confounding on the accuracy of sequence recognition.

As shown in Figure 3, the data processing method according to the embodiment of the present invention includes:

Step S2, calculating the mixing coefficient between the fluorescence signals of any molecular cluster A and the molecular cluster B meeting the predetermined conditions, the mixing coefficient C (A←B) of any molecular cluster A and molecular cluster B meeting the predetermined conditions is used to measure the molecular Severity of confounding in cluster A derived from molecular cluster B. The inventors found that, for any molecular cluster A and molecular cluster B with scrambling in the fluorescence signal of A, in any sequencing round and channel, the ratio of the scrambling of molecular cluster B in A to the fluorescence signal of molecular cluster B itself Approximately unchanged, therefore, the inventors used this ratio as the value of the confounding coefficient C(A←B) in the examples. The inventors also found that only the molecular clusters that are close to each other can mix with each other. Therefore, only the confounding coefficient between any molecular cluster and other molecular clusters whose distance does not exceed a predetermined pixel value is calculated. At the same time, since the fluorescence signal of molecular clusters without confounding can only have a large value in the channel corresponding to its sequence through preprocessing, and is approximately 0 in the remaining channels, the following formula is used to calculate the confounding coefficient C(A ←B):;

C(A←B)＝argmin _c (f(I _A -cI _B )+h(c));

Among them, h(c) is a pre-set monotone non-decreasing function, I _A and I _B are the fluorescent signals of molecular cluster A and molecular cluster B in the pre-specified sequencing round and sequencing channel, where n is the number of sequencing rounds, for the number of sequencing rounds j, _rj is a preset function for calculating the severity of confounding in sequencing round j, and _wj is the fluorescence in the jth sequencing round according to all molecular clusters The scalar calculated from the signal or a preset constant is the weight of the sequencing round j when calculating the confounding coefficient, and c is a real number within a predetermined interval.

For the fluorescent signal preprocessed by the preprocessing method of the embodiment, r _j can be in the following form:

Among them, r is the number of channels, and I(j,k) is the value of the input fluorescent signal in the jth sequencing round and the kth channel.

When calculating the confounding coefficient by the above formula, argmin _c (f(I _A -cI _B )+h(c)) can be obtained by using the quantile method to obtain the derivative function of f(I _A -cI _B )+h(c) The zero method is obtained.

In step S3, the fluorescence signals of different molecular clusters are processed according to the mixing coefficient.

In an embodiment, the inventors correct signal confounding in molecular cluster fluorescence signals by this confounding coefficient. The correction method used is to calculate the matrix _II composed of the fluorescent signals of the molecular clusters without signal confounding by the following formula:

C · I _I = I _O ;

Wherein in matrix _II , the element of each row corresponds to the fluorescent signal of a molecular cluster, and the element of each column corresponds to the fluorescent signal of all molecular clusters of one channel in a sequencing round; C is determined by the confounding coefficient between each molecular cluster I _O is a matrix formed by the fluorescent signals of molecular clusters that need to be corrected. In the matrix I _O , the elements in each row correspond to the fluorescent signals of a molecular cluster, and the elements in each column correspond to the fluorescence signals of a channel in a sequencing round. Fluorescence signals of all molecular clusters.

After correcting the signal confounding in the fluorescent signal of the molecular cluster by the confounding coefficient, the fluorescent signal of the molecular cluster that has been corrected for signal confounding can also be processed in a predetermined manner to complete sequence identification and calculation of related information.

In addition, before calculating the mixing coefficient between the fluorescent signals of different molecular clusters, according to the method used to calculate the mixing coefficient and the characteristics of the input data, it is necessary to perform corresponding preprocessing operations on the input data, including:

Step S1, before calculating the mixing coefficient between the fluorescent signals of different molecular clusters, the fluorescent signals of the molecular clusters are corrected by a predetermined correction method, and the predetermined correction method includes at least one of the following:

Step S121, correcting spectral cross-color;

Step S122, correcting phase loss;

Step S11, preprocessing the original image data to generate molecular cluster fluorescence signals.

Among them, preprocessing the raw image data to generate molecular cluster fluorescence signals includes:

Step S111, read the original image data and perform normalization, the specific method is as follows:

Calculate the fluorescence intensity scale of each position of the image in different channels according to the sequencing image data of the first k rounds, where k≥1, specifically, find the bright spots in the image, where the bright spots are in the same image, according to the preset Regularly screened pixels, and the fluorescence intensity of the screened pixels exceeds the fluorescence intensity of its surrounding pixels;

Divide the planar region of the image into multiple non-overlapping regions, and calculate the median of the fluorescence intensity of the bright spots contained in each region in the image in the previous k rounds of sequencing in each channel;

removing bright spots in areas that do not meet the predetermined rules within the predetermined range of the target area in the image according to predetermined rules;

Use the median of the bright spots calculated in the remaining area of the image to fit a high-order surface by the method of least squares, and calculate the fluorescence intensity scale in the remaining area of the image according to the high-order surface, wherein the surface of the high-order surface The number of times is proportional to the number of regions in the image.

Divide the light intensity value of each pixel of the image by the fluorescence intensity scale of the corresponding position of the current sequencing channel.

