Method for detecting fusion gene in high-throughput sequencing data

Abstract

The invention relates to the technical field of biological detection, and discloses a method for detecting a fusion gene in high-throughput sequencing, which comprises the following steps: S1, obtaining sequencing data and performing quality control to obtain filtered sequencing data; s2, comparing the high-quality sequencing data with a reference genome to obtain a comparison file; s3, screening abnormal alignment sequences according to the alignment file, and obtaining fusion breakpoint sequence fragments and candidate fusion genes; s4, performing statistics on fusion information according to the candidate fusion genes; s5, screening fusion events to obtain potential fusion gene pairs; s6, calculating the fusion confidence coefficient of the fusion gene pair; s7, judging whether the fusion gene is a true positive fusion gene or not according to the fusion confidence, and outputting a result; according to the method, the abnormal alignment sequence is screened and filtered according to the aligned parameter information, and the fusion gene pair with higher confidence is obtained by calculating the fusion confidence, so that the detection rate and correctness of the fusion gene can be effectively improved.

Description

Technical Field

[0001] The present invention relates to the field of biological detection technology, and in particular to a method for detecting fusion genes in high-throughput sequencing data. Background Art

[0002] Fusion genes refer to the rearrangement of two or more unrelated genes, which makes these genes under the control of the same set of regulatory elements, forming new gene products, chimeric genes, and acquiring new characteristics and functions. Fusion genes play an important role in the occurrence and development of cancer, which can lead to the inactivation of tumor suppressor genes or the activation of proto-oncogenes, and promote the growth and metastasis of cancer cells.

[0003] Traditional methods for identifying fusion genes include immunohistochemistry, fluorescence in situ hybridization, and PCR. With the development of sequencing technology, fusion genes can be effectively analyzed by sequencing genomic regions using high-throughput sequencing technology and using bioinformatics methods, and multiple fusion genes can be detected at the same time.

[0004] Currently, commonly used software for detecting fusion genes include FACTERA, Delly, and GeneFuse. However, due to the differences in fusion detection algorithms, the following problems may occur: (1) Different algorithms over-detect different types of potential fusion events, resulting in the detection of false positive fusion genes; (2) Some fusion events occur at a low frequency and may be covered or ignored by other high-frequency events, resulting in difficulty in detecting low-frequency fusion genes. Summary of the invention

[0005] Based on the above problems, the present invention proposes a method for detecting fusion genes in high-throughput sequencing data. The present invention screens and filters abnormal alignment sequences according to the alignment parameter information, and calculates the fusion confidence to obtain fusion gene pairs with higher confidence, which can effectively improve the detection rate and accuracy of fusion genes.

[0006] A method for detecting fusion genes in high-throughput sequencing data, the specific steps are as follows:

[0007] Step S1: obtaining high-throughput sequencing data, performing quality control filtering on the sequencing data, and obtaining filtered sequencing data;

[0008] Step S2: aligning the filtered sequencing data with the reference genome to obtain an alignment file;

[0009] Step S3: Screening abnormal alignment sequences according to the alignment file, and matching the abnormal alignment sequences to obtain candidate fusion sequence fragments and candidate fusion genes;

[0010] Step S4: Counting fusion information based on each candidate fusion gene;

[0011] Step S5: Screening fusion events according to the fusion information to obtain potential fusion gene pairs;

[0012] Step S6: calculating the fusion confidence of the fusion gene pair;

[0013] Step S7: Determine whether it is a true positive fusion gene based on the fusion confidence and output the result.

[0014] Preferably, the quality control filtering described in step S1 includes: filtering low-quality sequencing data and adapter sequence information.

[0015] Preferably, the situations of abnormally aligned sequences in step S3 include: situation 1: in paired-end sequencing, part of the read length is aligned to position 1 of the reference genome, and the other part is aligned to position 2 of the reference genome; situation 2: in paired-end sequencing, read length 1 is aligned to the reference genome, and read length 2 is aligned to other reference genomes; situation 3: in paired-end sequencing, read length 1 and read length 2 are respectively aligned to the same reference genome, but the alignment directions of the read length pairs are the same.

[0016] Preferably, the screening principle described in step S3 includes: (1) if the quality of the read pair in the double-end sequencing does not meet the quality control requirements, the abnormal alignment sequence information is deleted; (2) if at least one of the read pairs in the double-end sequencing belongs to Duplicates, the abnormal alignment sequence information is deleted; (3) if the read pair in the double-end sequencing belongs to case 1 of abnormal alignment sequence, the abnormal alignment sequence information with a shorter alignment part is deleted; (4) if the read pairs in the double-end sequencing are aligned to different positions of the reference gene respectively, if the position distance is less than the filtering threshold 1, the abnormal alignment sequence information is deleted.

[0017] Preferably, the candidate fusion gene statistics step described in step S4 is: according to the candidate fusion gene, all sequencing data aligned to the gene are selected from the alignment file, and statistics are performed based on the sequencing data on the candidate gene, and the statistical information includes: the average sequencing depth of the candidate gene, gene coverage, number of fusion breakpoints, sequencing depth corresponding to the fusion breakpoints, number of complete matches and number of abnormal alignments.

[0018] Among them, the number of fusion breakpoints refers to the number of breakpoint information contained in the abnormal alignment sequences detected for the gene, which includes the corresponding genes and their breakpoint numbers, and is represented by an array: <corresponding genes, number of breakpoints>; the number of complete matches refers to the number of sequences that do not belong to the abnormal alignment and are aligned to the gene; the number of abnormal alignments is the number of abnormal alignment sequences aligned to the gene.

[0019] Preferably, in step S5, when the gene coverage or average sequencing depth of the candidate gene is lower than the gene coverage threshold 1 and the average sequencing depth threshold 1, it is considered that there is a false positive in the sequencing of the candidate gene, and the candidate fusion gene is removed; when the gene coverage and the average sequencing depth of the candidate gene are both higher than the gene coverage threshold 1 and the average sequencing depth threshold 1, it is considered that the candidate fusion candidate gene can enter the screening step of the candidate gene pair.

[0020] Preferably, the gene coverage threshold 1 is the average coverage value of all genes, and the average sequencing depth threshold 1 is the average sequencing depth value of all genes.

[0021] The steps of screening candidate gene pairs include:

[0022] (1) Obtain the average sequencing depth, gene coverage, number of fusion breakpoints, sequencing depth of corresponding fusion breakpoints, number of complete matches, and number of abnormal alignments of fusion candidate gene A;

[0023] (2) Perform a t-test based on the number of breakpoints of the fusion candidate gene A to determine whether there is a significant difference. If there is a significant difference, the gene ranked first according to the number of breakpoints is selected as the suspected candidate gene B. If there is no significant difference, the top three gene loci are selected as suspected candidate genes B;

[0024] (3) Obtaining the fusion breakpoint number parameter information of candidate gene B and sorting them. If the fusion breakpoint number of candidate gene B with candidate gene A is higher than the breakpoint threshold, candidate gene A and candidate gene B are selected as a candidate gene pair;

[0025] (4) If the fusion breakpoint number ranking of candidate gene B with candidate gene A is lower than the breakpoint threshold, and the breakpoint number of candidate gene B is significantly different after the t-test, then candidate gene A and candidate gene B are filtered as a fusion gene pair;

[0026] (5) If the fusion breakpoint number ranking of candidate gene B with candidate gene A is lower than the breakpoint threshold, and there is no significant difference in the breakpoint number of candidate gene B after the t-test, then candidate gene A and candidate gene B are selected as the fusion gene pair;

[0027] (6) If the candidate gene B does not have the required number of fusion breakpoints, the sequencing depth and gene coverage of the gene are determined. If the sequencing depth and gene coverage meet the reporting criteria, the candidate gene A and the candidate gene B are not considered to be a fusion gene pair;

[0028] (7) If there is no fusion breakpoint in candidate gene B, the sequencing depth and gene coverage of the gene are determined. If the sequencing depth and gene coverage do not meet the reporting standard, it is considered that the fusion breakpoints may not be detected due to incomplete sequencing, and candidate gene A and candidate gene B are considered to be a fusion gene pair.

[0029] Preferably, the fusion confidence calculation formula in step S6 is:

[0030]

[0031] Preferably, the judgment rule in step S7 is: if the fusion confidence ≥ confidence threshold 1, the fusion gene pair is considered to be a true fusion breakpoint; if confidence threshold 1 < fusion confidence ≤ confidence threshold 2, the fusion gene is considered to be a potential fusion breakpoint; if confidence < confidence threshold 3, the fusion gene is considered to be a false fusion breakpoint.

[0032] Preferably, the output result in step S7 includes: fusion gene name, breakpoint location, fusion depth, fusion type, and fusion confidence.

[0033] Compared with the prior art, the present invention has the following beneficial effects:

[0034] (1) The present invention can effectively obtain unknown fusion genes by comparing and screening sequencing data with reference genomes without relying on known fusion gene information;

[0035] (2) The present invention uses the dual characteristics of sequencing depth and gene coverage to test the detection reliability of the gene, which can effectively filter out false positives of fusion gene detection caused by low-quality sequencing results;

[0036] (3) The present invention performs a double check on the comparison information of the two fusion genes, thereby increasing the reliability and accuracy of detecting the fusion gene pair. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 Flow chart of the method for detecting fusion genes in an embodiment of the present invention.

[0038] Figure 2 4 is a flow chart of the screening rules for candidate gene pairs in an embodiment of the present invention. DETAILED DESCRIPTION

[0039] The technical solution of the present invention is further described below in conjunction with the accompanying drawings and embodiments.

[0040] Example:

[0041] Step S1: Obtain RNA sequencing data, perform quality control on the sequencing data, and obtain high-quality sequencing data;

[0042] Specifically, in step S1 of this embodiment, fastp software is used for quality control filtering, and the step includes filtering low-quality sequencing data and adapter sequence information, wherein Q30 < 80% is considered to be a low-quality sequence and is filtered.

[0043] Step S2: Align the high-quality sequencing data with the reference genome to obtain the alignment bam file.

[0044] In this example, the GRCh37 (hg19) human genome was selected from the NCBI database (https: / / www.ncbi.nlm.nih.gov / ) as the reference genome, and the high-quality sequencing data was aligned with the reference genome using the BWA software to obtain an alignment file in the bam file format, wherein the bam file contains alignment information, base quality value, FLAG value, CIGAR tag, start site, alignment quality, TLEN tag, and SA tag.

[0045] Specifically, the CIGAR tag is composed of "M (match), I (insertion), D (deletion), N (skipped bases on the reference), S (soft clipping), H (hard clipping)" to record the alignment information. For example, "56M1I30M" means that the alignment of the sequence is: match 56 bases, insert 1 base, and match 30 bases; "FLAG" refers to the bit operation of the bam file, which is used to indicate the sequence alignment information; "TLEN" refers to Templatelength, which is a parameter used by the Bam file to estimate the fragment length. It is only available when the alignment file is double-end sequencing data. When the double-end sequencing data is aligned to different chromosome positions, the value is 0; "SA tag" refers to "Secondary Alignmenttag", which is a parameter in the bam file used to indicate the position of the matched chimera. When a sequence is aligned to two genomic positions that are far apart, two alignment records will be formed and recorded in SA In the tag, for example: "SA:Z:NC_000007.13,116415165,-,72M79S,60,0" means that part of the sequence is aligned to the genome NC_000007.13, the starting position is: 116415165, matched to the negative strand, the matching situation is: 72M79S, and the matching quality is 60.

[0046] Step S3: Screen the abnormal alignment sequences according to the bam file, and match the abnormal alignment sequences to obtain candidate fusion sequence fragments and candidate fusion genes.

[0047] Specifically, in this embodiment, a custom python code package (Get_abnormal_sequence.py) is used to obtain abnormal alignment sequences from bam files. The specific screening steps are: (1) According to the CIGAR value, if "S" appears in CIGAR, it indicates that shearing has occurred, and it is included in the abnormal alignment sequence; (2) When the parameter TLEN in the bam file is 0, and the parameter RNAME and the parameter MRNM are inconsistent, it is included in the abnormal alignment sequence; (3) The parameter FLAG value in the bam file is extracted, converted into binary and it is determined whether the fifth and sixth bits are consistent. When the fifth and sixth bits are consistent, it indicates that the alignment direction of the double-end sequencing is consistent, and it is included in the abnormal alignment sequence.

