Method for identifying subgenome in polyploidy genome assembly process
The modified dynamic programming algorithm recognizes and removes subgenomes in polyploid genome assembly, solves the problems of rearrangement events and logical vulnerabilities, and achieves more accurate and automated analysis results, avoiding the error of manual correction.
Patent Information
- Application Number
- CN202411201299.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-29
- Publication Date
- 2024-11-08
- Estimated Expiration
- 2044-08-29
AI Technical Summary
The prior art is difficult to accurately identify and remove subgenomes during the assembly of polyploid genomes, especially because of inaccurate analysis results caused by rearrangement events and "bypass" logic vulnerabilities. Conventional methods require manual inspection and manual correction, and local trimming. Possibly disrupt the genetic structure.
Using a modified dynamic programming algorithm, by obtaining the alignment matrix between Contig and target sequence, calculating the highest scored path and removing duplicate sequences, solving logical vulnerabilities and automating the processing of rearrangements, using dynamic programming algorithm to connect HSPs in series to maximize overall comparison Score.
More accurate subgenome recognition and removal are achieved, and the "bypass" incidents are solved automatically, which significantly improves the accuracy and automation of the analysis results, and avoids the errors caused by manual intervention.
Smart Images

Figure CN118918955A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of genome assembly, and in particular to a method for identifying subgenomes in a polyploid genome assembly process. Background Art
[0002] A polyploid genome is one in which an organism's cells contain three or more complete sets of chromosomes. This is different from diploid organisms (such as humans), which have only two sets of chromosomes in their cells. Polyploidy is particularly common in plants, but it also occurs in some animals.
[0003] The main types of polyploid genomes are:
[0004] 1. Autopolyploidy: multiple sets of chromosomes from the same species.
[0005] 2. Allopolyploidy: multiple sets of chromosomes from different but related species.
[0006] Characteristics of sequence similarity between polyploid chromosomes:
[0007] 1. High similarity: Chromosomes in autopolyploids usually have high sequence similarity, which may lead to increased complexity in genome assembly and analysis.
[0008] 2. Homologous regions: There are a large number of homologous regions between different chromosome sets. These regions may be almost identical in sequence, especially in recently formed polyploids.
[0009] 3. Progressive differentiation: Different chromosome sets may gradually diverge over time. This differentiation may lead to a decrease in sequence similarity and the generation of subgenome-specific variation.
[0010] 4. Gene redundancy: Multiple nearly identical copies of genes exist in different chromosome sets. This redundancy may lead to complex regulation of gene expression and function.
[0011] 5. Structural variation: Structural variations such as rearrangement, insertion, and deletion may occur between chromosomes. These variations may lead to differences in local sequence similarity.
[0012] 6. Homologous recombination: Homologous recombination may occur between highly similar chromosomes. This may lead to gene conversion and sequence assimilation.
[0013] 7. Subgenome preference: In some polyploids, a certain subgenome may be preferentially retained or expressed, which may lead to different degrees of sequence conservation between subgenomes.
[0014] 8. Distribution of repetitive sequences: The distribution of repetitive sequences such as transposons may be different in different chromosome sets, which may lead to differences in local sequence similarity.
[0015] 9. Functional redundancy and new functions: Highly similar sequences may lead to gene function redundancy. At the same time, it also provides opportunities for gene function differentiation and the generation of new functions.
[0016] 10. Sequence loss: After polyploidization, loss of genes or chromosome segments may occur. This phenomenon may cause originally similar regions to become different.
[0017] In summary, there is a certain similarity in the sequences between sister chromatids of polyploid chromosomes, but due to their independent evolution and other conditions, their genes have a large number of rearrangements, insertions and deletions, and other large differences. Understanding these characteristics is of great significance for accurately analyzing the structure, function and evolution of polyploid genomes, but it also complicates the assembly and analysis of polyploid genomes.
[0018] After genome assembly analysis, it is expected to obtain a consensus sequence representing the haploid genotype - after the genome of a polyploid genome data species is initially assembled by the assembly software, a contig or scaffold version of the genome is obtained, which can be considered as a fragmented sequence of the genome chromosome. These contigs must contain homologous sequences from different homology groups or different subgenomes. This will result in a large number of homologous sequences representing the same chromosome region in these homology groups. These sequences have certain similarities, but they also have certain differences, deletions and rearrangements between each other. The presence of these redundant homologous sequences will affect subsequent analysis (such as an increase in the number of genes and chromosome positioning errors), and they need to be removed as much as possible.
[0019] Conventional methods include:
[0020] 1. Reference sequence alignment method:
[0021] The reference sequence alignment method generally uses the reference sequence of closely related species, and aligns these contigs to the reference sequence using alignment tools with high mismatch tolerance such as Minimap2 and Lastz, and then sets a threshold. Sequences below this threshold will be excluded, and the remaining sequences will be retained as the result;
[0022] 2. Self-comparison elimination method:
[0023] All contigs are aligned in pairs using alignment tools, and then the local alignment results are linked together using Lastz's Chain algorithm. The longer of the two best aligned sequences is retained, and the shorter one is removed as a subgenome. Of course, some tools use coverage as a criterion to remove homologous contigs with low coverage as subgenomes, such as HaploMerger2, Redundans, purge_haplogs, etc.
[0024] 3. Spectrum comparison method:
[0025] Using the BioNano or HiC sequencing results, all contigs were aligned with the map. In the same alignment region, the contig with the highest score was retained, and other contigs in the same region were removed as subgenomes.
[0026] However, the alignment tools used by the first two methods, such as the Chain algorithm used by LastZ and Minimap2, are designed mainly to obtain colinearity results between genomes. For reverse complementation or jump rearrangements, the algorithm will not recognize the alignment results by default, which affects the final removal of subgenomes (such as Figure 1 ).
[0027] The spectrum comparison method also has certain limitations:
[0028] 1) Any sister chromatid in the polyploid used to construct the map may not be the same as the chromatid actually assembled;
[0029] 2) For small individual samples, all samples used to construct the map may not be the same samples used for sequencing, and sister chromatids may undergo homologous recombination during sexual reproduction;
[0030] 3) The assembled Contig is a consensus sequence, that is, its assembly result may be a mixed result of multiple homologous chromosome types, and there may be mis-spellings inside it.
[0031] Therefore, during the polyploid assembly process, there will inevitably be a large number of rearrangements and differences after the map and Contig are aligned. The conventional Chain algorithm's neglect of large-scale rearrangements will lead to a large number of subsequent analysis errors, which need to be manually corrected.
[0032] To address these challenges, the common method is to use dynamic programming algorithms and their derived heuristic or greedy optimization algorithms to concatenate HSPs (High-scoring Segment Pairs) obtained from local alignments (which can be alignments between Contigs, alignments between Contigs and reference sequences, or alignments between Contigs and genetic maps) (Chain algorithm) and calculate their similarity (alignment length after concatenation / total Contig length) to determine and remove redundant homologous Contig sequences. Figure 2 shown.
[0033] a) Assume there are two sequences, Ref1 (5400 bp) and Ref2 (7600 bp). Ref1 and Ref2 are highly homologous. If rearrangement is not considered, Ref1 can be completely contained by Ref2. Figure 2 The same color areas in the image represent the same sequence. The arrows from left to right represent 5'→3', and from right to left represent 3'←5'. The boxes represent the unique sequence of Ref2, which has no homology with Ref1. The gray areas are flipped, and the black areas are jump-flip rearrangements.
[0034] b) The ordinate represents the Ref1 alignment coordinate, and the abscissa represents the Ref2 alignment coordinate. The HSPs obtained by aligning Ref1 and Ref2 are sorted according to the coordinates to construct a matrix. If two regions can be aligned, their alignment scores are placed in the corresponding positions of the matrix.
[0035] c) Use dynamic programming to connect HSPs in series.
[0036] principle:
[0037] 1. Use dynamic programming algorithm to connect similar HSPs into longer sequences.
[0038] 2. Consider the distance, direction and score between HSPs to find the highest scoring location.
[0039] Algorithm steps:
[0040] 1. Sort all HSPs according to their starting positions in the reference sequence.