In addition, preprocessing the raw image data to generate molecular cluster fluorescence signals further includes:

Step S112, calculating the background light of the original image data, and removing the background light;

Step S113, generate an alignment template, the specific steps are: first align the channels with spectral cross-color in the predetermined sequencing round, then correct the spectral cross-color of the images of the channels with spectral cross-color, and convert each corrected spectral cross-color image Compare the fluorescence signals of the pixels at the same position in the same position, keep the value with the largest fluorescence signal in the same position, and generate an alignment template. The step of aligning any two images is to select the area with the same coordinates in the two images that need to be aligned and corrected for spectral cross-color, and perform a displacement operation on the selected area of one of the images; search for the entire point of the selected area The displacement of the coordinate, and the displacement coordinate corresponding to the maximum correlation is used as the initial point, and the displacement is positioned by BFGS or other algorithms for solving unconstrained optimization problems.

Step S114, performing a molecular cluster localization operation on the aligned images.

Specifically, the spectral cross-color is corrected on the aligned image, and then the bright spot in the corrected image is searched, and the target bright spot and the fluorescence signal of the pixels around the target bright spot are used to fit a parabola in two directions respectively, And calculate the symmetry axis of the parabola, and use the symmetry axis as the coordinates of the target bright spot;

The coordinates of the molecular clusters corresponding to each bright spot are calculated by the mean value of the coordinates of bright spots without neighbors, where the bright spots without neighbors are bright spots that meet the following conditions: there is no other than itself within the range of two unit pixels around the unit pixel containing the bright spot In addition to the included highlights, other highlights of the same channel and the same sequencing round.

Step S115, extracting fluorescence signals of molecular clusters. The specific method is to calculate the position of each molecular cluster in each image by aligning each image with the alignment template, so as to obtain the fluorescence signal of each molecular cluster.

Wherein, the preprocessing operation step S1 may also include:

Step S123, after correcting the spectral cross-coloring of the molecular cluster fluorescence signal, the adjacent group interference correction is performed on the molecular cluster fluorescence signal, wherein the adjacent group interference is the group category at the previous position of the molecular cluster and its successor group The phenomenon of different interferences generated by the fluorescent signal of the group.

Specifically, after correcting the spectral cross-color, for any group category a and category b, for all the molecular clusters of type a in the Lth sequencing round, calculate the molecular clusters of all the categories in the L+1th sequencing round as type b The average or median of the molecular fluorescence intensities on the channel corresponding to the molecular cluster is used to obtain the fluorescence of type b when the fluorescent label of type a in the Lth sequencing round interferes with the fluorescent signal of type b in the L+1 sequencing round. Average scale of markers, where L ≥ 1;

For any sequencing round M, where M≥2, according to the sequence category identified in the M-1 round, the fluorescence signal of the molecular cluster on each channel of the M-th round is divided by the category identified by the M-1 round The average scale of the fluorescent markers in the current channel under the interference.

The above-mentioned method of the present invention is applicable to the mixing coefficient with arbitrary characteristics between the fluorescence signals of any two molecular clusters. The above-mentioned method reduces the interference of signal mixing through the mixing coefficient, and improves the accuracy of molecular cluster sequence identification.

According to an embodiment of the present invention, the present invention also provides a data processing device, which can be applied in the field of molecular recognition, and is used to more accurately identify molecular sequences using the above-mentioned method of the present invention.

As shown in Figure 6, the device includes:

Calculating the mixing coefficient module D2, used to calculate the mixing coefficient between the fluorescent signals of different molecular clusters. Wherein, the mixing coefficient C (A←B) between any molecular cluster A and the fluorescent signal of the molecular cluster B meeting the predetermined conditions is used to measure the severity of the mixing of the molecular cluster B to the fluorescent signal of the molecular cluster A.

The processing module D3 is used to process the fluorescent signal of the molecular cluster through the mixing coefficient, so as to complete the recognition of the molecular sequence.

The device may also include a preprocessing module D1, configured to process the input data in a predetermined manner before calculating the mixing coefficient between the fluorescent signals of different molecular clusters.

Wherein, the module D2 for calculating the mixing coefficient is further used to calculate the following mixing coefficient: for any molecular cluster A and the molecular cluster B meeting the predetermined conditions, the mixing coefficient C (A←B) is E(A←B) and E(B← The ratio of B), wherein, E(A←B) is the fluorescent signal of the fluorescent label derived from molecular cluster B in the fluorescent signal of molecular cluster A, and E(B←B) is the fluorescent signal of molecular cluster B derived from the molecular Fluorescence signal of fluorescently labeled molecules in cluster B.

Calculating the mixing coefficient module D2 is further used to calculate C(A←B) by the following formula:

C(A←B)＝argmin _c (f(I _A -cI _B )+h(c));

Among them, h(c) is a pre-set monotone non-decreasing function, I _A and I _B are the fluorescent signals of molecular cluster A and molecular cluster B in the pre-specified sequencing round and sequencing channel, where n is the number of sequencing rounds, for the number of sequencing rounds j, r _j is a preset function, w _j is a scalar calculated from the fluorescence signals of all molecular clusters in the j-round sequencing or a preset constant , where j≥1, c is a real number within a predetermined interval.

The processing module D3 may further include a correction unit D31, which is used to correct the signal mixing in the fluorescence signal of the molecular cluster, and the correction method includes:

The matrix I _I composed of the fluorescent signals of different molecular clusters corrected for signal mixing is calculated by the following formula:

C · I _I = I _O ;

Wherein in matrix _II , the element of each row corresponds to the fluorescent signal of a molecular cluster, and the element of each column corresponds to the fluorescent signal of all molecular clusters of one channel in a sequencing round; C is determined by the confounding coefficient between each molecular cluster A matrix composed of; I _O needs to be corrected The matrix formed by the fluorescent signals of the molecular clusters, in the matrix I _O , the elements of each row correspond to the fluorescence signals of a molecular cluster, and the elements of each column correspond to all the elements of a channel in a sequencing round Fluorescence signal of molecular clusters.

The processing module D3 may further include a downstream processing unit D32, configured to perform subsequent processing on the fluorescent signal of the molecular cluster corrected for signal confusion, and then complete the identification of the molecular sequence.

Among them, the preprocessing module D1 includes an image processing unit D11 and a preprocessing unit D12. The image processing unit is used to process the image to generate molecular cluster fluorescence signals when the input data is an image obtained by sequencing. Cluster fluorescence signals are processed to meet the conditions required to calculate confounding coefficients.

Wherein, the image processing unit D11 is further used to use the method of the present invention to perform the following operations on the images obtained by sequencing to generate molecular cluster fluorescence signals: remove background light, normalize, generate alignment templates, locate molecular clusters and extract molecular clusters fluorescent signal.

Among them, the image processing unit D11 includes a correction subunit DC and a positioning subunit D114:

The proofreading unit DC is used to correct the spectral cross-color of the image corresponding to the channel with spectral cross-color;

The locating subunit D114 is used to perform molecular cluster locating operations on the image that has been corrected for spectral cross-color.

Different modules of the device can be realized by different hardware or software and combinations thereof. The device can be configured with multiple subunits with the same function, and the data processing speed can be accelerated by assigning tasks to these subunits for simultaneous processing. For example, the part of calculating each confounding coefficient in module D2 can be parallelized through OPENMP, or the part of calculating each confounding coefficient can be realized on GPU, FPGA or DSP so that multiple requests for calculating confounding coefficients can be processed at the same time, or through simultaneous Configure multiple instances of this device to speed up data processing.

In order to better understand the composition of the scheme of the present invention, a specific example will be described below. The example applies the present invention to the sequencing of DNA molecules, and improves the sequencing accuracy by processing the input data. It should be noted that the headlines of the following embodiments only express the contents described in the headlines, but the implementation sequence of the technical solutions of the present invention is not limited. Similarly, the steps in the examples only represent a feasible realization of the technical solution of the present invention, and the realization of having no substantial positive impact on the sequencing results by adjusting the order of the steps does not exceed the scope of the technical solution of the present invention.

Fig. 4 shows a schematic flowchart of a data processing method according to an embodiment of the present invention.

1. Data preprocessing and determination of the position of each molecular cluster:

There are differences in the changes of the average signal peak value between different channels in different areas. If it is not processed, the spectral color cross-over matrix in different areas will be inconsistent, so when the estimated spectral cross-color matrix is used to correct the cross-color, the deviation will appear and thus affect the outcome. However, because the signal peak value is affected by factors such as the number of molecules in the molecular cluster, the estimated average signal intensity variance of the sub-region is relatively large, so the present invention uses the sequencing data of the first four rounds to estimate it, and uses the method of polynomial fitting to estimate The value is smoothed.

The flow of this step is as follows:

Step S111, first read in the image data, and then use the data of the first four rounds to estimate the light intensity scale of each position of the image in different channels.

The estimation steps are as follows:

S1111. Find out the bright spots in each image. A pixel is considered a bright spot: if its light intensity value is greater than that of the surrounding 8 pixels and the light intensity value exceeds the mean value of the image light intensity plus a quarter of the standard deviation.

S1112. Cut the entire area into small squares, and in each channel, for each small square, calculate the median of the light intensity of the bright spots falling within the square in the first four rounds of data. Think of the median as an estimate of the scale of that small square.

S1113. Remove those estimated values that deviate too far from the estimated values of the light intensity scale of the surrounding squares. An estimate is considered to be too far away: if its value differs from the mean of at most 8 surrounding neighbors by more than the difference between the largest and smallest value in the neighbors.

S1114. In each channel, for the remaining estimated values, use least squares to fit a high-order surface, and use the value of the surface at each pixel as an estimate of the light intensity scale. The degree of the surface depends on the number of squares in each figure.

The background light of the incoming data is then estimated and subtracted, and each pixel is divided by the light intensity scale for the corresponding channel.

Step S112, the method for estimating background light is as follows:

S1121. Divide each picture into small squares. Use the kth smallest point of all light intensity values in the small square as the estimation of the background light of the small square.

S1122. Remove those estimated values that deviate too far from the estimated values of the surrounding square background light. The definition of "too far away" is the same as that in intensity scale estimation.

S1123. Use the mean value of the estimated background light of surrounding neighbors to replace the removed estimated value.

S1124. Use bilinear interpolation to calculate the background light of each pixel.

Next, generate an alignment template and align the first five rounds of images:

The basis for aligning the pictures is that the luminescent places in different pictures of the chip are the positions of the molecular clusters. Therefore, there is a correlation between the aligned photos, so that the displacement of the photos can be found by using the method of seeking the maximum correlation. However, because the positions of the A and C channel photos of the same round are illuminated, the G and T channels will not be illuminated, so the two cannot be directly aligned. At the same time, because the position of the same molecular cluster in the photos between different channels does not necessarily emit light at the same time, the correlation is weak. In order to achieve high-precision alignment, it is necessary to try to strengthen this correlation. quasi-precision.