[0048] In addition, this embodiment uses a code package (Get_potential_fusionSequenceAndGene.py) to further screen the obtained abnormal alignment sequence to obtain candidate fusion sequence fragments and their genes, wherein the screening principle is: (1) low-quality screening is performed according to the parameter MAPQ in the bam file. When MAPQ is less than 20, it indicates that the alignment quality is poor, and the abnormal alignment sequence is deleted; (2) binary conversion is performed according to the parameter FLAG value in the bam file and it is determined whether the eighth bit is 1. If it is 1, it is considered that the sequence belongs to Duplicate, and the abnormal alignment sequence is deleted; (3) according to the bam file, if the aligned sequence contains soft shearing, that is, CIGAR contains the "S" string, and it is determined whether the spanning region is greater than the spanning threshold. If it is less than the spanning threshold, it is considered that the sequence is sheared due to sequencing errors and the like, and the abnormal alignment sequence is deleted; in this embodiment, the spanning threshold is set to 1 kbp, which can be adjusted according to actual conditions.

[0049] Step S4: Counting fusion information based on each candidate fusion gene;

[0050] In this embodiment, the selected candidate fusion genes are statistically analyzed, and the specific steps are as follows: according to the obtained candidate fusion genes, all sequencing data aligned to the fusion gene are screened from the bam file, and low-quality sequence information is filtered out to obtain the final data information of the fusion gene, and statistics are performed. The statistical information includes: the average sequencing depth of the candidate genes, gene coverage, number of fusion breakpoints, sequencing depth corresponding to the fusion breakpoints, number of complete matches, and number of abnormal alignments.

[0051] Among them, the average sequencing depth of the candidate gene is the total sequencing depth of the candidate gene fragment divided by the base length of the candidate gene fragment; the gene coverage is the proportion of coverage on the genomic region; the number of fusion breakpoints refers to the number of breakpoint information contained in the abnormal alignment sequences detected for the gene, and the information includes the corresponding genes and their breakpoint numbers, which are represented by an array: <corresponding genes, number of breakpoints>, which are counted according to the candidate fusion sequence fragments; the sequencing depth of the corresponding fusion breakpoint refers to the sequencing depth of the breakpoint position of each corresponding gene; the number of complete matches refers to the number of sequencing sequences that do not belong to the abnormal alignment sequence and are aligned to the gene; the number of abnormal alignments is the number of abnormal alignment sequences aligned to the fusion gene.

[0052] Step S5: Perform fusion screening based on the fusion information to obtain potential fusion gene pairs.

[0053] Preferably, in step S5, the average sequencing depth and average coverage of the entire sequencing data are first counted, and the fusion genes are preliminarily filtered according to the average sequencing depth and average coverage of the entire sequencing data: when the gene coverage of the candidate gene is less than the average coverage of the entire sequencing data, or when the average sequencing depth of the candidate gene is less than the average sequencing depth of the entire sequencing data, the candidate gene is removed; when the gene coverage and average sequencing depth of the candidate gene are both greater than the gene coverage and average coverage of the entire sequencing data, it is considered that the candidate fusion gene meets the reportable standard and can enter the candidate gene pair screening step.

[0054] The screening rules for candidate gene pairs are as follows: Figure 2 As shown, the specific steps are as follows:

[0055] (1) Obtain the average sequencing depth, gene coverage, number of fusion breakpoints, sequencing depth of corresponding fusion breakpoints, number of complete matches, and number of abnormal alignments of candidate fusion gene A;

[0056] (2) Perform a t-test based on the breakpoint number of the fusion candidate gene A to determine whether there is a significant difference. If there is a significant difference, the gene ranked first according to the fusion breakpoint number is selected as the suspected candidate gene B. If there is no significant difference, the top three gene loci are selected as suspected candidate genes B;

[0057] (3) Obtaining the fusion breakpoint number parameter information of candidate gene B and sorting them. If the fusion breakpoint number of candidate gene B with candidate gene A is higher than the breakpoint threshold, gene A and gene B are selected as the candidate gene pair;

[0058] (4) If the fusion breakpoint number ranking of candidate gene B with candidate gene A is lower than the breakpoint threshold, and the breakpoint number of candidate gene B is significantly different after the t-test, then candidate gene A and candidate gene B are filtered as a fusion gene pair;

[0059] (5) If the fusion breakpoint number ranking of candidate gene B with candidate gene A is lower than the breakpoint threshold, and there is no significant difference in the breakpoint number of candidate gene B after the t-test, then candidate gene A and candidate gene B are selected as the fusion gene pair;

[0060] (6) If the candidate gene B does not have the required number of fusion breakpoints, the sequencing depth and gene coverage of the gene are determined. If the sequencing depth and gene coverage meet the reporting criteria, the candidate gene A and the candidate gene B are not considered to be a fusion gene pair;

[0061] (7) If there is no fusion breakpoint in candidate gene B, the sequencing depth and gene coverage of the gene are determined. If the sequencing depth and gene coverage do not meet the reporting standard, it is considered that the fusion breakpoints may not be detected due to incomplete sequencing, and candidate gene A and candidate gene B are considered to be a fusion gene pair.

[0062] Among them, when the P value <0.05, it indicates that there is a significant difference in the number of breakpoints, and when the P value ≥0.05, it indicates that there is no significant difference in the number of breakpoints.

[0063] In this embodiment, the breakpoint threshold is that the breakpoint number of the gene ranks 60% of the total number of fusion breakpoints. For example, if the fusion breakpoint numbers of 30 related genes are detected in fusion gene A, it is considered that the related genes starting from the 19th ranking are all below the breakpoint threshold.

[0064] Step S6: Calculate the fusion confidence of the fusion gene pair.

[0065] For the fusion gene judged as a fusion gene pair, the fusion confidence of the fusion gene is calculated. The specific calculation formula is as follows:

[0066]

[0067] The fusion confidences of fusion gene A and fusion gene B are calculated respectively, and the average fusion confidence is calculated according to the fusion confidences to obtain the fusion confidence of the fusion gene pair.

[0068] Step S7: Determine whether it is a true positive fusion gene based on the fusion confidence and output the result.

[0069] In this embodiment, the confidence judgment rule of the fusion gene in step S7 is: if the fusion confidence is ≥1, the fusion gene is considered to be a real fusion gene; if 1<fusion confidence≤0.5, the fusion gene is considered to be a potential fusion gene; if the confidence is <0.5, the fusion gene is considered to be a false fusion gene.

[0070] In this embodiment, the samples are Figure 1The fusion gene detection was performed according to the process shown in the figure. The detection results included the fusion gene name, breakpoint location, fusion depth, fusion type, and fusion confidence, as shown in Table 1:

[0071] Table 1: Fusion gene detection results:

[0072]

[0073] It can be seen that the ALK-EML4 fusion gene was detected in sample S1 with a confidence level of 1.24, indicating that the fusion gene is a true fusion gene and the fusion type of the fusion gene is inversion; the ROS1-CD74 fusion gene was detected in sample S2 with a confidence level of 0.65, indicating that the fusion gene is a potential fusion gene and the fusion type of the fusion gene is translocation; no credible fusion gene was detected in sample S3.

Claims

1. A method for detecting fusion genes in high-throughput sequencing data, characterized in that the steps include: Step S1: obtaining high-throughput sequencing data, performing quality control filtering on the sequencing data, and obtaining filtered sequencing data; Step S2: aligning the filtered sequencing data with the reference genome to obtain an alignment file; Step S3: Screening abnormal alignment sequences according to the alignment file, and matching the abnormal alignment sequences to obtain candidate fusion sequence fragments and candidate fusion genes; Step S4: Counting fusion information based on each candidate fusion gene; Step S5: Screening fusion events according to the fusion information to obtain potential fusion gene pairs; Step S6: calculating the fusion confidence of the fusion gene pair; Step S7: Determine whether it is a true positive fusion gene based on the fusion confidence and output the result.

2. The method for detecting fusion genes in high-throughput sequencing data according to claim 1, characterized in that: The situations of abnormal alignment sequences described in step S3 include: situation 1: in paired-end sequencing, part of the read length is aligned to position 1 of the reference genome, and the other part is aligned to position 2 of the reference genome; situation 2: in paired-end sequencing, read length 1 is aligned to the reference genome, and read length 2 is aligned to other reference genomes; situation 3: in paired-end sequencing, read length 1 and read length 2 are aligned to the same reference genome respectively, but the alignment directions of the read length pairs are the same.

3. The method for detecting fusion genes in high-throughput sequencing data according to claim 1, characterized in that: The screening principle described in step S3 is: (1) If the quality of the read pair in the double-end sequencing does not meet the quality control requirements, the abnormal alignment sequence is deleted; (2) If at least one of the read pairs in the double-end sequencing belongs to Duplicates, the abnormal alignment sequence is deleted; (3) If the read pair in the double-end sequencing belongs to case 1 of the abnormal alignment sequence, the abnormal alignment sequence information with a shorter alignment part is deleted; (4) If the read pairs of the double-end sequencing are aligned to different positions of the reference gene respectively, if the position distance is less than the filtering threshold 1, the abnormal alignment sequence information is deleted.

4. The method for detecting fusion genes in high-throughput sequencing data according to claim 1, characterized in that: The candidate fusion gene statistics step described in step S4 is: selecting all sequencing data aligned to the candidate fusion gene according to the alignment file, and performing statistics based on the sequencing data on the candidate fusion gene; The statistical fusion information described in step S4 includes: the average sequencing depth of the candidate genes, gene coverage, the number of fusion breakpoints, the sequencing depth of the corresponding fusion breakpoints, the number of complete matches and the number of abnormal alignments.

5. The method for detecting fusion genes in high-throughput sequencing data according to claim 1, characterized in that: In step S5, when the gene coverage or average sequencing depth of the candidate gene is lower than the gene coverage threshold 1 and the average sequencing depth threshold 1, it is considered that there is a false positive in the sequencing of the candidate gene, and the candidate fusion gene is removed; when the gene coverage and the average sequencing depth of the candidate gene are both higher than the gene coverage threshold 1 and the average sequencing depth threshold 1, it is considered that the candidate fusion candidate gene meets the reportable standard and can enter the candidate gene pair screening step; wherein, the gene coverage threshold 1 is the average coverage of all genes, and the average sequencing depth threshold 1 is the average sequencing depth of all genes.

6. The method for detecting fusion genes in high-throughput sequencing data according to claim 1, characterized in that: The fusion event screening rule described in step S5 is: (1) Obtain the average sequencing depth, gene coverage, number of fusion breakpoints, sequencing depth of corresponding fusion breakpoints, number of complete matches, and number of abnormal alignments of fusion candidate gene A; (2) Perform a t-test based on the breakpoint number of the fusion candidate gene A to determine whether there is a significant difference. If there is a significant difference, the gene ranked first according to the fusion breakpoint number is selected as the suspected candidate gene B. If there is no significant difference, the top three gene loci are selected as suspected candidate genes B; (3) Obtaining the fusion breakpoint number parameter information of candidate gene B and sorting them. If the fusion breakpoint number of candidate gene B with candidate gene A is higher than the breakpoint threshold, candidate gene A and candidate gene B are selected as a candidate gene pair; (4) If the fusion breakpoint number ranking of candidate gene B with candidate gene A is lower than the breakpoint threshold, and the breakpoint number of candidate gene B is significantly different after the t-test, then candidate gene A and candidate gene B are filtered as a fusion gene pair; (5) If the ranking of the number of fusion breakpoints between candidate gene B and candidate gene A is lower than the breakpoint threshold, and there is no significant difference in the number of breakpoints in candidate gene B after the t-test, then candidate gene A and candidate gene B are selected as the fusion gene pair; (6) If the candidate gene B does not have the required number of fusion breakpoints, the sequencing depth and gene coverage of the gene are determined. If the sequencing depth and gene coverage meet the reporting criteria, the candidate gene A and the candidate gene B are not considered to be a fusion gene pair; (7) If there is no fusion breakpoint in candidate gene B, the sequencing depth and gene coverage of the gene are determined. If the sequencing depth and gene coverage do not meet the reporting standard, it is considered that the fusion breakpoints may not be detected due to incomplete sequencing, and candidate gene A and candidate gene B are considered to be a fusion gene pair.

7. The method for detecting fusion genes in high-throughput sequencing data according to claim 1, characterized in that: The fusion confidence described in step S6 is calculated according to the fusion confidence calculation formula, which is: 。 8. The method for detecting fusion genes in high-throughput sequencing data according to claim 1, characterized in that: The fusion confidence judgment rule described in step S7 is: if the fusion confidence ≥ confidence threshold 1, the fusion gene pair is considered to be a true fusion breakpoint; if confidence threshold 1 < fusion confidence ≤ confidence threshold 2, the fusion gene is considered to be a potential fusion breakpoint; if confidence < confidence threshold 3, the fusion gene is considered to be a false fusion breakpoint.

Images