[0041] 2. Traverse the sorted HSPs from left to right: a. For each HSP, try to connect it with the previous HSPs (to the left or above). b. Use dynamic programming to calculate the best connection method.
[0042] 3. Dynamic programming score calculation: score(i) = max{score(j) + match_score(i) - gap_penalty(i,j)} where:
[0043] i is the current HSP, j is the previous HSP, match_score(i) is the score of the current HSP, gap_penalty(i,j) is the gap penalty when connecting i and j (for simplicity, all gap_penalties in this example are 0);
[0044] 4. Find the matrix position with the highest score
[0045] 4. Backtrack the score matrix to find the best chain path. (Red and green arrows)
[0046] 5. Filter out chains with lower scores based on the set threshold.
[0047] 6. Output the final chains result.
[0048] 7. Figure 2 The HSP represented by the gray arrow has part of its Ref1 coordinates (4500-5400) repeatedly counted, so only the one with the highest score in these two regions is retained for Chain calculation.
[0049] The core idea of this algorithm is to maximize the overall alignment score while maintaining sequence colinearity.
[0050] It can be found that under the premise of maintaining sequence colinearity, using the conventional algorithm, Ref1's 5400bp can be matched with Ref2, but its optimal calculation result is only 4300bp. Assuming that the threshold for merging homologous sequences is 85% similarity, the Ref1 sequence will be omitted, resulting in inaccurate calculation results.
[0051] In addition, for the self-alignment elimination method, since the current algorithm focuses on the 1vs 1 similarity comparison and processing between sequences, Figure 3 In the situation shown (the Contig in the dotted box is completely included in the other two Contigs, but in a 1vs1 comparison, no Contig can meet the screening threshold, which we call a "bridging" event), there is currently no algorithm that can automatically handle this logical loophole, and it relies on manual inspection by humans, which is time-consuming and laborious.
[0052] To solve this problem, the following technical solutions are currently used:
[0053] 1. Manually check the alignment results. Manually check the alignment results of each allele and manually delete the homologous sequences (refer to the comparative documents CN111445948A CN111584004A);
[0054] 2. Local trimming method, for the comparison results of genetic map and reference sequence. If multiple Contigs can correspond to the same reference sequence or a region of the genetic map, the sequencing reads that make up these Contigs are collected, and then locally reassembled or clustered, and then the reassembled Contigs are compared with the map again, until a Contig that meets the threshold is obtained, and other Contigs are discarded (refer to comparative documents CN109326323B CN110020726A CN112289382A);
[0055] 3. Sample pretreatment method: construct SNP linkage groups before sequencing. Select raw data according to linkage groups and construct genetic maps. Only haploid data that meet the same SNP linkage group are analyzed, thus solving the polyploidy problem from the source (refer to comparative documents CN112289382B, CN111816248A).
[0056] All three methods are time-consuming and difficult to repeat. At the same time, it is inevitable that there will be some differences between the genetic map and the Contig. Local trimming may not necessarily produce the best results. Excessive trimming may damage the gene structure and lead to a series of subsequent problems. Summary of the invention
[0057] 1. Technical issues to be resolved
[0058] In view of the shortcomings of the prior art, the present invention provides a new method for identifying subgenomes in the polyploid genome assembly process based on a modified dynamic programming algorithm, which can eliminate the influence of rearrangement events on the analysis results. Compared with the existing methods, this method can more accurately identify subgenomes in polyploids and perfectly solve the logical loophole of "bridging" events.
[0059] (II) Technical solution
[0060] To achieve the above objectives, the present invention is implemented by the following technical scheme: A method for identifying subgenomes during polyploid genome assembly, specifically comprising the following steps:
[0061] S1. Obtain the comparison result of any Contig (Query). For any Query, obtain the comparison coordinate matrix between it and any target Subject.
[0062] S2, the matrix is characterized by one dimension of coordinates being the comparison coordinates of Contig, and the other dimension of coordinates being the coordinates of Subject mapped onto Query;
[0063] S3. For any cell G(i,j), score(i,j)=max{score(m,n)+match_score(I,j)-gap_pendalty({m,n}(i,j))}, where i>0, j>0, m<=i, n<=j;
[0064] S4, scanning the entire matrix to find the highest scoring value, and using a dynamic programming algorithm to calculate the sum X of the base coordinates involved in the path with the highest score;
[0065] S5. Calculate the ratio of X to the length of the Query sequence. If it is greater than the set threshold, mark the Query as a homologous sequence of the Subject and remove it.
[0066] S6. For any contig set S, the above steps S1-S5 algorithm can split the contig into two sets, S 去除 and S 保留 ;
[0067] S7. In view of the logical loopholes in the self-comparison elimination method, 保留 Operate any C in the above code, query the previous dynamic programming results, and find the S with the highest coordinate sum with C. 保留 The first two contigs C x and C y , C x and C y The sequences are concatenated to obtain sequence C connect , C connect Compare with C, and then perform steps S1-S6 on both, and update S 保留 ;
[0068] S8. Traverse S 保留 Until the end.
[0069] Preferably, assuming there are two sequences Ref1 and Ref2, Ref1 and Ref2 are highly homologous, and if rearrangement is not considered, Ref1 can be completely contained by Ref2.
[0070] Preferably, the number of base pairs of Ref1 is 5400 bp, and the number of base pairs of Ref2 is 7600 bp.
[0071] Preferably, the ordinate represents the alignment coordinate of Ref1, the abscissa represents the alignment coordinate of Ref2 HSP mapped to Ref1, and the matrix is constructed by sorting the new coordinates. If two regions can be aligned, their alignment scores are placed in the corresponding positions of the matrix.
[0072] Preferably, HSPs are connected in series using dynamic programming, and the calculation formula is as follows:
[0073] score(i)=max{score(j)+match_score(i)-gap_penalty(i,j)};
[0074] Among them, the coordinates of the j cell are less than or equal to the coordinates of the i cell.
[0075] Where: i is the current HSP, j is the previous HSP, match_score(i) is the score of the current HSP, and gap_penalty(i,j) is the gap penalty when connecting i and j.
[0076] Preferably, the results show that the aligned region obtained by this Chain is 5200bp, which is significantly better than the 4300bp of the conventional algorithm.
[0077] (III) Beneficial effects
[0078] The present invention provides a method for identifying subgenomes during the assembly of a polyploid genome. Compared with the prior art, it has the following beneficial effects:
[0079] (1) The method for identifying subgenomes during the polyploid genome assembly process obtains the optimal result by connecting any query and subject HSP in series through a modified dynamic programming algorithm. The core innovation is to replace the coordinates of the subject HSP with the coordinates mapped to the query, so that the influence of subject rearrangement events on the HSP chain analysis results can be excluded with maximum likelihood.
[0080] (2) The method for identifying subgenomes during the assembly of the polyploid genome can connect HSPs from any source by using a modified planning algorithm, which may represent sequence alignment coordinates, genetic maps or SSR molecular marker map alignment coordinates, enzyme cleavage maps or HiC alignment matrices and other arbitrary data. As long as it involves a tensor or vector for similarity scoring and sorting, this method can be applied. Therefore, the scope of protection needs to be expanded. For the sake of convenience in understanding, the example given only uses the starting coordinate vector of the sequence alignment result as an example for ease of understanding, but in fact it can be applied to any similarity alignment system represented by a vector or tensor.
[0081] (3) This method accepts a sequence formed by randomly connecting multiple query sequences as input. Since this method can eliminate the impact of rearrangement events on the analysis results, it can automatically and perfectly solve the "bridging" problem. BRIEF DESCRIPTION OF THE DRAWINGS
[0082] Figure 1 A schematic diagram of the shortcomings of the existing LastZ and Minimap2 comparison results;
[0083] Figure 2 This is a schematic diagram of the dynamic programming Chain algorithm under the conventional framework;
[0084] Figure 3 This is a schematic diagram of the logical loopholes in the current homologous contig screening process;
[0085] Figure 4 Schematic diagram of HSP after different methods of the present invention are connected in series;
[0086] Figure 5 It is a schematic diagram of the framework of the Chain algorithm of the present invention;
[0087] Figure 6 This is the first demonstration diagram of the Chain algorithm of the present invention;
[0088] Figure 7 This is a second demonstration diagram of the Chain algorithm of the present invention;
[0089] Figure 8 This is the first schematic diagram of the scoring matrix obtained by the dynamic programming Chain algorithm under ideal conditions in the calculation case of the present invention;
[0090] Fig. 9 This is a second schematic diagram of the scoring matrix obtained by the dynamic programming Chain algorithm under ideal conditions in the calculation case of the present invention. DETAILED DESCRIPTION
[0091] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0092] See also Figure 1-9 The embodiment of the present invention provides a technical solution: a method for identifying subgenomes during polyploid genome assembly, specifically including the following embodiments:
[0093] A method for identifying subgenomes during polyploid genome assembly, comprising the following steps:
[0094] S1. Obtain the comparison result of any Contig (Query). For any Query, obtain the comparison coordinate matrix between it and any target Subject.
[0095] S2, the matrix is characterized by one dimension of coordinates being the comparison coordinates of Contig, and the other dimension of coordinates being the coordinates of Subject mapped onto Query;
[0096] S3. For any cell i,j, score(i,j)=max{score(m,n)+match_score(I,j)-gap_penalty({m,n}(i,j))}, where i>=0, j>=0, m<=i, n<=j;
[0097] S4, scanning the entire matrix to find the highest scoring value, and using a dynamic programming algorithm to calculate the sum X of the base coordinates involved in the path with the highest score;
[0098] S5, calculating the ratio of X to Query, if it is greater than the set threshold, marking Query as a homologous sequence of Subject and removing it;
[0099] S6. For any contig set S, the above steps S1-S5 algorithm can split the contig into two sets, S 去除 and S 保留 ;
[0100] S7. In view of the logical loopholes in the self-comparison elimination method, 保留 Operate any C in the above code, query the previous dynamic programming results, and find the S with the highest coordinate sum with C. 保留 The first two contigs C x and C y , C x and C y The sequences are concatenated to obtain sequence C connect , C connect Compare with C, and then perform steps S1-S6 on both, and update S 保留 ;
[0101] S8. Traverse S 保留 Until the end.
[0102] In the embodiment of the present invention, it is assumed that there are two sequences Ref1 and Ref2, and Ref1 and Ref2 are highly homologous. Without considering rearrangement, Ref1 can be completely contained by Ref2. Figure 4The same color areas in the figure represent the same sequence. The arrows from left to right represent 5'→3', and from right to left represent 3'←5'. The boxes represent the unique sequences of Ref2, which have no homology with Ref1. The light gray area is flipped, and the gray area is rearranged by jump flipping. The number of base pairs of Ref1 is 5400bp, and the number of base pairs of Ref2 is 7600bp. The ordinate reaches the Ref1 alignment coordinate, and the abscissa represents the alignment coordinate of Ref2 HSP mapped to Ref1. The matrix is constructed according to the new coordinates. If the two regions can be aligned, their alignment scores are placed in the corresponding position of the matrix. The HSPs are connected in series using dynamic programming. The calculation formula is as follows:
[0103] score(i)=max{score(j)+match_score(i)-gap_penalty(i,j)};
[0104] Where: i is the current HSP, j is the previous HSP, match_score(i) is the score of the current HSP, gap_penalty(i,j) is the gap penalty when connecting i and j (for simplicity, all gap_penalties in our example are 0).
[0105] The results show that the aligned region obtained by this Chain is 5200bp, which is significantly better than the 4300bp of the conventional algorithm.
[0106] The data of polyploid chelicerate samples were compared using this method and the conventional method, and the results are shown in Table 1.
[0107] Table 1 Statistical table of embodiments
[0108]
[0109]
[0110] like Figure 1Figure A shows the loopholes of the relevant algorithms in the Lastz algorithm. Assuming that the color box is the reverse complementary alignment result, Lastz will ignore the intermittent reverse complementary alignment results during the HSPs (High-scoring Segment Pair) concatenation process. Figure B shows the common shortcomings of Minimap and Lastz. For the jump rearrangement that is common in sub-chromosomes, the algorithm of this software will ignore it by default. In addition, these tools need to annotate the repetitive sequences in advance before analysis to prevent the repetitive sequences from affecting their analysis results. Finally, for large-scale rearrangements, such as the overall loss of part of the chromosome region, or jumping to other positions, these algorithms will consider it to have lost homology, and the final alignment results will not be recognized by the Chain algorithm, thus affecting the analysis accuracy.
[0111] Figure 5 The left picture in the middle is the LastZ Chain algorithm, and the right picture is this method (the results of the LastZ algorithm are retained for easy inspection and comparison).
[0112] Figure 6 Schematic diagram of the comparison results of the two sequences in . Assume that there are two sequences, Ref1 (5400bp) and Ref2 (7600bp). Ref1 and Ref2 are highly homologous. If rearrangement is not considered, Ref1 can be completely contained by Ref2. The areas with the same color in the figure represent the same sequence. The arrows from left to right represent the 5'→3' direction, and from right to left represent the 3'←5' direction. The boxes represent the unique sequences of Ref2, which have no homology with Ref1. The gray-white area is flipped, and the gray area undergoes jump-flip rearrangement.
[0113] Meanwhile, the contents not described in detail in this specification belong to the prior art known to those skilled in the art.
[0114] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.
[0115] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.
Claims
1. A method for identifying subgenomes during polyploid genome assembly, characterized in that: The specific steps include: S1. Obtain the comparison result of any Contig, and for any Query, obtain the comparison coordinate matrix between it and any target Subject; S2, the matrix is characterized by one dimension of coordinates being the comparison coordinates of Contig, and the other dimension of coordinates being the coordinates of Subject mapped onto Query; S3. For any cell i,j, score(i,j)=max{score(m,n)+match_score(i,j)-gap_penalty({m,n}(i,j))}, where i>=0, j>=0, m<=i, n<=j; S4, scanning the entire matrix to find the highest scoring value, and using a dynamic programming algorithm to calculate the sum X of the base coordinates involved in the path with the highest score; S5, calculating the ratio of X to Query, if it is greater than the set threshold, marking Query as a homologous sequence of Subject and removing it; S6. For any contig set S, the above steps S1-S5 algorithm splits the contig into two sets, S 去除 and S 保留 ; S7. In view of the logical loopholes in the self-comparison elimination method, 保留 Operate any C in the above code, query the previous dynamic programming results, and find the S with the highest coordinate sum with C. 保留 The first two contigs C x and C y , C x and C y The sequences are concatenated to obtain sequence C connect , C connect Compare with C, and then perform steps S1-S6 on both, and update S 保留 ; S8. Traverse S 保留 Until the end.
2. The method for identifying subgenomes in a polyploid genome assembly process according to claim 1, characterized in that: The ordinate represents the comparison coordinate of Ref1, and the abscissa represents the comparison coordinate of Ref2 HSP mapped to Ref1. The matrix is constructed by sorting the new coordinates. If two regions can be compared, their comparison scores are placed in the corresponding positions of the matrix.
3. The method for identifying subgenomes in a polyploid genome assembly process according to claim 2, characterized in that: Use dynamic programming to connect HSP in series, and the calculation formula is as follows: score(i)=max{score(j)+match_score(i)-gap_penalty(i,j)}; Where: i is the current HSP, j is the previous HSP, match_score(i) is the score of the current HSP, and gap_penalty(i, j) is the gap penalty when connecting i and j.
4. The method for identifying subgenomes during polyploid genome assembly according to claim 3, characterized in that: The results show that the region that this Chain can obtain for comparison is 5200bp.
Citation Information
Patent Citations
Genomic sequence assembling method, structural variation detection method and relevant system
CN107133493A
Genome stack array, genome framework, genome sequence assembling method and genome sequence assembling system
CN115440302A
Polyploid transcriptome subgenome typing method and computer readable storage medium
CN116779035A
Method for typing assembly and variation identification of homologous polyploid genome
CN118538291A
Method and system for multiple dot plot analysis
KR101832834B1