During the alignment process, when non-integer pixels are involved, the light intensity value is estimated by segmented cubic interpolation successively in the x-axis and y-axis directions. Step S113, the method of generating an alignment template and aligning the first five rounds of pictures is as follows:

S1131. Through step S11R, align the pictures of channel C and channel A in each round. Estimation of spectral crosstalk between A and C channels. Correct the cross-color of the alignment picture, and then generate the AC channel template corresponding to the sequencing round by taking the maximum value of every two pictures A and C by pixel, that is, compare the light intensity at the same position of each picture, and keep the one with the largest value , resulting in an alignment template.

S1132. Align the template of the second round with the template of the first round. Align the fourth round template with the third round. The maximum value of each pixel of the aligned first-round and second-round templates is used to generate template one, and the templates of the third and fourth rounds are used to generate template two. Align template two with template one.

S1133. Align the G and T channel pictures of the first two rounds with template 2, and align the remaining pictures with template 1.

Step S11R, the algorithm for aligning the two pictures is as follows:

S11R1. Align the small piece in the middle of the two pictures. The criterion for alignment is that the correlation value between the two graphs is the largest at this time. First search the displacement of the whole grid point, and then use the displacement corresponding to the maximum correlation as the initial point to search for a more accurate displacement with the BFGS method.

S11R2. Taking the displacement of the small square in the middle of the two pictures as the initial point, search for the displacement between the small squares located near the four corners of the two pictures by the method of maximizing correlation respectively.

S11R3. Treat the coordinate difference between the two images as an affine transformation, and use Robust regression to calculate the transformation in the x-axis direction and the y-axis direction respectively to calculate the affine transformation between the two images.

Finally, the position of each molecular cluster is identified, and the light intensity scale corresponding to each molecular cluster in each channel is calculated.

Step S114, the steps of identifying molecular clusters are as follows:

S1141. Estimating spectral cross-color through step SC. And correct spectral cross-color. The correction method is to treat the light intensity values of the four channels of each pixel as a four-dimensional vector, and then multiply by left to estimate the inverse of the cross-color matrix.

S1142. Find the bright spots in each picture. Use the center of the bright spot and its five light intensity values up, down, left, and right to determine more accurate coordinates of the bright spot by fitting a parabola in two directions and calculating the symmetry axis of the parabola. A pixel is determined as a bright spot: if its light intensity value is greater than the light intensity values of the surrounding 8 adjacent pixels and its light intensity value exceeds a certain threshold determined based on the entire picture.

S1143. Treat each pixel as a grid, and put the found bright spots into these grids. If two adjacent grids meet: there is at most one channel with a bright spot in each round, the two grids will be merged. Merging refers to moving the bright spots contained in the grid with a low total light intensity value to another grid.

S1144. Delete the grid whose sum of light intensity values of all bright spots in the surrounding grids is too low. Grids with excessively large light intensity values and no significant change in light intensity during the five rounds of sequencing were deleted. Delete cells that contain bright spots with low mean intensity compared to neighboring cells.

S1145. Treat all remaining lattices containing light spots as molecular clusters. The coordinate mean of the included spots located in different channels from the adjacent grid is used as the coordinates of the molecular cluster.

In step SC, the method for estimating spectral crosstalk between m channels is as follows:

SC1. Normalize each channel so that the variance across the different channels is the same. Treat the input as a population of m-dimensional vectors.

SC2. Using the unit vectors on the four channels as initial points, perform k-means clustering with k=m on all input vectors. The distance used in clustering is defined as d(x,y)=1-cos<x,y>

SC3. Calculate the median of each class on each channel, so as to obtain the estimation of each class vector. These vectors are used to form the cross-color matrix of the normalized data.

SC4. Calculate the cross-color matrix before normalization according to the normalization information.

2. Step S115, extracting molecular cluster fluorescence signals

The flow of this step is as follows:

For each image that is read in, first pass through S112 to remove its background light, and then use S11R to calculate the transformation needed to align it with the template. Then the coordinates of each molecular cluster on this map are calculated according to the affine transformation. The light intensity of each molecular cluster is calculated using an interpolation algorithm, and then this light intensity is divided by the average size of the molecular cluster corresponding to the corresponding channel. The above content of related algorithms has been introduced or can be directly implemented according to the description, and will not be repeated here.

3. Step S12, preprocessing of molecular cluster fluorescence signals

Each molecular cluster in the CIF file contains a series of discrete numbers, a total of n rows and 4 columns, and each number represents the light intensity on one channel of one sequencing round. When dealing with spectral bleed-through and phase dephasing, the following probabilistic model for the i-th molecular cluster is widely accepted:

I _i ＝λ _i PS _i M ^T +N+ε _i

Here I _i represents the light intensity value recorded in the CIF file, S _i represents the base sequence of the molecular cluster, it is the same as I _i , it is a matrix of n rows and 4 columns, each row has only one element as 1, and the remaining three elements The positions where both are 0 and 1 correspond to the base type of the molecular cluster in the sequencing round indicated by the row. P is an n×n phase matrix, where the element in the jth row and the lth column represents the probability that the base at the lth position emits light in the jth round of sequencing. And M is a 4×4 spectral cross-color matrix, and the elements in the jth row and the lth column represent the fluorescence intensity of the lth base in the jth channel. ε _i is a white noise matrix with n rows and 4 columns, which represents the measurement error.

The flow of this step is as follows:

Step S121, estimating and correcting spectral cross-color, the specific steps are:

Step S1211, use the SC to estimate the cross-color matrix, and step S1212, correct the spectral cross-color.

Step S122, fully calculating and correcting phase loss. The specific steps are:

Step S1221, estimating the phase matrix. Using this phase matrix as an initial value, a more accurate 4m × 4m matrix including phase and spectral cross-color phenomena is estimated by iterative weighted least squares algorithm. Here m refers to the number of sequencing rounds.

Step S1222, using the new matrix to calibrate the fluorescence signal.

Step S123, correcting the adjacent base interference phenomenon, the steps for correcting this phenomenon are as follows:

Step S1231. Determine the base type of each molecular cluster according to the maximum light intensity value of each sequencing round.

Step S1232. Using the data of the first four rounds, calculate the median value of the light intensity of each base in the current round on the corresponding channel when the current round is a certain base.

Step S1233. For each round of data of each molecular cluster, according to the identified base type in the previous round, divide the data of each channel in the current round by the corresponding median light intensity. Then complete the current round of identification again.

Wherein, step S12 can be replaced by:

In step S12R, other existing methods are used to correct other problems in the fluorescent signal of the molecular cluster except signal mixing.

4. Step S2, correcting signal mixing between molecular clusters

This step depends on the model:

Wherein M is a spectral cross-color matrix, P is a phase matrix, and both are defined in step S12; C is a signal mixing matrix, and the length of its two dimensions is equal to the number of molecular clusters; ξ is a three-dimensional array composed of observation errors , S is a three-dimensional state array composed of sequences that are either 0 or 1, representing the sequence of all molecular clusters, O is a three-dimensional array composed of extracted light intensities, and the lengths of the three dimensions of the above three three-dimensional arrays are respectively the number of molecular clusters , the number of sequencing rounds and the number of channels. The specific meanings of M and P will not be described in detail. The element in the i-th row and l-column of C represents the luminescence of the fluorescent label of the l-th molecular cluster in the CIF data of the i-th molecular cluster, which is recorded as the confounding coefficient C(i ←l), or c _il . Fix the subscripts of the other two dimensions except the r-th dimension in H, and multiply the matrix A by the vector obtained by traversing the r-th dimension to obtain the vector at the corresponding position in the new array. The properties that this operation satisfies include the consistency of the same dimension operation Exchangeability when operating in different dimensions Reversibility (for reversible A, )Wait. And by using the exchangeability of this operation (that is, which dimension is calculated first and then which dimension is calculated), the result can be obtained:

in I is the data corrected for spectral bleed and phase issues. Therefore, other phenomena can be corrected before estimating that the molecular clusters are mixed with each other. and by solving or calculate directly Correction for confounding is done.

When estimating the signal confounding matrix, the confounding coefficient between two molecular clusters can be determined by establishing an objective function to measure the signal quality of molecular clusters, and then optimizing this function, so as to estimate the confounding matrix and solve the model equation to remove confounding. Specifically, it is first set that the elements on the diagonal of the mixing matrix are all 1, and there is no mutual mixing between molecular clusters that are far away (the value is 0). For close-range molecular clusters, taking molecular cluster 1 and molecular cluster 2 as examples, the following two molecular cluster models are used:

Obtained by deformation:

I ₁ ＝c ₁₂ I ₂ +(1-c ₁₂ c ₂₁ )S ₁ +(ξ ₁ -c ₁₂ ξ ₂ )

Here ξ ₁ -c ₁₂ ξ ₂ is expected to be 0, and S ₁ is 0 in the channel except the base category corresponding to the first molecular cluster. Therefore, the base category of each position of the first molecular cluster can be found, and then the corresponding channel can be removed, and the estimation of c ₁₂ can be completed in the remaining channels. This estimation can be realized by establishing an objective function and calculating its extreme value. When correcting the mutual mixing of molecular cluster signals, a larger mixing coefficient will bring additional precision loss to the light intensity data of the four channels. Therefore, it is necessary to introduce a penalty for a large mixing coefficient in the objective function. And notice that on the channel other than the channel corresponding to each base of molecular cluster 1, exist When its expected value is 0, so you can choose an objective function of the form g(1,2)(t)=f(I ₁ -tI ₂ )+h(t), where h(t) is a monotonically increasing function And the function f can be written as follows: w _j is a measure of the accuracy of the j-th round of sequencing, and the function r _j measures the severity of the j-th round of signal being confounded.

Estimation of the proportion of confounding is done by using the weighted LAD method, assuming The channel of the largest signal in each round of sequencing corresponds to the base category of the position of molecular cluster 1, and let h(t) take a linear function, then the objective function can be obtained:

g(1,2)(t)=f(I ₁ -tI ₂ )+ut

Among them, u is a normal number calculated according to the weight or the observation error of the fluorescence signal of the molecular cluster, and the function f is defined as follows:

It represents a measure of the purity of the input signal. By optimizing the objective function, the estimation algorithm of each confounding coefficient can be obtained.

The method of step S2 is as follows:

After completing the initial corrections for problems other than intermixing, proceed as follows. Assume that each image to be processed contains n molecular clusters.

Step S21, perform preprocessing work, and calculate the parameters required for calculating the confounding coefficient, the steps are as follows:

S211. For each molecular cluster, take out the other three signal values that are not the maximum signal in each sequencing round, calculate the median value of these signals, and then estimate the variance through the median value.

S212. For each round of sequencing j, calculate C is any normal number whose value does not affect the calculation result; It is the variance at the jth round of sequencing estimated in the previous step.

S213. For the parameter ink (given in advance, within the range of 0.5 to 0.8, the higher the value, the sequencing accuracy will be slightly improved but the sequence repetition rate will be increased, and the lower the reverse), calculate thr=(-0.75+1.5ink) ∑ w _j .

Step S214, creating an empty sparse matrix S. Assign the molecular cluster number to the element corresponding to the molecular cluster position in an array with the same size as the picture. For each molecular cluster, find all molecular clusters with a distance of no more than a certain pixel through the array, and then estimate the mixing of these molecular clusters to it.

Step S22, for any molecular cluster i and the molecular cluster j whose distance to it is smaller than a predetermined constant, estimate the confounding coefficient C(i←j), ie c _ij . The estimation method is as follows:

S221. If i= _j , assign cij as 1; otherwise, proceed to the following steps.

S222. Definition here I _i and I _j are the light intensities of molecular cluster i and molecular cluster j corrected for other confounding, respectively. Set variable l to 0, r to 1, and then proceed to the next step.

S223. Calculate g(0.6l+0.4r), if its value is greater than thr, then change the value of l to 0.6l+0.4r, otherwise change the value of r to 0.6l+0.4r, and then if |l-r|＞ 0.001, repeat this step, otherwise proceed to the following steps.

S224. Assign l to c _ij .

Wherein, the estimation of different confounding coefficients in step S2 can be completed in parallel. This parallelism can be achieved through GPU programming, multi-core CPUs or FPGAs.

Four, step S3, carry out follow-up processing

This step includes:

Step S31, after the completion of step S2 to obtain the estimation of C, for the input molecular cluster fluorescence signal without any processing, or the molecular cluster fluorescence signal O obtained through step S115, solve CD=O to obtain the light intensity corrected for mutual mixing D.

Step S32, repeating step S12 for the light intensity data that has been corrected for the intermixing of molecular cluster signals, so as to perform spectral cross-color and phase mismatch correction operations.

Step S33, for each round of data of each molecular cluster, determine the base type of the corresponding position according to the channel where the maximum light intensity value is located. Determine the mass value of the molecular cluster signal according to its purity. Output base class and quality value.

Wherein, steps S2 and S31 can be completed in the following manner:

In step S2P, the plane area where the molecular cluster coordinates are located is divided by a predetermined method, and for each sub-area, all the molecular clusters contained in the sub-area and all the molecular clusters whose distance from the sub-area does not exceed a predetermined value are selected, and the selected Steps S2 and S31 are executed for the molecular clusters, and then the calculation results of the molecular clusters included in the sub-region are used as the corrected signal mixing light intensity. The operations on each sub-area can be completed in parallel, and steps S211 to S213 can be performed on each sub-area separately or before step S2P.

Wherein, step S31 and step S32 can be replaced by:

Step S3R1, solving CS=I for the molecular cluster fluorescence signal I obtained in step S123 to obtain a signal that can be directly used for base identification.

Step S32 and step S33 can be replaced by:

Step S3R2, output D, use a third-party tool, such as AYB (Massingham&Goldman, 2012) to complete the sequencing.

The inventor used the technical solution of the present invention to perform a simulation test on the fluorescent signal data of molecular cluster sequencing, as shown in Figure 5: Figure 5 is a schematic diagram of the data processing results according to the embodiment of the present invention, where the horizontal axis represents the distance from the nearest molecular cluster The distance, the vertical axis represents the quantity, and the black part (CACC improved PF reads) is the improvement of sequencing accuracy after adopting the embodiment of the present invention. The x-axis coordinate represents the distance from the center of the nearest molecular cluster. The long bar on the left is the ratio of perfect matching sequences after data processing by the present invention, the middle bar is the result of the solution of the present invention without correcting the mutual mixing of molecular cluster signals, and the right is the total number of identified molecular clusters. It can be seen that the mapping accuracy of the molecular clusters whose distance from the nearest molecular cluster is 1 to 3 pixels increases most significantly.

At the same time, the inventor produced software applying the technical solutions of the present invention. The software can input sequencing image data or molecular cluster fluorescence signal data, complete the correction of signal mixing by calculating the mixing coefficient, and output the molecular cluster fluorescence signal or sequence recognition results and quality values corrected for signal mixing. According to the technical solution of the present invention, the software is divided into a preprocessing module, a mixing coefficient calculation module and a processing module, which are respectively used for preprocessing the input data, calculating the mixing coefficient and performing subsequent processing on the input data according to the mixing data. The preprocessing module is divided into an image processing unit and a preprocessing unit. The image processing unit is used to process the case where the input data is a sequencing image, and the preprocessing unit is used to complete the preprocessing of the data so that it meets the conditions for calculating the confounding coefficient. The specific content of the software is shown in the above steps and will not be repeated here. One version of the software is implemented by compiling C++ code, and the other version of the software is implemented by Matlab program. The steps of each part of the software are processed in parallel through OPENMP, which speeds up the execution speed.

To sum up, with the help of the above-mentioned technical solution of the present invention, through the self-adaptive correction of signal mixing between adjacent molecular clusters, the identification of molecular sequences can be completed more accurately. In addition, the present invention can also read in the original image data or molecular cluster fluorescence signal data, and output the molecular cluster fluorescence signal data corrected for signal mixing, or output the final molecular sequence with quality evaluation. This technology can be directly applied to processing Data generated by bridge amplification technology DNA sequencing instruments can be applied to process data generated by other devices that identify the structure or sequence of multiple molecules.

The above descriptions are only preferred embodiments of the present invention, and are not intended to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present invention shall be included in the scope of the present invention. It is within the scope of protection, and this work has been supported by the National Natural Science Foundation of China Major Research Program Cultivation Project 91130008.

references

Anastasi, C.(2008). Accurate whole human genome sequencing using reversible terminator chemistry. Nature, 456(7218), 53-59.

Bentley, D.R., Balasubramanian, S., Swerdlow, H.P., Smith, G.P., Milton, J., Brown, C.G.,...&

Li, L., & Speed, T.P.(1999). An estimate of the crosstalk matrix in four-dye fluorescence-based DNA sequencing. Electrophoresis, 20(7), 1433-1442.

Massingham, T., & Goldman, N. (2012). All Your Base: a fast and accurate probabilistic approach to base calling. Genome Biol, 13, R13.

Whiteford, N., Skelly, T., Curtis, C., Ritchie, ME, A.,Zaranek,AW,...&Brown,C.(2009).Swift: primary data analysis for the Illumina Solexasequencing platform.Bioinformatics,25(17),2194-2199.

Claims (14)

1. A data processing method for multiple molecular signals, characterized in that it comprises: Calculating the mixing coefficient C (A←B) between any molecular cluster A and the fluorescent signal of the molecular cluster B meeting the predetermined condition, wherein the molecular cluster B meeting the predetermined condition is in the fluorescent signal of the molecular cluster A having promiscuous said molecular cluster B; According to the mixing coefficient, the fluorescent signals of different molecular clusters are processed; Wherein, for any of the molecular cluster A and the molecular cluster B meeting the predetermined conditions, the C(A←B) is used to measure the severity of the confounding in the fluorescent signal of the molecular cluster A originating from the molecular cluster B degree, the mixing refers to the fluorescent signal of the fluorescent label in the molecular cluster B that appears in the fluorescent signal of the molecular cluster A; a molecular cluster is a collection of specific molecules, which contains the specific molecules with the same sequence , and the average distance between these specific molecules is smaller than the average distance between molecules of different molecular clusters; for any molecular cluster A, its fluorescence signal refers to the obtained by a predetermined method, which can be used to detect the molecular cluster A Data containing the sequence or subsequence of a molecule for identification; the sequence of a molecule is the type of molecular basic element at one or more predetermined positions in the molecule; Wherein, the processing of fluorescent signals of different molecular clusters according to the mixing coefficient includes: correcting for signal confounding in the fluorescent signals of the different molecular clusters by the confounding coefficient; Wherein, the signal mixing refers to the occurrence of fluorescent signals belonging to molecular fluorescent labels in other molecular clusters among the fluorescent signals of any molecular cluster; The correction of the signal confounding in the fluorescent signals of the different molecular clusters by the confounding coefficient comprises: The matrix _II composed of the fluorescent signals of the different molecular clusters corrected for signal mixing is calculated by the following formula: C · I _I = I _O ; Wherein in the matrix _II , the elements of each row correspond to the fluorescence signals of a molecular cluster, and the elements of each column correspond to the fluorescence signals of all molecular clusters of one channel in a sequencing round; C is the mixture between each molecular cluster A matrix composed of coefficients; I _O is a matrix composed of the fluorescent signals of the molecular clusters that need to be corrected. In the matrix I _O , the elements in each row correspond to the fluorescent signals of a molecular cluster, and the elements in each column correspond to Fluorescent signal of all molecular clusters in one channel in one sequencing round. 2. according to the described method of claim 1, it is characterized in that, for any described molecular cluster A and described molecular cluster B, described confounding coefficient C (A←B) is E(A←B) and E(B ←B), wherein, the E(A←B) is the fluorescent signal of the fluorescent label derived from the molecular cluster B in the fluorescent signal of the molecular cluster A, and the E(B←B) is the The fluorescent signal of the molecular cluster B is derived from the fluorescent signal of the molecular fluorescent label in the molecular cluster B. 3. according to the described method of claim 1, it is characterized in that, calculate described confounding coefficient C (A←B) by following formula: C(A←B)＝argmin _c (f(I _A -cI _B )+h(c)); Wherein, h(c) is a preset monotone non-decreasing function, I _A and I _B respectively represent the fluorescent signals of the molecular cluster A and the molecular cluster B in the pre-specified sequencing round and sequencing channel, where n is the number of sequencing rounds, for the number of sequencing rounds j, r _j is a preset function, w _j is a scalar calculated from the fluorescence signals of all molecular clusters in the j-round sequencing or a preset constant , c is a real number in a predetermined interval; The function formula of r _j is as follows: Among them, r is the number of channels, and I(j,k) is the value of the input fluorescent signal in the jth sequencing round and the kth channel. 4. according to the described method of claim 3, it is characterized in that, argmin _c (f(IA _{- cIB} )+h(c)) obtains f(IA _{- cIB} )+h( c) The zero point of the derivative function is obtained. 5. The method according to claim 1, further comprising: before calculating the mixing coefficient between the fluorescent signals of the different molecular clusters: The input data is processed in a predetermined manner, and the predetermined manner includes at least one of the following: Correct spectral cross-color, correct phase loss, and preprocess raw image data to generate fluorescence signals of molecular clusters. 6. The method according to claim 5, wherein the raw image data is preprocessed to generate a fluorescent signal of a molecular cluster, comprising: Correct the spectral cross-color of the image corresponding to the channel with spectral cross-color; performing a molecular cluster localization operation on the image corrected for spectral cross-color, Wherein, the molecular cluster positioning operation refers to determining the molecular clusters meeting the predetermined conditions in the image, and determining the coordinates of the molecular clusters meeting the predetermined conditions, and the determining molecular clusters meeting the predetermined conditions are located by target bright spots in each image. corresponding molecular clusters. 7. The method according to claim 1, wherein the processing of fluorescent signals of different molecular clusters according to the mixing coefficient further comprises: The sequences of the molecules in the clusters are identified by the cluster fluorescence signals corrected for signal scrambling. 8. A data processing device for multiple molecular signals, comprising: Calculate the mixing coefficient module, which is used to calculate the mixing coefficient C (A←B) between the fluorescence signal of any molecular cluster A and the molecular cluster B meeting the predetermined conditions, wherein the molecular cluster B meeting the predetermined conditions is in the The fluorescent signal of molecular cluster A has mixed molecular cluster B; A processing module, configured to process the fluorescent signals of different molecular clusters according to the mixing coefficient; Wherein, for any of the molecular cluster A and the molecular cluster B meeting the predetermined conditions, the C(A←B) is used to measure the severity of the confounding in the fluorescent signal of the molecular cluster A originating from the molecular cluster B degree, the mixing refers to the fluorescent signal of the fluorescent label in the molecular cluster B that appears in the fluorescent signal of the molecular cluster A; a molecular cluster is a collection of specific molecules, which contains the specific molecules with the same sequence , and the average distance between these specific molecules is smaller than the average distance between molecules of different molecular clusters; for any molecular cluster A, its fluorescence signal refers to the obtained by a predetermined method, which can be used to detect the molecular cluster A Data containing the sequence or subsequence of a molecule for identification; the sequence of a molecule is the type of molecular basic element at one or more predetermined positions in the molecule; Wherein, the correction unit is used to correct the signal confounding in the fluorescent signals of the different molecular clusters through the confounding coefficient, Wherein, the signals are mixed with the fluorescent signals belonging to molecular fluorescent labels in other molecular clusters appearing in the fluorescent signals of any molecular clusters; The correction unit is further used to calculate the matrix I _I composed of the fluorescent signals of the different molecular clusters that have corrected signal mixing by the following formula: C · I _I = I _O ; Wherein in the matrix _II , the elements of each row correspond to the fluorescence signals of a molecular cluster, and the elements of each column correspond to the fluorescence signals of all molecular clusters of one channel in a sequencing round; A matrix composed of mixing coefficients; the I _O is a matrix formed by the fluorescent signals of the molecular clusters that need to be corrected, in the matrix I _O , the elements in each row correspond to the fluorescent signals of a molecular cluster, and each column The elements correspond to the fluorescent signals of all molecular clusters in one channel in one sequencing round. 9. The device according to claim 8, wherein the confounding coefficient C(A←B) is the ratio of E(A←B) to E(B←B), wherein the E(A←B) B) is the fluorescent signal of the molecular fluorescent marker belonging to the molecular cluster B in the fluorescent signal of the molecular cluster A, and the E(B←B) is the fluorescent signal of the molecular cluster B belonging to the molecular cluster B Fluorescent signal of a mid-molecule fluorescent label. 10. according to the described device of claim 8, it is characterized in that, described calculation confounding coefficient module is further used for, calculates described confounding coefficient C (A←B) by following formula: C(A←B)＝argmin _c (f(I _A -cI _B )+h(c)); Among them, h(c) is a preset monotone non-decreasing function, I _A and I _B are the fluorescent signals of molecular cluster A and molecular cluster B in the pre-specified sequencing round and sequencing channel, where n is the number of sequencing rounds, for the number of sequencing rounds j, r _j is a preset function, w _j is a scalar calculated from the fluorescence signals of all molecular clusters in the j-round sequencing or a preset constant , c is a real number in a predetermined interval; The function formula of r _j is as follows: Among them, r is the number of channels, and I(j,k) is the value of the input fluorescent signal in the jth sequencing round and the kth channel. 11. according to the described device of claim 10, it is characterized in that, argmin _c (f(IA _{- cIB} )+h(c)) finds f(IA _{- cIB} )+h( c) The zero point of the derivative function is obtained. 12. The apparatus of claim 8, further comprising: A preprocessing module, used to process the input data in a predetermined manner before calculating the mixing coefficient between the fluorescent signals of the different molecular clusters, and the predetermined manner includes at least one of the following: Correct spectral cross-color, correct phase loss, and preprocess raw image data to generate fluorescence signals of molecular clusters. 13. The device according to claim 12, wherein the preprocessing module further comprises: The image processing unit is used to preprocess the raw image data to generate the fluorescence signal of the molecular cluster; and the image processing unit further includes: A correcting subunit, configured to correct spectral cross-color of images corresponding to channels with spectral cross-color; a positioning subunit, configured to perform a molecular cluster positioning operation on the image corrected for spectral cross-color, Wherein, the molecular cluster positioning operation refers to determining the molecular clusters meeting the predetermined conditions in the image, and determining the coordinates of the molecular clusters meeting the predetermined conditions, and the determining molecular clusters meeting the predetermined conditions are located by target bright spots in each image. corresponding molecular clusters. 14. The device according to claim 8, wherein the processing module further comprises: The downstream processing unit is used to identify the sequence of the molecules in the molecular cluster according to the fluorescent signal of the molecular cluster corrected by the correction unit for signal confusion.