A method, reagent kit, and detection device for nucleic acid detection of lower respiratory tract infection pathogens.
By constructing dual-label libraries and evaluating sequence reliability, the problems of non-standard analysis procedures, long analysis times, and high false positive rates in mNGS technology for lower respiratory tract infection detection were solved, enabling rapid and accurate pathogen detection and improving detection efficiency and accuracy.
Patent Information
- Application Number
- CN202611287582.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-08-24
- Publication Date
- 2026-11-17
AI Technical Summary
Existing mNGS technology for detecting lower respiratory tract infections suffers from problems such as lack of standardized analysis procedures, poor reproducibility of results, excessively long analysis time, high false positive rate, and lack of targeted optimization. Furthermore, library construction and bioinformatics analysis lack integration and optimization.
A dual-label library construction method was adopted, combined with universal adapter ligation and magnetic bead in situ PCR amplification. Positive controls, negative controls and internal standards were set up for batch testing. Through sequence alignment and species differentiation, combined with sequence quality and repeatability evaluation, a rapid and accurate pathogen detection process was achieved.
It enables rapid and accurate detection throughout the entire process from nucleic acid sample to pathogen detection report, shortening analysis time by 6-20 times, increasing host sequence removal rate by 100 times, significantly improving detection sensitivity and specificity, and reducing false positive rate.
Smart Images

Figure CN123108133A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the fields of in vitro diagnostics, high-throughput sequencing, and bioinformatics analysis. Specifically, it relates to a method for detecting lower respiratory tract pathogens through metagenomic sequencing of DNA samples from bronchoalveolar lavage fluid, a library construction kit, a combination of detection products, and a detection device. In particular, it relates to a dry-wet synergistic detection system that combines universal adapter ligation, in situ PCR amplification with dual-tag magnetic beads, positive and negative controls, internal standards, template-free blank controls, and sequence-level reliability evaluation. Background Technology
[0002] Lower respiratory tract infections are among the most common infectious diseases in clinical practice, with a complex and diverse spectrum of pathogens, encompassing multiple categories such as bacteria, viruses, fungi, mycoplasma, chlamydia, and mycobacteria. Furthermore, mixed infections are increasingly common. Rapid and accurate identification of the causative microorganism is crucial for guiding clinical anti-infective treatment.
[0003] Currently used clinical methods for pathogen detection have significant limitations. Traditional culture methods are time-consuming, typically requiring 3-7 days, and have a high false negative rate; immunological detection methods have a limited recognition range and low sensitivity; while existing molecular detection technologies (such as PCR) have high sensitivity, their detection range is limited to specific pathogens, making them unable to detect unknown or rare pathogens and lacking flexibility.
[0004] Metagenomics Next-Generation Sequencing (mNGS) technology detects the nucleic acid sequences of all microorganisms in a sample and obtains the composition of pathogenic microorganisms in the sample through bioinformatics analysis. It has the advantages of wide detection range of pathogens, high sensitivity, high accuracy and fast response time, and is gradually becoming a first-line clinical method for detecting pathogen infection.
[0005] However, the translation of sequencing data into clinical diagnostic results using mNGS technology highly depends on complex bioinformatics analysis processes. A standard mNGS test typically generates approximately 20 M-40 M sequencing reads, with a data volume of 1-2 GB. It requires steps such as data quality control, host sequence removal, species identification, and result interpretation, each of which involves complex algorithm selection, parameter optimization, and quality control.
[0006] Currently, mNGS technology faces the following key technical challenges:
[0007] First, the analytical procedures lack standardization, resulting in poor reproducibility of results. Different laboratories use different analytical tools, algorithms, and parameters, leading to differences of 10-20% in results for the same sample.
[0008] Second, the analysis time is too long. Analytical strategies that pursue high sensitivity require an hour or even longer, which cannot meet the rapid diagnostic needs of critically ill patients.
[0009] Third, the false positive rate is high. Due to factors such as high sequence homology among closely related species and inconsistent quality of pathogen reference databases, false positives are often reported, affecting diagnostic accuracy.
[0010] Fourth, there is a lack of targeted optimization. The general analysis workflow cannot be optimized for specific pathogens in the lower respiratory tract, and cannot maximize analysis efficiency while ensuring accuracy.
[0011] Furthermore, in the library construction stage, the quality of the lower respiratory tract sample nucleic acid library construction directly affects the effective information content of subsequent sequencing data, thereby affecting the final pathogen detection performance. The starting amount, fragmentation method, adapter ligation efficiency, and PCR amplification cycle number of the library construction all need to be systematically optimized and validated for the mNGS detection scenario.
[0012] In existing technologies, library construction and bioinformatics analysis are usually implemented as independent steps, lacking systematic integration and optimized validation for the detection of pathogens causing lower respiratory tract infections.
[0013] In summary, there is an urgent need to establish a mNGS detection system for lower respiratory tract infection pathogens that organically combines library construction, internal standard and control quality control, sequence-level contamination identification, and species-level background correction, so as to achieve rapid and accurate detection throughout the entire process from nucleic acid samples to pathogen detection reports. Summary of the Invention
[0014] The purpose of this invention is to address the technical problems of existing mNGS detection technologies, such as the independent front-end library construction, experimental quality control, and back-end bioinformatics analysis, low standardization of the analysis process, excessively long analysis time, high false positive rate, and lack of targeted optimization. This invention provides a method and detection device for identifying lower respiratory tract infection pathogens that integrates library construction and bioinformatics analysis. Based on the above objective, this invention provides a method, kit, product combination, and detection device for nucleic acid detection of lower respiratory tract pathogens based on dual-label library construction and reliability evaluation of detected sequences. The method obtains a sequencing library through universal adapter ligation, in-situ dual-label PCR amplification with magnetic beads, and batch settings for positive control, negative control, internal standard, and template-free blank control (NTC). After pathogen identification, the reliability of each detected sequence is evaluated using sequence alignment, species differentiation, sequence quality, duplication, cross-sample consistency, and batch background characteristics. The results are then combined with effective sequence number standardization and NTC background correction to generate a report for reviewing candidate results.
[0015] A metagenomic sequencing method for detecting nucleic acids of pathogenic microorganisms in lower respiratory tract samples includes the following steps:
[0016] S1. Library Construction and Sequencing: Nucleic acids were extracted from the lower respiratory tract sample to be tested, positive control, negative control, and template-free blank control. Internal standards were added to the nucleic acids of the sample to be tested, positive control, and negative control, but no internal standards were added to the template-free blank control. Each nucleic acid was fragmented, end-repaired, A-added, adapter-ligated, and amplified by PCR to obtain a sequencing library containing sample tag information. The sample library, positive control library, negative control library, and template-free blank control library were subjected to high-throughput sequencing in the same testing batch to obtain raw sequencing data that could be distinguished according to samples and controls. The sample tag information is used to determine the sample or control source corresponding to the sequencing sequence. The detection results of the positive control, negative control, internal standard, and template-free blank control, as well as the sequence distribution information between different samples and controls in the same testing batch, are used for subsequent reliability evaluation of detected sequences, batch background identification, and reporting rule judgment.
[0017] S2. Sequencing data quality control: Perform quality control on the raw sequencing data, remove sequencing sequences that do not meet the preset quality requirements, and obtain high-quality sequencing sequences;
[0018] S3. Host sequence removal: The high-quality sequencing sequence is compared with the host sequence database, and sequencing sequences identified as originating from the host are removed to obtain host-free sequencing sequences.
[0019] S4. Pathogen identification: Based on the lower respiratory tract pathogen reference database, the host-de-host sequencing sequences are first preliminarily classified using the k-mer classification method. Then, the sequencing sequences that are not classified by the k-mer classification method are compared with reference sequences. The results of the preliminary classification and the results of the reference sequence comparison are integrated to obtain candidate pathogens and the set of detection sequences corresponding to the candidate pathogens.
[0020] S5. Detection sequence reliability assessment: Extract sequence-level features for each detected sequence in the detected sequence set, perform hard filtering on each detected sequence based on the sequence-level features, and score the reliability of detected sequences that do not trigger the hard filtering rules; determine the detected sequences that reach the preset reliability score threshold as reliable detected sequences, and determine the detected sequences that trigger the hard filtering rules or do not reach the preset reliability score threshold as unreliable detected sequences;
[0021] S6. Reliable Detection Sequence Statistics and Report Assistance Generation: The reliable detection sequences are aggregated according to the candidate pathogens to obtain the number of reliable detection sequences corresponding to each candidate pathogen. The number of reliable detection sequences is standardized to a preset sequencing data volume based on the total number of valid sequences in the sample. Combined with the background detection sequence number of the corresponding candidate pathogen in the template-free blank control and the preset reporting rules, it is determined whether the candidate pathogen is included in the report review candidate list, and the report review candidate result is generated.
[0022] In step S2, the raw sequencing data is subjected to quality control using the FastP tool. This quality control includes:
[0023] Remove adapter sequences, low-complexity sequences, PCR repetitive sequences, and short sequences less than 36 bp in length; and
[0024] A sliding window with a window size of 5 bp is used to remove low-quality regions with an average quality value lower than Q15.
[0025] The host sequence database is constructed through the following steps:
[0026] S31. Obtain and integrate human genome sequences, human transcript sequences, and human mitochondrial sequences from multiple sources;
[0027] S32. The integrated sequence is split into sequence segments of a preset length, and the sequence segments are assembled and clustered to remove redundancy.
[0028] S33. Identify and remove endogenous viral integration sequences, microbial contamination sequences, and sequence regions highly homologous to the target pathogen from the database to obtain a host sequence database for host sequence removal.
[0029] In step S31, the human genome sequence, human transcript sequence, and human mitochondrial sequence include the complete human genome sequence T2T-CHM13v1.1, the Chinese population-specific genome sequence YH2.0, the GRCh38.p13 reference genome sequence, human sequences from the NCBI NT library, human mRNA sequences, fusion gene sequences, and human mitochondrial sequences.
[0030] In step S32, the integrated sequence is split into sequence segments of 500 bp in length with adjacent sequence segments overlapping by 50 bp, and clustering is performed to remove redundancy using a sequence similarity of not less than 98%.
[0031] The host sequence database was compressed from 70 GB to 4.2 GB after redundancy removal.
[0032] Step S33 includes:
[0033] A representative genome of the pathogenic microorganism was selected, and a sequencing sequence with a length of 150 bp and a coverage of 5× was generated based on the representative genome.
[0034] The simulated sequencing sequences are aligned to the host sequence database to be processed.
[0035] The host database regions with an alignment similarity of not less than 95% and an alignment length of not less than 100 bp with the simulated sequencing sequences are identified as homologous regions of pathogenic microorganisms, and these homologous regions are removed from the host sequence database.
[0036] The lower respiratory tract pathogen reference database is constructed through the following steps:
[0037] S41. Obtain the genome sequence of the target pathogen from a public genome database;
[0038] S42. Based on the multi-dimensional quality characteristics and weighted scoring model of the genome sequence, the genome sequence is evaluated and screened to obtain a set of high-quality candidate genome sequences;
[0039] S43. Cluster the high-quality candidate genome sequence set and select representative genome sequences from each cluster category to form a non-redundant representative genome set;
[0040] S44. Verify the species classification accuracy of the representative genome sequences by average nucleic acid consistency analysis, and remove genome sequences with incorrect species classification;
[0041] S45. The verified genome sequence is subjected to contamination fragment identification and removal to obtain a high-quality genome sequence for constructing the lower respiratory tract pathogen reference database;
[0042] S46. Construct an index based on the high-quality genome sequence for k-mer classification, reference sequence alignment, and similarity search.
[0043] In step S41, 578,817 genome sequences from 23,241 species were obtained from the NCBI GenBank database, and 52,388 high-quality genomes were obtained through screening in steps S42 to S45.
[0044] The multidimensional quality characteristics include genome integrity, contamination level, N base ratio, sequencing depth, and species annotation quality.
[0045] In step S46, the Kraken2 index, Bowtie2 index, and BLAST index are constructed respectively. The Kraken2 index is used for rapid k-mer classification and pre-screening of pathogenic microorganisms, the Bowtie2 index is used for accurate sequence alignment and species identification, and the BLAST index is used for similarity search and differentiation of closely related species.
[0046] In step S4, Kraken2 is used to perform k-mer rapid classification of the hostless sequencing sequences based on the Kraken2 index, and Bowtie2 is used to perform reference sequence alignment of the sequencing sequences that were not classified by Kraken2 based on the Bowtie2 index. The Kraken2 classification results and Bowtie2 alignment results are then integrated.
[0047] The detected sequences obtained through Kraken2 classification are merged with those obtained through Bowtie2 reference sequence alignment to form a candidate detected sequence set. BLAST is used to perform a similarity search on the candidate detected sequences in the candidate detected sequence set based on the BLAST index, obtaining the best hit, second best hit, effective alignment length, alignment similarity, alignment score, reference sequence number, alignment position, genomic hit information within the target species, and hit information of closely related species for each candidate detected sequence. Based on the Kraken2 classification results, Bowtie2 reference sequence alignment results, and BLAST similarity search results, the candidate pathogenic microorganisms corresponding to each candidate detected sequence are determined, obtaining the candidate pathogenic microorganisms and their corresponding detected sequence sets. The BLAST similarity search results are then used for sequence-level feature extraction in step S5.
[0048] The sequence-level features extracted in step S5 include:
[0049] Effective alignment length (EAL), alignment similarity (IDY), best-to-second best alignment score difference (BSD), target species-specific k-mer ratio (SSKR), number of species with the same best hit (EBHC), lowest common ancestor classification level (LCA), ratio of closely related species (RSHR), ratio of genomes within the target species (IGHR), origin region type (RegionType), average Phred quality score (APQ), Q30 base ratio (Q30R), N base ratio (NR), low complexity ratio (LCR), PCR duplicate marker (PCRdup) within the same sample, optical duplicate marker (OdupX) between samples, NTC hit marker (CtrlHit) within the same batch, and high-frequency background sequence marker (HFBG) within the same batch.
[0050] The EAL is the base length of the effective alignment between the detected sequence and the reference sequence of the candidate pathogenic microorganism; when there are multiple local alignment segments, the overlapping intervals on the detected sequence are merged and the EAL is calculated.
[0051] The IDY is the proportion of the number of matching bases between the detected sequence and the candidate pathogen reference sequence to the EAL.
[0052] The BSD is calculated according to the following formula: BSD=(BS1-BS2) / BS1; where BS1 is the best alignment score of the detected sequence against the species to which the candidate pathogenic microorganism belongs, and BS2 is the highest alignment score of the detected sequence against other species; when there is no second-best effective hit, BSD is 1.
[0053] SSKR splits the detected sequence into k-mers according to a preset k value, removes k-mers containing N bases and low-complexity k-mers, and only supports the proportion of k-mers of the species to which the candidate pathogenic microorganism belongs to the total number of effective k-mers.
[0054] The EBHC refers to the number of different species whose alignment score is not less than 98% of the best alignment score and which meets the preset minimum effective alignment conditions.
[0055] The LCA is the lowest taxonomic level that the detected sequence can support after performing a lowest common ancestor analysis on a set of species that meet the EBHC criteria.
[0056] The RSHR is the proportion of the number of closely related species reference genomes to which the detected sequence can be effectively matched out of the total number of closely related species reference genomes. Effective match means that the EAL is not less than 36 bp and the IDY is not less than 90%.
[0057] The IGHR is the proportion of the number of reference genomes of the candidate pathogenic microorganism whose detected sequence can be effectively matched to the total number of reference genomes of that species. Effective match means that the EAL is not less than 36 bp and the IDY is not less than 90%.
[0058] The RegionType is determined based on the position of the detected sequence in the reference sequence, including core genome regions, species-specific regions, rRNA or ITS regions, plasmid regions, phage or prephage regions, transposon element regions, repeat regions, and reference blacklist regions.
[0059] The APQ is the average Phred quality value of all bases in the detected sequence;
[0060] Q30R is the proportion of the number of bases in the detected sequence with a quality value of Q30 to the total number of bases in the detected sequence.
[0061] The NR is the ratio of the number of N bases in the detected sequence to the length of the detected sequence;
[0062] The LCR is the ratio of the length of the low-complexity region in the detected sequence to the length of the detected sequence.
[0063] The PCRdup is used to indicate whether the detected sequence is a representative sequencing sequence retained for a PCR repeat set in step S2. The PCR repeat set information is used to record the PCR repeat set corresponding to the representative sequencing sequence, and the same PCR repeat set is counted as an independent molecular evidence in step S6. The PCRdup and the corresponding PCR repeat set information do not participate in the weighted reliability score.
[0064] The OdupX is used to indicate whether the detected sequence in the current sample is completely consistent with the sequencing sequences of other samples in the same batch in the valid sequence region;
[0065] The CtrlHit is used to indicate whether the detected sequence appears in the negative control or blank control of the same batch of samples.
[0066] The HFBG is used to indicate whether the detected sequence appears repeatedly in multiple samples of the same batch.
[0067] A detected sequence is directly classified as an unreliable detected sequence and no reliability score is applied if it meets any of the following conditions:
[0068] The detected sequence appeared in the negative control or blank control of the same batch of samples.
[0069] The detected sequences belong to the same batch of high-frequency background sequences.
[0070] The detected sequence is an optically repeatable sequence between samples;
[0071] The detected sequence is derived from a reference blacklist region; or
[0072] The number of equivalent best-hit species of the detected sequence is not less than 3, and its lowest common ancestor classification level cannot support species-level resolution.
[0073] For a detected sequence q that does not trigger the hard filtering rule, the reliability score R(q) is calculated according to the following formula:
[0074] in, Let be the weight of the i-th sequence-level scoring indicator. Assign a score to the i-th sequence-level scoring indicator; the sum of the weights of each sequence-level scoring indicator is 1, the value of the assigned score ranges from 0 to 1, and the value of the reliability score R(q) ranges from 0 to 100.
[0075] When R(q) is not less than 70 points, the detected sequence q is determined to be a reliable detected sequence; when R(q) is less than 70 points, the detected sequence q is determined to be an unreliable detected sequence and is not included in the number of reliable detected sequences.
[0076] The weights and scoring rules for the sequence-level scoring indicators are as follows:
[0077] The weight of EAL is 8%; when EAL is not less than 45 bp, it is assigned a score of 1; when EAL is not less than 36 bp but less than 45 bp, it is assigned a score of (EAL-36) / 9; when EAL is less than 36 bp, it is assigned a score of 0.
[0078] The weight of IDY is 8%; when IDY is not lower than 95%, the score is 1; when IDY is not lower than 90% but lower than 95%, the score is (IDY-90%) / 5%; when IDY is lower than 90%, the score is 0.
[0079] The weight of BSD is 8%; when BSD is not less than 5%, the score is 1; when BSD is not less than 1% but less than 5%, the score is (BSD-1%) / 4%; when BSD is less than 1%, the score is 0.
[0080] The weight of SSKR is 10%; when SSKR is not less than 30%, the score is 1; when SSKR is not less than 10% but less than 30%, the score is (SSKR-10%) / 20%; when SSKR is less than 10%, the score is 0.
[0081] The weight of EBHC is 6%; when EBHC is 1, the score is 1; when EBHC is 2, the score is 0.5; when EBHC is not less than 3, the score is 0.
[0082] The weight of LCA is 5%; when LCA supports the species level, it is assigned a score of 1; when LCA supports the genus level, it is assigned a score of 0.5; when LCA only supports the family level or higher taxonomic levels, it is assigned a score of 0.
[0083] The weight of RSHR is 6%; when RSHR is not higher than 1%, the score is 1; when RSHR is higher than 1% but not higher than 10%, the score is (10%-RSHR) / 9%; when RSHR is higher than 10%, the score is 0.
[0084] The weight of IGHR is 10%; when IGHR is not lower than 80%, the score is 1; when IGHR is not lower than 20% but lower than 80%, the score is (IGHR-20%) / 60%; when IGHR is lower than 20%, the score is 0.
[0085] The weight of RegionType is 12%; when the detected sequence originates from the core genome region or species-specific region, it is assigned a score of 1; when the detected sequence originates from the rRNA or ITS region, it is assigned a score of 0.5; when the detected sequence originates from the plasmid, phage, prophage, transposon element, or repeat region, it is assigned a score of 0.3; when the detected sequence originates from the reference blacklist region, it is assigned a score of 0; detected sequences originating from the reference blacklist region are processed according to the hard filtering rules described above and are not included in the weighted reliability score.
[0086] The APQ weight is 5%; when APQ is not less than 30, the score is 1; when APQ is not less than 20 but less than 30, the score is (APQ-20) / 10; when APQ is less than 20, the score is 0.
[0087] The weight of Q30R is 4%; when Q30R is not less than 80%, it is assigned a score of 1; when Q30R is not less than 50% but less than 80%, it is assigned a score of (Q30R-50%) / 30%; when Q30R is less than 50%, it is assigned a score of 0.
[0088] The weight of NR is 3%; when NR is not higher than 1%, it is assigned a score of 1; when NR is higher than 1% but not higher than 5%, it is assigned a score of (5%-NR) / 4%; when NR is higher than 5%, it is assigned a score of 0.
[0089] The weight of LCR is 4%; when LCR is not higher than 20%, it is assigned a score of 1; when LCR is higher than 20% but not higher than 50%, it is assigned a score of (50%-LCR) / 30%; when LCR is higher than 50%, it is assigned a score of 0.
[0090] The hard filtering rules, the weights of the sequence-level scoring metrics, the scoring thresholds, and the preset reliability scoring thresholds are determined in the following ways:
[0091] Using preset library construction conditions, libraries were constructed and sequenced for lower respiratory tract clinical samples, positive controls, negative controls, template-free blank controls, simulated samples, and known contaminated samples to form a training dataset.
[0092] The preset library construction conditions include at least universal adapter ligation, magnetic bead in situ dual-tag PCR amplification, internal standard addition, and template-free blank control from the same batch.
[0093] The detected sequences in the training dataset are labeled as reliable or unreliable detected sequences, and the corresponding sequence-level features are extracted.
[0094] Through statistical analysis, parameter search, and cross-validation, the weights, scoring thresholds, and preset reliability scoring thresholds of the hard filtering rules and sequence-level scoring indicators are determined.
[0095] In step S6, only the reliable detection sequence is counted in the number of reliable detection sequences of the candidate pathogenic microorganism;
[0096] When the number of reliable detected sequences of a candidate pathogen reaches the requirements specified in the preset reporting rules, the candidate pathogen is added to the report review candidate list.
[0097] If the number of reliable detected sequences of a candidate pathogen does not meet the requirements specified in the preset reporting rules, the candidate pathogen will not be added to the report review candidate list.
[0098] The number of standardized, reliable detectable sequences of candidate pathogenic microorganisms is calculated using the following formula:
[0099] Standardized reliable detection sequence count = Reliable detection sequence count of candidate pathogenic microorganisms × 20,000,000 ÷ Total effective sequence count of the sample;
[0100] The total number of valid sequences in the sample refers to the total number of sequences from the original sequencing data that were used for pathogen comparison and analysis after quality control.
[0101] For target pathogenic microorganisms other than Mycobacterium tuberculosis complex, Mycoplasma pneumoniae, Chlamydia pneumoniae, and Chlamydia psittaci:
[0102] When the number of detected sequences of the target pathogen in the template-free blank control is not 0, if the number of standardized reliable detected sequences of the target pathogen in the test sample is not less than 10 times the number of corresponding detected sequences in the template-free blank control, then the target pathogen will be added to the report review candidate list.
[0103] When the number of detected sequences of the target pathogen in the template-free blank control is 0, if the number of reliable detected sequences of the target pathogen in the test sample is not less than 2 and the number of standardized reliable detected sequences is not less than 3, then the target pathogen will be added to the report review candidate list.
[0104] For Mycobacterium tuberculosis complex, Mycoplasma pneumoniae, Chlamydia pneumoniae and Chlamydia psittaci, the corresponding target pathogen will be added to the report review candidate list when at least one reliable detection sequence is present in the sample to be tested.
[0105] The validity of the test results must at least meet the following requirements: all preset target pathogens in the positive control are detected; none of the preset target pathogens in the negative control are detected; internal standards are detected in the test sample, positive control, and negative control; and the template-free blank control meets the preset batch background quality control requirements.
[0106] Step S6 does not perform species-level comprehensive reliability scoring on candidate pathogenic microorganisms, nor does it classify them as high or low reliability based on the species-level comprehensive reliability score.
[0107] The candidate results for report review are used to assist in report review. The final test report is determined by combining experimental quality control results, sample type, clinical relevance of pathogenic microorganisms, manual review results, and the preset report rules.
[0108] A library construction kit for nucleic acid metagenomic sequencing detection of lower respiratory tract pathogens, comprising:
[0109] Fragmentation, end repair, and A-addition reaction solution, wherein the reaction solution includes fragmentation enzyme, DNA polymerase, and polynucleotide kinase;
[0110] LRTI ligase buffer;
[0111] LRTI ligase;
[0112] Universal connector;
[0113] DNA amplification enzyme;
[0114] Dual-tag amplification primers containing i5 and i7 tag sequences;
[0115] Positive control;
[0116] Negative control; and
[0117] Internal standards from non-target microorganism sources.
[0118] The library construction kit according to the preceding claims is characterized in that the positive control comprises genomic DNA of Klebsiella pneumoniae, Staphylococcus aureus, avian intracellular mycobacterial complex, Candida albicans, human adenovirus type C and Mycoplasma pneumoniae, as well as genomic DNA of human cell lines.
[0119] The negative control includes Cryptococcus glottii genomic DNA and human cell line genomic DNA, and the internal standard includes Trichoderma citrinum genomic DNA.
[0120] A metagenomic sequencing detection device for pathogenic microorganism nucleic acids in lower respiratory tract samples, comprising:
[0121] The sequencing data acquisition module is used to acquire raw sequencing data obtained from nucleic acid library construction and high-throughput sequencing of lower respiratory tract samples;
[0122] The data quality control module is used to control the quality of the raw sequencing data to obtain high-quality sequencing sequences;
[0123] The host sequence removal module is used to align the high-quality sequencing sequence to a host sequence database, remove the sequencing sequence from the host, and obtain a host-free sequencing sequence.
[0124] The pathogen identification module is used to perform preliminary classification of the host-de-host sequencing sequences based on the lower respiratory tract pathogen reference database and the k-mer classification method, and to perform reference sequence alignment on the unclassified sequencing sequences to obtain candidate pathogens and their corresponding set of detected sequences.
[0125] The detection sequence reliability determination module is used to perform sequence-level feature extraction, hard filtering, and reliability scoring on each detection sequence in the detection sequence set, remove unreliable detection sequences, and obtain a reliable detection sequence set.
[0126] A reliable detection sequence statistics module is used to aggregate and statistically analyze the reliable detection sequences according to candidate pathogenic microorganisms to obtain the number of reliable detection sequences corresponding to each candidate pathogenic microorganism; and
[0127] The report review assistance generation module is used to generate candidate results for report review based on the number of reliable detected sequences and preset report rules.
[0128] The detection device also includes a processor and a memory;
[0129] The sequencing data acquisition module, data quality control module, host sequence removal module, pathogen identification module, detection sequence reliability determination module, reliable detection sequence statistics module, and report review and auxiliary generation module are stored as computer programs in the memory.
[0130] When the computer program is executed by the processor, it enables the identification of pathogenic microorganism nucleic acids in lower respiratory tract samples, reliable detection sequence statistics, and report review assistance generation.
[0131] The data quality control module uses FastP to remove adapter sequences, low-complexity sequences, PCR repetitive sequences, and short sequences less than 36 bp in length, and uses a sliding window with a window size of 5 bp to remove low-quality regions with an average quality value lower than Q15.
[0132] The host sequence removal module uses Bowtie2 to align high-quality sequencing sequences to the host sequence database and removes human sequencing sequences with an alignment similarity of not less than 95% and an alignment length of not less than 40 bp.
[0133] The pathogen identification module uses Kraken2 for initial k-mer classification and Bowtie2 for reference sequence alignment of sequencing sequences that were not classified by Kraken2.
[0134] The detected sequence reliability determination module is configured as follows:
[0135] Extract the alignment quality characteristics, species discrimination ability characteristics, intra-species genome hit characteristics, sequence origin region characteristics, sequence quality and complexity characteristics, repetitive sequence characteristics, cross-sample consistency characteristics, and contamination background characteristics of the detected sequences;
[0136] Hard filtering was performed based on the hit rate of template-free blank controls, high-frequency background sequences within the same batch, optical repeatability between samples, hit rate of reference blacklist regions, and hit rate of equivalent best species; and
[0137] Detected sequences that do not trigger hard filtering rules are assigned a weighted reliability score, and those sequences with a reliability score of not less than 70 are determined to be reliable detected sequences.
[0138] The detection device further includes an experimental quality control module, a sequence number standardization module, and a template-free blank control background correction module. The experimental quality control module is used to determine whether the positive control, negative control, and internal standard meet the preset validity requirements. The sequence number standardization module is used to standardize the number of reliably detected sequences to 20,000,000 valid sequences. The template-free blank control background correction module is used to compare the target pathogen signal of the sample to be tested with the background signal of the template-free blank control in the same batch of tests.
[0139] A computer-readable storage medium having a computer program stored thereon, the computer program, when executed by a processor, performing the following steps:
[0140] Quality control was performed on the raw metagenomic sequencing data of lower respiratory tract samples to obtain high-quality sequencing sequences;
[0141] The high-quality sequencing sequences are compared with the host sequence database and the host-derived sequencing sequences are removed to obtain host-free sequencing sequences.
[0142] