Method, device and storage medium for noise reduction of single-cell immune repertoire sequencing data
By barcode and UMI correction of the sequencing data of single-cell immune library and filtering the background sequence according to preset filter conditions, the problem of false positive cells and low pairing rates in existing tools is solved, improving data quality and analysis accuracy.
Patent Information
- Application Number
- CN202410947442.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-16
- Publication Date
- 2024-11-29
- Estimated Expiration
- 2044-07-16
AI Technical Summary
The existing single-cell immune library sequencing data analysis tools have a narrow scope of application, and there is a problem of low pairing rates of false positive cells and heavy and light chains, especially the failure to effectively remove background sequences or contaminated sequences in VDJ library enriched fragments, resulting in Production of false positive cells.
Provide a noise reduction method for single-cell immune library sequencing data. After barcode correction and UMI correction, VDJ-related reads are extracted, and the background sequence or contaminated sequence is filtered out according to preset filter conditions, retaining the heavy and heavy VDJs. barcode of chain genes, thereby improving data quality and pairing rate of heavy and light chains.
Effectively removing false positive cells and low pairing rate problems has improved the accuracy of single-cell immune library analysis results and heavy and light chain pairing rate, providing a more reliable and accurate data basis.
Abstract
Description
Technical Field
[0001] The present invention belongs to the field of bioinformatics, and in particular relates to a noise reduction method, device and storage medium for single-cell immune repertoire sequencing data. Background Art
[0002] In recent years, with the continuous optimization of single-cell transcriptomics and immune group sequencing technologies (Aldridge, S.&Teichmann, SA Single cell transcriptomics comes of age. Nat Commun 11,4307, doi:10.1038 / s41467-020-18158-5(2020); Vandereyken, K., Sifrim, A.,Thienpont, B.&Voet, T. Methods and applications for single-cell and spatialmulti-omics. Nat Rev Genet 24, 494-515, doi:10.1038 / s41576-023-00580-2(2023).), the sequencing cost has gradually decreased, and the technology used to study the V(D)J gene recombination of single cells in the immune system, single-cell V(D)J analysis, has begun to penetrate into all aspects of basic life research, such as tumorigenesis (Barthel, S.,Falcomata, C., Rad, R., Theis, FJ&Saur, D. Single-cell profiling to explore pancreatic cancer heterogeneity, plasticity and response to therapy. Nat Cancer 4, 454-467, doi:10.1038 / s43018-023-00526-x (2023).), disease diagnosis (Miranda, AMA et al. Single-cell transcriptomics for the assessment of cardiac disease. Nat Rev Cardiol 20, 289-308, doi:10.1038 / s41569-022-00805-7(2023).), etc. V(D)J gene recombination refers to the combination of V, D, and J gene fragments in immune cells to form variable regions of immunoglobulins or T cell receptors through recombination. This variable region determines whether immune cells can recognize and respond to specific antigens.
[0003] Single-cell transcriptome sequencing based on microfluidics can utilize micro-scale pipelines and control systems to efficiently and accurately operate and analyze single cells. For example, BD Rhapsody™ technology uses card-type chips to achieve single-cell capture and molecular labeling of mRNA transcripts on magnetic oligonucleotide barcode-labeled microspheres, and then merges these microspheres into a single tube for cDNA amplification and library construction, which can meet the automatic sorting, amplification and library construction of 100 to 10,000 cells. In addition, the 10x Genomics Chromium platform based on the principle of oil-in-water micro-reaction system is also widely used to capture, dissolve, react and sequence single cells to achieve transcriptome analysis of single cells. In terms of workflow, the 10xGenomics Chromium platform first uses a microfluidic chip to capture individual cells, ensuring that each cell is separated and fixed in a specific position, then lyses each single cell, extracts RNA, and synthesizes cDNA. The synthesized cDNA is amplified and built into a library, and barcode tags and unique molecular identifier (UMI) sequences are added. Then, high-throughput sequencing is performed and the sequencing data is analyzed, including gene expression levels, cell type identification, and population structure analysis. Among them, the barcode tag is a short sequence used to distinguish RNA fragments from different cells. Each cell is accompanied by a unique barcode tag when the library is built, so that the RNA fragments can be accurately attributed to the corresponding cells in the sequencing data to avoid confusion and crossover. Each barcode tag is also connected to a molecular identifier UMI composed of a nucleotide sequence, which is used to correct the deviations and duplications introduced by PCR amplification. Each RNA molecule is accompanied by a specific UMI sequence when the library is built. The UMI can distinguish between the repeated fragments generated by PCR amplification and the real RNA molecules, thereby improving the accuracy and reliability of the data. Therefore, in the sequencing data, based on the barcode label and UMI tag, it can be determined whether each sequence in the result comes from the same cell and the same mRNA. This method can reduce the preference of PCR on different molecules. By matching and counting barcodes and UMIs, gene expression information is summarized in the counting matrix, thereby obtaining the transcriptome expression profile of a single cell.
[0004] For the bioinformatics analysis of single-cell immune group sequencing data, researchers first focus on the CDR3 domain sequence at the single-cell level. The enrichment efficiency of the immune group library data, the fragment size of the VDJ heavy and light chains, and the ratio of heavy and light chain-related gene data are directly related to the accuracy of the analysis results. For example, the library constructed by Drop-Seq has a barcode sequence, a UMI sequence, and Poly T at the 5' end of read 1, and the efficiency of single-cell library construction is greatly improved through microfluidics technology. The 10x Genomics single-cell 5' RNA library construction method introduces the TSO sequence after the UMI sequence on the basis of 3' RNA, eliminating the preference for 3' RNA library construction (refer to the 10x Genomics official website). The VDJ library construction is based on 5' RNA and uses nested PCR amplification to obtain the sequence of the above structure, which contains all VDJ genes and part of the C gene.
[0005] According to different library structures and their characteristics, researchers have developed multiple analysis tools and built different analysis processes, such as MiXCR5, BD Rhapsody™ Analysis Pipeline, TRUST4, Cell Ranger, etc. Among them, MiXCR and CellRanger are tools for analyzing single-cell immune repertoire sequencing data, which can directly output potential immune cells and VDJ heavy and light chain information in cells. The disadvantage is that if the quality of the added library is poor or the enrichment efficiency is low, more false positive cells will be obtained, and the proportion of cells with heavy and light chain pairing is also low. BD Rhapsody™ Analysis Pipeline can process BD Rhapsody™ Whole Transcriptome Data, determine the T cells and B cells to be analyzed based on the cell results of the transcriptome, and cannot directly analyze the individual VDJ library data, so the usage scenarios are limited. TRUST4 removes low-complexity sequences from the data, then enriches the reads that can be aligned to the VDJ reference sequence in the transcriptome data, and then performs VDJ transcript splicing to obtain VDJ heavy and light chain information. The disadvantage is that the quality of the spliced transcripts is low, and the quality of transcripts and annotations in the analysis results of low-quality libraries is also low, so its scope of application is also limited.
[0006] Specifically, Cell Ranger, which is mainly adapted for the analysis of 10x single-cell immune group data, has a whole technical solution covering the whole process from data download, filtering, alignment, VDJ gene splicing, VDJ heavy and light chain pairing, and clonal type identification. The input file of Cell Ranger is a fastq format file obtained by second-generation sequencing. The technology will automatically determine whether the reads obtained by sequencing contain a valid barcode sequence, that is, whether the 16 consecutive bases at the 5' end of the read 1 obtained by sequencing exist in the barcode whitelist (the barcode whitelist is matched with the 10x 5' transcriptome / VDJ kit). For reads containing valid barcodes, potential VDJ gene reads are filtered by matching the amplification primers of VDJ, and then the number of UMIs of the sequencing reads corresponding to each barcode sequence is evaluated to match the VDJ gene. The UMI will be corrected by the correction algorithm first, and contig splicing and annotation will be performed based on the reads in each barcode, and the contig will be annotated to the reference sequence of VDJ. The effective cell identification is performed according to the annotation results and the number of readsde1UMI. Finally, the annotation information of the light and heavy chains in the effective cells and the pairing information are used for clonal identification, and finally the annotation information and clonal identification information of the VDJ gene transcripts are obtained. Subsequently, the html report and vloupe visualization file are generated.
[0007] However, Cell Ranger is currently only applicable to single-cell immune repertoire sequencing data from the 10x Genomics single-cell platform, and no noise reduction is performed on the VDJ library enrichment fragments, which results in a certain degree of false positive cells in the VDJ library construction data of some samples, and there is also a problem of low matching rate. Specifically, it is because the CellRanger method does not remove background sequences or contamination sequences in the VDJ library enrichment fragments, so that the background sequences or contamination sequences are mistakenly regarded as VDJ gene fragments from cells, resulting in the generation of false positive cells. Among them, the contamination sequence is generally a non-VDJ gene sequence amplified by the VDJ primers used in the experimental system.
[0008] Therefore, there is an urgent need for a noise reduction method that can improve the quality of VDJ library analysis results and support separate VDJ library data so that the analysis results can achieve both accuracy in cell identification and a higher level of pairing rate for heavy and light chains. Summary of the invention
[0009] In order to solve the problems of narrow application scope, false positive cells and low pairing rate of heavy and light chains in single-cell immune group library data analysis tools in the prior art, the present invention provides a denoising method, device and storage medium for single-cell immune group library sequencing data, which can perform quality analysis such as correction, alignment and quantification on single-cell immune cell receptor gene data in single-cell immune group library sequencing data of any single-cell platform of various tissue type samples, and filter all barcodes in VDJ sequencing data according to the number of UMIs of VDJ gene-related reads extracted under each barcode and the number of UMIs of total reads, remove background sequences or contamination sequences that may exist in VDJ sequencing reads, and solve the problems of false positive cells and low pairing rate.
[0010] The technical solution adopted by the present invention is: a method for denoising single-cell immune repertoire sequencing data, comprising the following steps:
[0011] S1. extract reads from VDJ sequencing data, perform barcode correction and UMI correction on the reads in sequence to obtain corrected reads, and extract VDJ_reads mapped to the VDJ gene from the corrected reads; wherein the VDJ_reads include VDJ_light_reads mapped to the VDJ light chain gene, and / or VDJ_heavy_reads mapped to the VDJ heavy chain gene;
[0012] S2. Filter the barcodes in the reads according to the preset filtering conditions to obtain the retained barcodes; wherein the preset filtering conditions are set according to the number of UMIs of the corrected reads under each barcode, the number of UMIs of VDJ_light_reads, and the number of UMIs of VDJ_heavy_reads;
[0013] S3. Re-extracting reads containing the retained barcode from the VDJ sequencing data to obtain retained reads; wherein the retained reads are used for single-cell immune repertoire analysis.
[0014] In the present invention, in order to remove the background sequences or contamination sequences that may exist in the VDJ sequencing reads, the corrected reads after barcode correction and UMI correction are first compared with the gene annotation file of the species corresponding to the sample, so as to extract the VDJ_light_reads and / or VDJ_heavy_reads that are related to the VDJ light and heavy chain genes, and then the UMI number of the reads related to the VDJ gene extracted under each barcode and the total UMI number of the corrected reads are counted. It can be understood that under each barcode, the sum of the UMI number of VDJ_light_reads and the UMI number of VDJ_heavy_reads should be equal to the UMI number of VDJ_reads, and the UMI number of VDJ_reads should be less than or equal to the UMI number of the corrected reads.
[0015] After many experiments, the present invention proposes to set preset filtering conditions for filtering barcodes according to the statistical results of the number of UMIs of the correction reads, VDJ_reads, VDJ_light_reads, and VDJ_heavy_reads under each barcode. Specifically, the barcode containing more VDJ light and heavy chain genes is retained, that is, the number of UMIs of VDJ_reads, VDJ_light_reads, and VDJ_heavy_reads under this barcode accounts for a high proportion, while the number of UMIs of sequences unrelated to the VDJ gene accounts for a low proportion, thereby filtering out invalid barcodes with more background sequences or contamination sequences, reducing the impact of poor quality cells, false positive cells, etc. contained in the reaction system on subsequent data analysis. At the same time, quantitative analysis of the number of UMIs of VDJ_light_reads and VDJ_heavy_reads under each barcode can also exclude single droplet multiple cells in the reaction system. After filtering, B cells or T cells with retained barcodes will be regarded as potential immune cells for noise reduction or background removal, and their reads containing retained barcodes will be further used for subsequent single-cell immune repertoire analysis.
[0016] On this basis, the present invention provides a better preset filtering condition. The preset filtering condition is: the barcode that meets any of the following conditions is retained:
[0017] Condition 1: 25≤the number of UMIs of the corrected reads≤100000, and the number of UMIs of VDJ_heavy_reads≥5, and the proportion of the number of UMIs of VDJ_heavy_reads≥5%, and the number of UMIs of VDJ_light_reads≥5, and the proportion of the number of UMIs of VDJ_light_reads≥10%;
[0018] Condition 2: The UMI number of VDJ_heavy_reads accounts for ≥ 40%, and the UMI number of VDJ_light_reads accounts for ≥ 40%.
[0019] From the parameter settings of condition 1 and condition 2, it can be seen that the purpose of the preset filtering condition is to filter out barcodes with more VDJ light and heavy chain genes, exclude invalid barcodes with more background sequences or contamination sequences, so as to achieve the purpose of noise reduction, which can effectively solve the problems of false positive cells and low pairing rates, and also reduce the impact of poor quality cells and single droplet multi-cells contained in the reaction system on subsequent data analysis. It can be understood that based on the above-mentioned preferred preset filtering condition, those skilled in the art can also adjust the parameter range of the UMI number and the proportion of the UMI number of corrected reads, VDJ_reads, VDJ_light_reads, and VDJ_heavy_reads under each barcode in the preset filtering condition to adapt to the single-cell immune repertoire analysis of VDJ sequencing data of samples from different tissue sources. Among them, the proportion of UMI number of VDJ_heavy_reads should be understood as the proportion of the UMI number of VDJ_heavy_reads to the UMI number of correction reads; the proportion of UMI number of VDJ_light_reads should be understood as the proportion of the UMI number of VDJ_light_reads to the UMI number of correction reads.
[0020] Specifically, in step S1, no less than 1 million pairs of reads are extracted from the VDJ sequencing data. To ensure better noise reduction effect, the method provided by the present invention should analyze and filter no less than 1 million pairs of reads. To further improve the noise reduction effect and maintain a reasonable processing speed, it is preferably to extract 1 million to 10 million pairs of reads, and more preferably 8 million pairs of reads. As is well known to those skilled in the art, each pair of reads includes read1 and read2, and both read1 and read2 contain a barcode sequence and a UMI sequence, which are used to determine the source of each read in the sequencing results at the cellular level and the mRNA level, respectively. Among them, the VDJ sequencing data can be VDJ sequencing data from any single-cell platform, and the VDJ sequencing data generally comes from the VDJ library construction kit to build a library and sequence the sequencing file.
[0021] Specifically, in step S1, the method of sequentially performing barcode correction and UMI correction on reads includes: retaining valid reads containing valid barcodes after performing barcode correction on reads; aligning read2 in the obtained valid reads to the genome of the species corresponding to the sample to obtain a comparison file, wherein the comparison file contains the position information of the valid reads on the genome; sorting the obtained comparison file according to the position information of the valid reads on the genome and their barcodes; and performing UMI correction on the valid reads located in the same gene and having the same barcode according to the sorting result to obtain corrected reads. Among them, read2 in the valid reads is selected for comparison because read2 in the single-cell immune group library sequencing data has more information than read1, which improves the accuracy of the comparison. The obtained comparison file should contain the following information: the position information of the valid reads on the genome, the barcode sequence of the valid reads, the original read name of the valid reads in the VDJ sequencing data, and the UMI sequence of the valid reads. More specifically, bowtie2 can be used to align read2 in the valid reads to the genome of the species corresponding to the sample to obtain the comparison file bam. To facilitate the subsequent sorting of the comparison file bam, it is preferred to sort the information in the comparison file bam in the following format:
[0022] The location information of valid reads on the genome - the barcode sequence of valid reads - the original read name of valid reads in VDJ sequencing data - the UMI sequence of valid reads.
[0023] Therefore, for example, the comparison file bam after format adjustment may have the following format:
[0024] chr12_000113271463-ACCTTCGCGCGCCGTTAGTT-A00456:1122:HHJCFDSX5:1:1101:22462:1031-CTAGGAATCG.
[0025] Among them, this format is divided into four fields. The first segment chr12_000113271463 is the location of the valid reads mapped to the genome, the second segment ACCTTCGCGCGCCGTTAGTT is the valid reads with a barcode sequence length of 20 bp, the third segment A00456:1122:HHJCFDSX5:1:1101:22462:1031 is the read name of the valid reads in the VDJ sequencing data fastq file, and the fourth segment CTAGGAATCG is the valid reads with a UMI sequence length of 10 bp.
[0026] More specifically, the UMI correction described in step S1 is performed based on Hamming Distance. When the Hamming Distance between two UMIs is ≤1, the UMI with fewer valid reads is corrected to a UMI with more valid reads. Hamming Distance represents the number of different characters at corresponding positions of two strings of equal length. We can represent the Hamming Distance between strings x and y by d(x,y). It can be understood that the purpose of UMI correction based on Hamming Distance is to correct UMI mismatches generated during PCR or sequencing, thereby retaining more valid reads under the same UMI and improving accuracy. It can also be understood that when UMI correction is performed based on Hamming Distance, the correction threshold of Hamming Distance can be adjusted according to the length of the UMI sequence. For example, for UMIs with short sequence lengths, such as 10 bp, setting Hamming Distance≤1 can achieve better correction accuracy, that is, the number of different characters at corresponding positions of two UMIs should not be greater than 1.
[0027] Specifically, in step S1, the method of extracting VDJ_reads aligned to the VDJ gene from the correction reads includes: aligning the correction reads with the VDJ light chain gene and / or the VDJ heavy chain gene in the gene annotation file of the species corresponding to the sample, and extracting VDJ_light_reads aligned to the VDJ light chain gene and / or VDJ_heavy_reads aligned to the VDJ heavy chain gene. It can be understood that the VDJ_reads include VDJ_light_reads and VDJ_heavy_reads.
[0028] More specifically, the denoising method for single-cell immune repertoire sequencing data comprises the following steps:
[0029] S1-1. Extract no less than 1 million read pairs from VDJ sequencing data, each pair of reads includes read1 and read2, and both read1 and read2 contain barcode sequences and UMI sequences;
[0030] S1-2. Obtain the gene annotation file of the species corresponding to the sample and extract the light chain information and the light and heavy chain information from it;
[0031] S1-3. Perform barcode correction on the reads extracted in S1-2 and retain valid reads containing valid barcodes;
[0032] S1-4. Use bowtie2 to align read2 in the valid reads obtained in S1-3 to the genome of the species to obtain an alignment file containing the position information of the valid reads on the genome;
[0033] S1-5. Sort the alignment files obtained in S1-4 according to the genome location information and barcodes of the valid reads;
[0034] S1-6. According to the sequencing results obtained in S1-5, the valid reads located in the same gene and having the same barcode are subjected to UMI correction based on Hamming Distance to obtain corrected reads;
[0035] S1-7. Compare the corrected reads obtained in S1-6 with the light chain information and light and heavy chain information obtained in S1-2, and extract VDJ_light_reads that are aligned to the VDJ light chain gene and / or VDJ_heavy_reads that are aligned to the VDJ heavy chain gene;
[0036] S1-8. According to the comparison results of S1-7, the quantitative information of the number of UMIs of the corrected reads, VDJ_light_reads, and VDJ_heavy_reads under each barcode is counted; that is, the number of UMIs of the corrected reads, VDJ_light_reads, and VDJ_heavy_reads with the same barcode is counted;
[0037] S2. Filter the barcodes in the reads according to the preset filtering conditions to obtain the retained barcodes; the preset filtering conditions are: the barcodes that meet any of the following conditions are retained:
[0038] Condition 1: 25≤the number of UMIs of the corrected reads≤100000, and the number of UMIs of VDJ_heavy_reads≥5, and the proportion of the number of UMIs of VDJ_heavy_reads≥5%, and the number of UMIs of VDJ_light_reads≥5, and the proportion of the number of UMIs of VDJ_light_reads≥10%;
[0039] Condition 2: The UMI number of VDJ_heavy_reads accounts for ≥ 40%, and the UMI number of VDJ_light_reads accounts for ≥ 40%;
[0040] S3. Re-extracting reads containing the retained barcode from the VDJ sequencing data to obtain retained reads; wherein the retained reads are used for single-cell immune repertoire analysis.
[0041] The present invention also provides a denoising device for single-cell immune repertoire sequencing data, the device comprising:
[0042] A preprocessing module, used to perform barcode correction and UMI correction on the reads extracted from the VDJ sequencing data in sequence to obtain corrected reads, and extract VDJ_reads aligned to the VDJ gene from the corrected reads; wherein the VDJ_reads include VDJ_light_reads aligned to the VDJ light chain gene, and / or VDJ_heavy_reads aligned to the VDJ heavy chain gene;
[0043] A filtering module is used to filter the barcodes in the reads according to a preset filtering condition to obtain the retained barcodes; wherein the preset filtering condition is set according to the number of UMIs of the corrected reads, the number of UMIs of VDJ_light_reads, and the number of UMIs of VDJ_heavy_reads under each barcode;
[0044] The analysis module is used to re-extract reads containing the retained barcode from the VDJ sequencing data to obtain retained reads; wherein the retained reads are used for single-cell immune repertoire analysis.
[0045] Specifically, the preprocessing module extracts no less than 1 million pairs of reads from the VDJ sequencing data, preferably 1 million to 10 million pairs of reads, and more preferably 8 million pairs of reads. Each pair of reads includes read1 and read2, and both read1 and read2 contain a barcode sequence and a UMI sequence.
[0046] Specifically, the preprocessing module includes:
[0047] A correction unit is used to perform barcode correction and UMI correction on reads in sequence and obtain corrected reads; wherein, after the reads are barcode corrected, valid reads containing valid barcodes are retained; read2 in the obtained valid reads is aligned to the genome of the species corresponding to the sample to obtain an alignment file, wherein the alignment file contains the position information of the valid reads on the genome; the obtained alignment file is sorted according to the position information of the valid reads on the genome and their barcodes; according to the sorting results, UMI correction is performed on the valid reads located in the same gene and having the same barcode to obtain corrected reads;
[0048] The extraction unit is used to compare the corrected reads with the VDJ light chain gene and / or VDJ heavy chain gene in the gene annotation file of the species corresponding to the sample, and extract VDJ_light_reads that are aligned to the VDJ light chain gene and / or VDJ_heavy_reads that are aligned to the VDJ heavy chain gene.
[0049] More specifically, the alignment file obtained in the correction unit should contain the following information: location information of valid reads on the genome, barcode sequences of valid reads, original read names of valid reads in VDJ sequencing data, and UMI sequences of valid reads.
[0050] More specifically, the UMI correction in the correction unit is performed based on the Hamming Distance. When the Hamming Distance between two UMIs is ≤1, the UMI with fewer valid reads is corrected to a UMI with more valid reads.
[0051] Specifically, the preset filtering condition adopted by the filtering module is: the barcodes that meet any of the following conditions are retained:
[0052] Condition 1: 25≤the number of UMIs of the corrected reads≤100000, and the number of UMIs of VDJ_heavy_reads≥5, and the proportion of the number of UMIs of VDJ_heavy_reads≥5%, and the number of UMIs of VDJ_light_reads≥5, and the proportion of the number of UMIs of VDJ_light_reads≥10%;
[0053] Condition 2: The UMI number of VDJ_heavy_reads accounts for ≥ 40%, and the UMI number of VDJ_light_reads accounts for ≥ 40%;
[0054] Among them, the proportion of UMI number of VDJ_heavy_reads should be understood as the proportion of the UMI number of VDJ_heavy_reads to the UMI number of correction reads; the proportion of UMI number of VDJ_light_reads should be understood as the proportion of the UMI number of VDJ_light_reads to the UMI number of correction reads.
[0055] The present invention also provides a terminal device, including a memory and a processor, wherein the memory stores a computer program that can be run on the processor, and the processor implements the steps of the above method when executing the computer program.
[0056] The present invention also provides a computer-readable storage medium on which a computer program is stored, and the computer program implements the steps of the above method when executed by a processor.
[0057] Beneficial effects of the present invention:
[0058] The present invention provides a noise reduction method, device and storage medium for single-cell immune repertoire sequencing data, which solves the problem of background sequences or contamination sequences that may exist in VDJ sequencing reads by correcting, aligning, quantitatively analyzing and filtering VDJ sequencing data, and effectively removes false positive cells and low pairing rates. The present invention sets preset filtering conditions according to the statistical results of the number of UMIs of corrected reads, VDJ_light_reads, and VDJ_heavy_reads under each barcode and the proportion of the number of UMIs, retains barcodes containing more VDJ light and heavy chain gene information, excludes background sequences or contamination sequences, reduces the impact of poor quality cells, false positive cells, and single droplet multi-cells on data analysis, achieves the purpose of improving data quality and reducing interference factors, effectively solves the problem of false positive cells and low pairing rates, and provides a more reliable and accurate data basis for single-cell immune repertoire analysis. DETAILED DESCRIPTION
[0059] The following is an explanation of the embodiments of the present invention by specific specific examples, and those skilled in the art can easily understand other advantages and effects of the present invention from the contents disclosed in this specification. The present invention can also be implemented or applied by other different specific embodiments, and the details in this specification can also be modified or changed in various ways based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that, in the absence of conflict, the features in the following embodiments and embodiments can be combined with each other. In the embodiments of the present invention, the methods used are all conventional methods unless otherwise specified, and the reagents used can be obtained from commercial sources.
[0060] Embodiment 1:
[0061] A sample from a mouse was constructed and sequenced using the VDJ library construction kit to obtain single-cell immune group library sequencing data Sample01 of BCR. Sample01 was subjected to noise reduction processing using the noise reduction method for single-cell immune group library sequencing data provided by the present invention, specifically comprising the following steps. The following implementation steps are all performed under the Linux operating system.
[0062] S1-1. Extract 8,000,000 reads from the raw data of sample01, where sample01_R1.fastq.gz and sample01_R2.fastq.gz are the read1 (R1) and read2 (R2) data files of the raw data:
[0063] zcat sample01_R1.fastq.gz | head -n 32000000>Tiny_VDJ_data_R1.fastq;
[0064] zcat sample01_R2.fastq.gz | head -n 32000000>Tiny_VDJ_data_R2.fastq.
[0065] S1-2. Since sample01 is a single-cell VDJ sequencing data from mice, it is necessary to obtain the mouse gene annotation file for subsequent comparison. Download the mouse gene annotation file (https: / / ftp.ebi.ac.uk / pub / databases / gencode / Gencode_mouse / release_M35 / gencode.vM35.primary_assembly.annotation.gtf.gz) to the Linux server, and use the gunzip command to decompress it into a gtf file. Extract the light chain and light and heavy chain information in the gtf file using IG_V_gene as the key from the obtained gtf file and convert it into a bed format file:
[0066] grep "IG_V_gene" / Data / Reference / GRCm39 / genes / genes.gtf | grep -e "Igk" -e "Igl" | awk '{print$1"\t"$4"\t"$5}' | sort -k1,1V -k2,2n -k3,3n |bedtools / bedtools2 / bin / bedtools merge -i ->igv_light.bed;
[0067] grep "IG_V_gene" / Data / Reference / GRCm39 / genes / genes.gtf | awk '{print$1"\t"$4"\t"$5}' | sort -k1,1V -k2,2n -k3,3n |bedtools / bedtools2 / bin / bedtoolsmerge -i ->igv_all.bed;
[0068] Among them, igv_all.bed represents the bed file of all VDJ-related genes of mouse B cell receptor, and igv_light.bed represents the bed file of all light chain-related genes of VDJ-related genes of mouse B cell receptor. The formats of the two obtained bed files, igv_light.bed and igv_all.bed, are shown in Table 1.
[0069] Table 1. BED file format
[0070] Chromosome location Bed_Region_Start Bed_Region_End chr6 67532481 67533200 chr6 67586697 67587492 chr6 67701886 67702645 chr6 67736684 67737397 chr6 67743020 67743756 chr6 67768030 67768495 chr6 67816777 67817250 chr6 67838137 67838643 chr6 67873156 67873626 chr6 67890557 67891036 chr6 67919058 67919524 chr6 67931212 67931698 chr6 67993702 67994472 chr6 68013796 68014302 chr6 68026967 68027440 … … …
[0071] S1-3. Perform barcode correction on the reads obtained in S1-1. Extract read1 and read2 containing valid barcodes based on the position information of the barcode on read1, and save the barcode sequence and UMI information on the read name of read1 and read2.
[0072] S1-4. Use bowtie2 to align read2 in the valid reads to the genome of the species corresponding to the sample to obtain the alignment file bam.
[0073] The comparison method of Bowtie2 comparison tool is as follows:
[0074] ${bowtie2} --very-sensitive-local -I 1 -X 1000 -p 12 -x ${btIndex} -U01.converted / ${fileID} _R2.fastq -S 02.aligned / ${fileID}.sam 2>02.aligned / ${fileID}_mapStats.xls.
[0075] To facilitate the subsequent sorting of the comparison file bam, it is preferred to sort the information in the comparison file bam in the following format:
[0076] The location information of valid reads on the genome - the barcode sequence of valid reads - the original read name of valid reads in VDJ sequencing data - the UMI sequence of valid reads.
[0077] Therefore, for example, the comparison file bam after format adjustment may have the following format:
[0078] chr12_000113271463-ACCTTCGCGCGCCGTTAGTT-A00456:1122:HHJCFDSX5:1:1101:22462:1031_CTAGGAATCG.
[0079] Among them, this format is divided into four fields. The first segment chr12_000113271463 is the location of the valid reads mapped to the genome, the second segment ACCTTCGCGCGCCGTTAGTT is the valid reads with a barcode sequence length of 20 bp, the third segment A00456:1122:HHJCFDSX5:1:1101:22462:1031 is the read name of the valid reads in the VDJ sequencing data fastq file, and the fourth segment CTAGGAATCG is the valid reads with a UMI sequence length of 10 bp.
[0080] S1-5. Use samtools to sort the alignment file bam obtained in S1-4.
[0081] Samtools sorting method is as follows:
[0082] samtools sort -@2 -n 02.aligned / ${fileID}_nameReformat.sam -o03.region_count / id_sorted _R2.sam.
[0083] S1-6. Correct the UMI of the comparison file bam sorted in S1-5 based on the Hamming Distance to obtain corrected reads. When the Hamming Distance between two UMIs is ≤1, the UMI with fewer valid reads is corrected to a UMI with more valid reads.
[0084] S1-7. Use samtools to extract VDJ_light_reads mapped to the VDJ light chain gene and VDJ_heavy_reads mapped to the VDJ heavy chain gene from the UMI-corrected alignment file bam obtained in S1-6, and save them as two new bam files.
[0085] Samtools extraction method is as follows:
[0086] samtools view -L ${bed_ig_all} 03.region_count / new_umi_deduped.sam>03.region_count / igv_all_regions.sam;
[0087] samtools view -L ${bed_ig_light} 03.region_count / new_umi_deduped.sam>03.region_count / igv_light_regions.sam;
[0088] Among them, ${bed_ig_all} is the igv_all.bed of the first step, and ${bed_ig_light} is the igv_light.bed of the third step.
[0089] S1-8. Statistical analysis of the UMI quantity of the calibration reads, VDJ_light_reads, and VDJ_heavy_reads under each barcode of valid reads using a self-compiled tool named script_stat, and the output result is umi_stats.xls. The format of the obtained umi_stats.xls file is shown in Table 2, where the UMI quantity of VDJ_reads is the sum of the UMI quantities of VDJ_light_reads and VDJ_heavy_reads.
[0090] The statistical method of script_stat is as follows:
[0091] python ${script_stat} 03.region_count / new_umi_deduped.sam 03.region_count / igv_all_regions.sam 03.region_count / igv_light_regions.sam>03.region_count / umi_stats.xls.
[0092] Table 2. umi_stats.xls file format
[0093] barcode sequence UMI number of reads UMI number of VDJ_reads UMI number of VDJ_light_reads The number of UMIs for VDJ_heavy_reads CTCTACAGTGCTAGCTATAC 1539 729 421 308 CAGCAGCGGTTTGTACAATA 1774 823 504 319 GTGTGAGTGTATCCTGTATT 1 1 0 1 ATCACCATCGGACCACTATT 691 257 156 101 CCAATCAGTCTGCCTAACTT 818 399 230 169 GTGCCGTACGCTCGAGTATA 625 318 206 112 TGACGACCGCACATCGACGT 807 421 190 231 ACGCAATCTTCCTGATGCAT 1 1 0 1 CCTAGGTCATAGAACTGTGT 538 243 144 99
[0094] S2. Based on the statistical information of the UMI in Table 2, the preset filtering conditions are set to filter the barcodes in the reads, filter out the barcode information with low background signal or noise, and obtain the retained barcodes. The preset filtering conditions are: the barcodes that meet any of the following conditions are retained:
[0095] Condition 1: 25≤the number of UMIs of the corrected reads≤100000, and the number of UMIs of VDJ_heavy_reads≥5, and the proportion of the number of UMIs of VDJ_heavy_reads≥5%, and the number of UMIs of VDJ_light_reads≥5, and the proportion of the number of UMIs of VDJ_light_reads≥10%;
[0096] Condition 2: The UMI number of VDJ_heavy_reads accounts for ≥ 40%, and the UMI number of VDJ_light_reads accounts for ≥ 40%;
[0097] Among them, the proportion of UMI number of VDJ_heavy_reads should be understood as the proportion of the UMI number of VDJ_heavy_reads to the UMI number of correction reads; the proportion of UMI number of VDJ_light_reads should be understood as the proportion of the UMI number of VDJ_light_reads to the UMI number of correction reads.
[0098] S3. Re-extract reads containing the retained barcode from the VDJ sequencing data to obtain retained reads; wherein the retained reads are used for single cell immune repertoire analysis. The analysis output result file content is as follows:
[0099] ├── 01.converted
[0100] │├── corrected_test1_R1.fastq
[0101] │└── corrected_test1_R2.fastq
[0102] ├── 02.aligned
[0103] │├── corrected_test1_mapStats.xls
[0104] │├── corrected_test1_nameReformat.sam
[0105] │├── corrected_test1.sam
[0106] │├── igv_all.bed
[0107] │└── igv_light.bed
[0108] └── 03.region_count
[0109] │├── id_sorted_R2.sam
[0110] │├── igv_all_regions.sam
[0111] │├── igv_light_regions.sam
[0112] │├── new_umi_deduped.sam
[0113] │├── new_umi_deduped.sam_umi_stats.xls
[0114] │└── umi_stats.xls
[0115] Among them, filtered.bcell.txt or filtered.tcell.txt obtained by filtering umi_stats.xls obtained in S1-8 are potential immune cells for noise reduction or background removal.
[0116] Embodiment 2:
[0117] The method provided in Example 1 was used to perform noise reduction processing on the single-cell immune repertoire sequencing data Sample02 of a mouse-derived BCR.
[0118]
Performance test
[0119] The sequencing data Sample01 and Sample02 of Examples 1 and 2 were analyzed, and the results after noise reduction were verified. The single cell analysis indicators of Sample01 and Sample02 before noise reduction are shown in Table 3, and the statistics of various indicators after noise reduction are shown in Table 4.
[0120] Table 3. Single cell analysis indicators of Sample01 and Sample02 before noise reduction
[0121] Sample ID Estimated Number of Cells Mean Reads per Cell Productive pair cell Number Productive pair ratio Sample01 4415 8810 2904 65.8% Sample02 3307 11830 2490 75.3%
[0122] Table 4. Single cell analysis indicators after noise reduction of Sample01 and Sample02
[0123] Sample ID Estimated Number of Cells Mean Reads per Cell Productive pair cell Number Productive pair ratio Sample01 3490 4365 2977 85.3% Sample02 3076 6443 2739 89.0%
[0124] After noise reduction, the analysis results of the single-cell immune repertoire data show that while the number of cells decreased after noise reduction, the number of light and heavy chain paired cells increased, and the pairing rate was significantly improved.
[0125] In one embodiment, a denoising device for single-cell immune repertoire sequencing data is also provided, the device comprising a preprocessing module, a filtering module and an analysis module. The preprocessing module is used to perform barcode correction and UMI correction on the reads extracted from the VDJ sequencing data in sequence and obtain the corrected reads, and extract VDJ_reads aligned to the VDJ gene from the corrected reads; wherein the VDJ_reads include VDJ_light_reads aligned to the VDJ light chain gene, and / or VDJ_heavy_reads aligned to the VDJ heavy chain gene. The filtering module is used to filter the barcode in the reads according to the preset filtering conditions to obtain the retained barcode; wherein the preset filtering conditions are set according to the number of UMIs of the corrected reads under each barcode, the number of UMIs of the VDJ_light_reads, and the number of UMIs of the VDJ_heavy_reads. The analysis module is used to re-extract reads containing the retained barcode from the VDJ sequencing data to obtain the retained reads; wherein the retained reads are used for single-cell immune repertoire analysis.
[0126] In one embodiment, the preprocessing module extracts no less than 1 million pairs of reads from the VDJ sequencing data, preferably 1 million to 10 million pairs of reads, and more preferably 8 million pairs of reads. Each pair of reads includes read1 and read2, and both read1 and read2 contain a barcode sequence and a UMI sequence.
[0127] In one embodiment, the preprocessing module includes a correction unit and an extraction unit. The correction unit is used to perform barcode correction and UMI correction on reads in sequence and obtain corrected reads; wherein, after the reads are barcode corrected, valid reads containing valid barcodes are retained; read2 in the obtained valid reads is aligned to the genome of the species corresponding to the sample to obtain an alignment file, wherein the alignment file contains the position information of the valid reads on the genome; the alignment file obtained is sorted according to the position information of the valid reads on the genome and their barcodes; according to the sorting result, the valid reads located in the same gene and having the same barcode are UMI corrected to obtain corrected reads. The alignment file obtained in the correction unit should contain the following information: the position information of the valid reads on the genome, the barcode sequence of the valid reads, the original read name of the valid reads in the VDJ sequencing data, and the UMI sequence of the valid reads. The UMI correction in the correction unit is based on the Hamming Distance. When the Hamming Distance between two UMIs is ≤1, the UMI containing fewer valid reads is corrected to a UMI containing more valid reads. The extraction unit is used to compare the corrected reads that have been corrected by barcode and UMI in sequence with the VDJ light chain gene and / or VDJ heavy chain gene in the gene annotation file of the species corresponding to the sample, and extract VDJ_light_reads that are aligned to the VDJ light chain gene and / or VDJ_heavy_reads that are aligned to the VDJ heavy chain gene.
[0128] In one embodiment, the preset filtering condition adopted by the filtering module is: barcodes that meet any of the following conditions are retained:
[0129] Condition 1: 25≤the number of UMIs of the corrected reads≤100000, and the number of UMIs of VDJ_heavy_reads≥5, and the proportion of the number of UMIs of VDJ_heavy_reads≥5%, and the number of UMIs of VDJ_light_reads≥5, and the proportion of the number of UMIs of VDJ_light_reads≥10%;
[0130] Condition 2: The UMI number of VDJ_heavy_reads accounts for ≥ 40%, and the UMI number of VDJ_light_reads accounts for ≥ 40%;
[0131] Among them, the proportion of UMI number of VDJ_heavy_reads should be understood as the proportion of the UMI number of VDJ_heavy_reads to the UMI number of correction reads; the proportion of UMI number of VDJ_light_reads should be understood as the proportion of the UMI number of VDJ_light_reads to the UMI number of correction reads.
[0132] In one embodiment, a terminal device is also provided, including a memory and a processor, wherein a computer program that can be run on the processor is stored on the memory, and the processor implements the following steps when executing the computer program: extracting reads from VDJ sequencing data, performing barcode correction and UMI correction on the reads in sequence to obtain corrected reads, and extracting VDJ_reads aligned to the VDJ gene from the corrected reads; wherein the VDJ_reads include VDJ_light_reads aligned to the VDJ light chain gene, and / or VDJ_heavy_reads aligned to the VDJ heavy chain gene; filtering the barcodes in the reads according to preset filtering conditions to obtain retained barcodes; wherein the preset filtering conditions are set according to the number of UMIs of the corrected reads, the number of UMIs of the VDJ_light_reads, and the number of UMIs of the VDJ_heavy_reads under each barcode; re-extracting reads containing the retained barcodes from the VDJ sequencing data to obtain retained reads; wherein the retained reads are used for single-cell immune repertoire analysis.
[0133] In one embodiment, a computer-readable storage medium is also provided, on which a computer program is stored, and when the computer program is executed by a processor, the following steps are implemented: extracting reads from VDJ sequencing data, performing barcode correction and UMI correction on the reads in sequence to obtain corrected reads, and extracting VDJ_reads aligned to the VDJ gene from the corrected reads; wherein the VDJ_reads include VDJ_light_reads aligned to the VDJ light chain gene, and / or VDJ_heavy_reads aligned to the VDJ heavy chain gene; filtering the barcodes in the reads according to preset filtering conditions to obtain retained barcodes; wherein the preset filtering conditions are set according to the number of UMIs of the corrected reads, the number of UMIs of the VDJ_light_reads, and the number of UMIs of the VDJ_heavy_reads under each barcode; re-extracting reads containing the retained barcodes from the VDJ sequencing data to obtain retained reads; wherein the retained reads are used for single-cell immune repertoire analysis.
[0134] The embodiments described above are merely descriptions of preferred implementations of the present invention and are not intended to limit the scope of the present invention. Without departing from the design spirit of the present invention, various modifications and improvements made to the technical solutions of the present invention by ordinary technicians in this field should all fall within the protection scope of the present invention.
Claims
1. A method for denoising single-cell immune repertoire sequencing data, characterized in that: The steps include: S1. extract reads from VDJ sequencing data, perform barcode correction and UMI correction on the reads in sequence to obtain corrected reads, and extract VDJ_reads mapped to the VDJ gene from the corrected reads; wherein the VDJ_reads include VDJ_light_reads mapped to the VDJ light chain gene, and / or VDJ_heavy_reads mapped to the VDJ heavy chain gene; S2. Filter the barcodes in the reads according to the preset filtering conditions to obtain the retained barcodes; wherein the preset filtering conditions are set according to the number of UMIs of the corrected reads under each barcode, the number of UMIs of VDJ_light_reads, and the number of UMIs of VDJ_heavy_reads; S3. Re-extracting reads containing the retained barcode from the VDJ sequencing data to obtain retained reads; wherein the retained reads are used for single cell immune repertoire analysis; In step S1, the method of sequentially performing barcode correction and UMI correction on reads includes: After barcode correction of reads, valid reads containing valid barcodes are retained; Align read2 in the obtained valid reads to the genome of the species corresponding to the sample to obtain an alignment file, wherein the alignment file contains the position information of the valid reads on the genome; Sort the obtained alignment files according to the location information of valid reads on the genome and their barcodes; According to the sequencing results, UMI correction is performed on the valid reads located in the same gene and having the same barcode to obtain corrected reads; the UMI correction is performed based on Hamming Distance. When the Hamming Distance between two UMIs is ≤1, the UMI with fewer valid reads is corrected to a UMI with more valid reads.
2. The method according to claim 1, characterized in that In step S1, no less than 1 million pairs of reads are extracted from the VDJ sequencing data.
3. The method according to claim 1, characterized in that In step S1, the method of extracting VDJ_reads aligned to the VDJ gene from the correction reads includes: The corrected reads are aligned with the VDJ light chain gene and / or VDJ heavy chain gene in the gene annotation file of the species corresponding to the sample, and VDJ_light_reads aligned to the VDJ light chain gene and / or VDJ_heavy_reads aligned to the VDJ heavy chain gene are extracted.
4. The method according to claim 1, characterized in that: In step S2, the preset filtering condition is: Barcodes that meet any of the following conditions are retained: Condition 1: 25≤the number of UMIs of the corrected reads≤100000, and the number of UMIs of VDJ_heavy_reads≥5, and the proportion of the number of UMIs of VDJ_heavy_reads≥5%, and the number of UMIs of VDJ_light_reads≥5, and the proportion of the number of UMIs of VDJ_light_reads≥10%; Condition 2: The UMI number of VDJ_heavy_reads accounts for ≥ 40%, and the UMI number of VDJ_light_reads accounts for ≥ 40%.
5. A denoising device for single-cell immune repertoire sequencing data, characterized in that: The device comprises: A preprocessing module, used to perform barcode correction and UMI correction on reads extracted from VDJ sequencing data in sequence to obtain corrected reads, and extract VDJ_reads aligned to VDJ genes from the corrected reads; wherein the VDJ_reads include VDJ_light_reads aligned to VDJ light chain genes, and / or VDJ_heavy_reads aligned to VDJ heavy chain genes; A filtering module is used to filter the barcodes in the reads according to a preset filtering condition to obtain the retained barcodes; wherein the preset filtering condition is set according to the number of UMIs of the corrected reads, the number of UMIs of VDJ_light_reads, and the number of UMIs of VDJ_heavy_reads under each barcode; An analysis module is used to re-extract reads containing the retained barcode from the VDJ sequencing data to obtain retained reads; wherein the retained reads are used for single-cell immune repertoire analysis; Wherein, the preprocessing module includes: A correction unit is used to perform barcode correction and UMI correction on reads in sequence and obtain corrected reads; wherein, after the barcode correction is performed on the reads, valid reads containing valid barcodes are retained; read2 in the obtained valid reads is aligned to the genome of the species corresponding to the sample to obtain an alignment file, wherein the alignment file contains the position information of the valid reads on the genome; the obtained alignment file is sorted according to the position information of the valid reads on the genome and their barcodes; according to the sorting result, UMI correction is performed on valid reads located in the same gene and having the same barcode to obtain corrected reads; the UMI correction is performed based on Hamming Distance, and when the Hamming Distance between two UMIs is ≤1, the UMI containing fewer valid reads is corrected to a UMI containing more valid reads; The extraction unit is used to compare the corrected reads with the VDJ light chain gene and / or VDJ heavy chain gene in the gene annotation file of the species corresponding to the sample, and extract VDJ_light_reads that are aligned to the VDJ light chain gene and / or VDJ_heavy_reads that are aligned to the VDJ heavy chain gene.
6. A terminal device, comprising a memory and a processor, wherein the memory stores a computer program that can be run on the processor, characterized in that: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 4 are implemented.
7. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 4 are implemented.
Citation Information
Patent Citations
Method for analyzing BCR characteristics of B lymphocytes based on single-cell multi-omics sequencing
CN116153408A
Method for detecting rare cell, apparatus, and use thereof
WO2023236121A1
