Mutation detection method and device and application thereof

Abstract

The invention belongs to the field of bioinformatics, particularly relates to a mutation detection method and device and application thereof, and more particularly relates to a low-frequency mutation detection method and device and application thereof. By constructing a new sequence set and combining a single-chain consistency sequence and a double-chain consistency sequence correction method, high mutation detection sensitivity and specificity can be obtained, and meanwhile, due to the fact that more sequences are reserved, the effective amount of data is greatly increased.

Description

Technical Field

[0001] The present invention belongs to the field of bioinformatics, and specifically, relates to a mutation detection method, device and application thereof, and more specifically, relates to a low-frequency mutation detection method, device and application thereof. Background Art

[0002] The detection of tissue samples is widely used in the field of tumor detection, but there are some limitations. Tissue samples cannot fully reflect the heterogeneity of tumors. They are an invasive sampling method and some samples are difficult to obtain. The development of technologies such as liquid biopsy biomarkers ctDNA has provided a new way to overcome the limitations of tumor tissue sample detection. ctDNA can reflect the heterogeneity within the tumor and help understand the complexity and progression of the tumor. Like tissue samples, ctDNA can be used for targeted drug guidance, efficacy monitoring, etc. At the same time, it can detect mutation signals earlier, conduct early intervention and treatment, and dynamically monitor patient recurrence. Its application in small residual lesions has gradually been expanded, and patient efficacy evaluation can be performed earlier than imaging evaluation.

[0003] The ability of plasma ctDNA to detect mutations is related to tumor burden. 3 , 10cm 3 , 100cm 3 For tumors of different ages, the average VAF of plasma is 0.006%, 0.1% and 1.3% respectively. 1ml of plasma generally corresponds to 1500 genome copies. Generally, 4ml of plasma can be obtained from 10ml of blood of patients, with a total of 6000 genome copies. The theoretical detection limit that can be achieved is 0.01%. The abundance of ctDNA in plasma is low, and the demand for low-frequency detection needs to be met.

[0004] In the prior art, dual sequencing based on unique molecule identifier (UMI) can effectively suppress high-throughput sequencing errors and is a method that can detect and quantify extremely low-frequency mutations. There are two methods for sequencing data analysis. One is to combine and aggregate multiple reads extended from the same DNA template by the same tag sequence in the sense strand and the negative strand to form a single-strand consensus sequence (SSCS) of the sense strand and the antisense strand, but for low-frequency mutations, the false positive rate is high; the second is to compare the generated sense strand SSCS with the complementary antisense strand SSCS, further generate a duplex consensus sequence (DCS), and compare the DCS with the reference genome again to identify mutations or sequencing errors, but when the sequencing depth is not high enough, the sensitivity of low-frequency variation detection is low, and the data efficiency of this method is less than 2.5%. In view of this, there is a method in the prior art that combines the above two methods to identify mutations or sequencing errors, but the sensitivity and specificity of mutation detection of this method still need to be improved, and its data efficiency is still low.

[0005] Therefore, there is a need in the art for an analysis method that has high sensitivity and specificity for mutation detection and high data efficiency. Summary of the invention

[0006] In view of this, in a first aspect, the present invention provides a mutation detection method, comprising the following steps:

[0007] Obtain sequencing data;

[0008] The sequencing data is compared with the reference genome, and the read families of the sense strand and the antisense strand are established according to the comparison results and the labels, and the read families are distinguished according to the following criteria based on the threshold of the number of read segments in the read family:

[0009] First data: the threshold of the number of read segments in the read family is greater than or equal to 2;

[0010] Second data: the threshold value of the number of reads in the read family is equal to 1 and can be complementary to the reads from the other strand of the same template molecule; and

[0011] Third data: the threshold of the number of reads in the read family is equal to 1 and cannot be complementary to the reads from the other strand of the same template molecule;

[0012] Wherein, the read family is derived from the same template molecule;

[0013] Filtering the read family in the third data to obtain a final read family;

[0014] Construct a single-strand consensus sequence and a double-strand consensus sequence for the read families in the first data and the second data;

[0015] The obtained single-strand consensus sequence and double-strand consensus sequence and the final readfamily obtained by filtering are used to construct a sequence set;

[0016] The constructed sequence set was used for mutation detection.

[0017] The applicant creatively discovered that by constructing a new sequence set and combining single-stranded consensus sequence and double-stranded consensus sequence correction methods, higher mutation detection sensitivity and specificity can be achieved. At the same time, because more sequences are retained, the effective amount of data is greatly improved.

[0018] In some specific embodiments, the tag is a UMI.

[0019] Specifically, by using UMI sequence tags, molecules with matching UMI sequence tags are derived from the same template molecule.

[0020] Furthermore, the read family in the third data is filtered according to the following criteria:

[0021] The Q30 ratio of the reads is greater than or equal to 95%; and / or

[0022] SNV / InDel mutation sites with VAF greater than 0.1% were retained in the read segments.

[0023] Using the filtering criteria described above, reads of the present invention that are greater than the threshold of the number of retained reads equal to 1 and cannot be complementary to reads from another chain of the same template molecule can also be further accurately used for mutation detection, greatly increasing the sensitivity and specificity of mutation detection.

[0024] Furthermore, the method may also include: performing cleaning processing on the sequencing data, including removing low-quality reads and bases, repeated reads and contaminated adapter sequences to obtain cleaned data.

[0025] Furthermore, when the read family described in the first data constructs a single-stranded consensus sequence and a double-stranded consensus sequence, if there are differences in the bases at the same position of the read segments in the read family, the bases with a base ratio greater than or equal to 0.7 in all read segments are used as consistent bases.

[0026] Furthermore, the constructed single-stranded consensus sequence and double-stranded consensus sequence have a quality value greater than Q30.

[0027] In a second aspect, the present invention provides a mutation detection device, comprising:

[0028] An acquisition module, used for acquiring sequencing data;

[0029] The alignment module is used to align the sequencing data with the reference genome, and to establish read families of the sense strand and the antisense strand respectively according to the alignment results and the labels, and to distinguish them according to the following criteria based on the threshold of the number of read segments in the read family:

[0030] First data: the threshold of the number of read segments in the read family is greater than or equal to 2;

[0031] Second data: the threshold value of the number of reads in the read family is equal to 1 and can be complementary to the reads from the other strand of the same template molecule; and

[0032] Third data: the threshold value of the number of reads in the read family is equal to 1 and cannot be complementary to the reads from another strand of the same template molecule; wherein the read family is derived from the same template molecule;

[0033] A filtering module, used for filtering the read family in the third data and obtaining a final readfamily;

[0034] A construction module, used to construct a single-stranded consensus sequence and a double-stranded consensus sequence from the read families in the first data and the second data;

[0035] An output module is used to construct a sequence set using the obtained single-strand consensus sequence and double-strand consensus sequence and the final read family obtained by the filtering; and the constructed sequence set is used for mutation detection.

[0036] In the filtering module, the filtering is performed according to the following standards:

[0037] The Q30 ratio of the reads is greater than or equal to 95%; and / or

[0038] SNV / InDel mutation sites with VAF greater than 0.1% were retained in the read segments.

[0039] Furthermore, the acquisition module may also include: performing cleaning processing on the sequencing data, including removing low-quality reads and bases, repeated reads and contaminated adapter sequences to obtain cleaned data.

[0040] Furthermore, when the read family described in the first data in the construction module constructs a single-stranded consensus sequence and a double-stranded consensus sequence, if there are differences in the bases at the same position of the read segments in the read family, the bases in all read segments with a base ratio greater than or equal to 0.7 are taken as consistent bases.

[0041] Furthermore, the device also includes a nucleic acid extraction module, and the nucleic acid extraction module is used to extract nucleic acid from the sample.

[0042] In a third aspect, the present invention provides a use of the above-mentioned analysis method or device in preparing a kit or device for mutation detection.

[0043] In a fourth aspect, the present invention provides a device, comprising:

[0044] at least one processor; and

[0045] a memory communicatively connected to at least one of the processors; wherein,

[0046] The memory stores instructions that can be executed by the processor, and the instructions are used to be executed by the processor to implement any of the mutation detection methods described above.

[0047] In some embodiments, the device further includes at least one input device and at least one output device; in the device, the processor, memory, input device, and output device are connected via a bus.

[0048] In a fifth aspect, a storage medium is provided, wherein the storage medium stores computer instructions, and the computer instructions are used to be executed by the computer to implement any of the mutation detection methods described above.

[0049] In some embodiments, the storage medium is a computer-readable storage medium.

[0050] In a sixth aspect, the present invention provides a kit comprising:

[0051] Sample nucleic acid extraction reagents and sequencing reagents; and

[0052] The apparatus, device or storage medium as described above. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Figures 1-2 is a schematic diagram example of the data analysis method of the present invention;

[0054] Figure 3 This is the data efficiency of different data analysis methods of the present invention. DETAILED DESCRIPTION

[0055] The present invention will be described in detail below in conjunction with specific embodiments and examples, and the advantages and various effects of the present invention will be more clearly presented. It should be understood by those skilled in the art that these specific embodiments and examples are used to illustrate the present invention, rather than to limit the present invention.

[0056] Terms involved in this invention:

[0057] Circulating tumor DNA (ctDNA): small fragments of DNA released into the blood by tumor cells, showing non-random fragmentation, about 160 nt. It carries the genetic information of the tumor and can be used for tumor diagnosis, treatment and detection.

[0058] Unique molecular identifiers (UMI): A randomized or specific nucleotide sequence, usually designed as a completely random nucleotide chain. In library construction, it is used as a molecular barcode to specifically label each molecular template.

[0059] Q30: This indicator is used to evaluate the accuracy of sequencing data. Q30 means that the probability of misidentification is 0.1%, that is, there may be one error in every 1,000 bases.

[0060] Single-strand consensus sequence (SSCS): Redundant reads from the same molecule are determined based on UMI sequence tags, and the consensus sequence is obtained after correcting inconsistent sequences.

[0061] Duplex consensus sequence (DCS): After obtaining SSCS based on UMI sequence tags, if the sequence from the positive strand and the sequence from the negative strand are confirmed to be from the same DNA segment based on UMI, sequence correction can be performed to obtain the consensus sequence, which is the DCS.

[0062] Single nucleotide variant (SNV): a substitution of a single nucleotide base pair in a DNA fragment.

[0063] Insertion and deletion (InDel): The insertion / deletion of several nucleotides in a DNA fragment, if located in the coding region, will lead to an increase / decrease in the encoded amino acids, resulting in a frameshift mutation or a non-frameshift mutation.

[0064] The schematic diagram of the mutation detection method of the present invention is shown in FIG. Figures 1-2 As shown, Figure 1 The schematic diagram of the analytical method of the present invention is as follows:

[0065] Specifically, after obtaining the sequencing data, the read families of the sense chain and antisense chain from the same template are constructed according to the alignment results and labels, and they are distinguished into three different data according to the standards. Some of the data are filtered, and consensus sequences are constructed for other data. Finally, the filtered read families and the constructed consensus sequences are constructed into a sequence set for mutation detection.

[0066] Figure 2 A schematic diagram of a specific embodiment of the present invention:

[0067] Specifically, for reads with the same UMI, they are classified as read families derived from the same template molecule. For the read families derived from the same template molecule, at least two or more read families are error corrected to generate SSCS consensus sequences. If the SSCS consensus sequence exists and its complementary sequence template is also SSCS, the positive and negative strands are corrected by DCS to generate consensus sequences;

[0068] If the read family with the same template molecule source has a read segment of 1 (i.e., a single template molecule) and its complementary template molecule SSCS exists, the single template molecule is corrected by SSCS to form a DCS consensus sequence; if the complementary template molecule of the single template molecule is also a single template molecule, a consensus sequence is generated by a correction method based on the single template molecule.

[0069] If the read family of the same template molecule originates from a read family with a read segment of 1 (i.e., a single template molecule) and there is no single template molecule of its complementary template molecule, those single template molecules whose Q30 ratio of reads is greater than or equal to 95% are retained.

[0070] The DCS consensus sequence, SSCS consensus sequence and single template molecule were merged to generate a sequence set.

[0071] Example 1: Analysis of samples using the mutation detection method of the present invention

[0072] The present invention selected two Twist ctDNA pan-cancer standards, one for detecting specificity and one for detecting sensitivity, both of which contain the same validation sites, and the specific information is shown in Table 1.

[0073] Table 1

[0074]

[0075] For two standard samples, targeted hybridization libraries were constructed, and two libraries were constructed for each standard sample. The original sequencing data were obtained by sequencing using Illumina NovaSeq6000, as shown in Table 2. After basic quality control of the original data, the hg19 reference genome was used for comparison.

[0076] Table 2

[0077]

[0078] The following processing was performed on the 4 samples tested

[0079] For paired-end sequencing data, extract the UMI sequences of read1 and read2 (i.e., the sense strand and antisense strand) and save them in the sequence identifier row.

[0080] Using hg19 as the reference genome, the data of the four test samples were aligned, and the reads that were not aligned to the reference genome in both read1 and read2, the reads that were not aligned at either end of read1 or read2, and the reads that were aligned to multiple positions of the reference genome were removed to generate the initial bam file.

[0081] For the alignment files of the above four samples, based on the UMI sequence tags, the readfamily with the same template molecule source is established. The read family with a read segment greater than or equal to 2 (i.e., the first data) in the read family is subjected to consensus sequence correction to generate SSCS. If there are differences in the bases at the same position in the read family, the base with a ratio greater than or equal to 0.7 in all reads is retained as the consensus sequence, and the quality value of the base is greater than Q30, otherwise the base at the position is judged as N.

[0082] For the alignment files of the above four samples after SSCS correction, the data is further corrected based on the double-stranded characteristics of the DNA molecule and the characteristics of the molecular source determined by the UMI sequence tag. If the generated SSCS, its complementary chain from the same template source is also SSCS, a consensus sequence is generated through DCS correction; if the sequencing data contains only one read segment, but the complementary chain from the same template source is SSCS, a single template molecule correction is performed through SSCS to generate a consensus sequence. If the sequencing data contains only one read segment, and the complementary chain from the same template source also contains only one read segment, a consensus sequence (i.e., the second data) is generated based on a single template molecule.

[0083] For the alignment files of the above 4 samples, the read molecules that could not be corrected by the above method had only one read in the sequencing data, and there was no complementary read from the same template source (i.e., the third data). Only the reads with a Q30 ratio greater than or equal to 95% were retained.

[0084] The reads corrected by the above strategy were merged for subsequent mutation detection.

[0085] Example 2. Detection specificity of different mutation detection methods

[0086] In order to further highlight the effect of the present invention, three other mutation detection methods are used for comparison:

[0087] Benchmarking method 1: UMI-based SSCS data correction (using SSCS only)

[0088] Benchmarking method 2: UMI-based DCS data correction (using DCS only)

[0089] Benchmarking method 3: A method for detecting low-frequency mutations based on double sequencing data (CN114530199A, using SSCS+DCS, but not performing the third data filtering of the present invention).

[0090] The corrected average depths of the four mutation detection methods are shown in Table 3.

[0091] Table 3. Average depth of different samples after correction by different methods

[0092]

[0093] It can be seen from Table 3 that the average depth obtained by the method of the present invention is the highest, which is significantly better than the three compared methods.

[0094] Next, the mutation detection software VarDict was used to perform mutation detection on the data corrected by different methods, and the SNV / InDel mutation sites with VAF greater than 0.1% were retained, and the detection specificity was calculated based on the ctDNA wild-type standard verification site, and the results are shown in Tables 4 to 6. It can be seen from the table that compared with the benchmark method 1, the benchmark methods 2 and 3 and the method used in the present invention have good specificity for detecting various mutation types.

[0095] Table 4. Specificity of mutation detection of ctDNA wild-type standards based on different calibration methods

[0096]

[0097]

[0098] Table 5. Specificity of mutation detection (SNV) of ctDNA wild-type standard based on different correction methods

[0099]

[0100] Table 6. ctDNA wild-type standard mutation detection specificity (InDel) based on different calibration methods

[0101]

[0102]

[0103] Example 3. Detection sensitivity of different mutation detection methods

[0104] Three comparative methods were used for analysis in accordance with Example 2. The detection sensitivity was calculated based on the ctDNA mutant standard validation site. The results are shown in Tables 7 to 9. As can be seen from the table, compared with benchmark methods 2 and 3, benchmark method 1 and the method used in the present invention have good sensitivity for detecting various mutation types.

[0105] Table 7. Sensitivity of mutation detection of ctDNA mutant standards based on different calibration methods

[0106]

[0107] Table 8. Sensitivity of mutation detection of ctDNA mutant standard products based on different calibration methods (SNV)

[0108]

[0109]

[0110] Table 9. Sensitivity of mutation detection of ctDNA mutant standards based on different calibration methods (InDel)

[0111]

[0112] Example 4. Data validity of different mutation detection methods

[0113] This example selects two samples of the two standard samples selected in Example 1 for gradient sampling, with three random samplings for each gradient, i.e., 20M, 40M, 60M…220M, and three random samplings for each gradient. The maximum amount of sampling data is no more than the amount of sample sequencing data.

[0114] The three comparison methods consistent with those in Example 2 were used for analysis and the effective ratios were statistically analyzed. The results are shown in Tables 10 and Figure 3 As shown in Table 10, it can be seen that the use of the method of the present invention for analysis can greatly increase the effective amount of data, and compared with the other three methods, the data efficiency is significantly improved.

[0115] Table 10. Valid data ratios of different samples and different data volumes

[0116]

[0117]

[0118]

[0119]

[0120]

Claims

1. A mutation detection method comprising the following steps: Obtain sequencing data; The sequencing data is compared with the reference genome, and the read families of the sense strand and the antisense strand are established according to the comparison results and the labels, and the read families are distinguished according to the following criteria based on the threshold of the number of read segments in the read family: First data: the threshold of the number of read segments in the read family is greater than or equal to 2; Second data: the threshold value of the number of reads in the read family is equal to 1 and can be complementary to the reads from the other strand of the same template molecule; and Third data: the threshold of the number of reads in the read family is equal to 1 and cannot be complementary to the reads from the other strand of the same template molecule; in, The read family is derived from the same template molecule; Filtering the read family in the third data to obtain a final read family; Construct a single-strand consensus sequence and a double-strand consensus sequence for the read families in the first data and the second data; The obtained single-strand consensus sequence and double-strand consensus sequence and the final read family obtained by filtering are used to construct a sequence set; The constructed sequence set was used for mutation detection.

2. The method according to claim 1, characterized in that The read family in the third data is filtered according to the following criteria: The Q30 ratio of the reads is greater than or equal to 95%; and / or SNV / InDel mutation sites with VAF greater than 0.1% were retained in the read segments.

3. The method according to claim 1, characterized in that When the read family in the first data constructs a single-strand consensus sequence and a double-strand consensus sequence, if there are differences in the bases at the same position of the read segments in the read family, the bases with a base ratio greater than or equal to 0.7 in all read segments are used as the consensus bases.

4. The method according to claim 1, characterized in that: The method also includes cleaning the sequencing data, including: removing low-quality reads and bases, repeated reads and contaminated adapter sequences to obtain cleaned data.

5. The method according to claim 1, characterized in that The quality values ​​of the constructed single-stranded consensus sequence and double-stranded consensus sequence are greater than Q30.

6. A device for mutation detection, comprising: An acquisition module, used for acquiring sequencing data; The alignment module is used to align the sequencing data with the reference genome, and to establish read families of the sense strand and the antisense strand respectively according to the alignment results and the labels, and to distinguish them according to the following criteria based on the threshold of the number of read segments in the read family: First data: the threshold of the number of read segments in the read family is greater than or equal to 2; Second data: the threshold value of the number of reads in the read family is equal to 1 and can be complementary to the reads from the other strand of the same template molecule; and Third data: the threshold value of the number of reads in the read family is equal to 1 and cannot be complementary to the reads from another strand of the same template molecule; wherein the read family is derived from the same template molecule; A filtering module, used for filtering the read family in the third data and obtaining a final readfamily; A construction module, used to construct a single-stranded consensus sequence and a double-stranded consensus sequence from the read families in the first data and the second data; An output module is used to construct a sequence set using the obtained single-strand consensus sequence and double-strand consensus sequence and the final read family obtained by the filtering; and the constructed sequence set is used for mutation detection.

7. The device according to claim 6, characterized in that In the filtering module, the filtering is performed according to the following standards: The Q30 ratio of the reads is greater than or equal to 95%; and / or SNV / InDel mutation sites with VAF greater than 0.1% were retained in the read segments.

8. The device according to claim 7, characterized in that When the readfamily described in the first data in the construction module constructs a single-stranded consensus sequence and a double-stranded consensus sequence, if there are differences in the bases at the same position of the reads in the read family, the bases with a base ratio greater than or equal to 0.7 in all reads are used as the consensus bases.

9. A device for mutation detection, comprising: at least one processor; as well as a memory communicatively connected to at least one of the processors; wherein, The memory stores instructions executable by the processor, and the instructions are used to be executed by the processor to implement the mutation detection method according to any one of claims 1 to 5. 10 . A storage medium storing computer instructions, wherein the computer instructions are used to be executed by the computer to implement the mutation detection method according to any one of claims 1 to 5.

Images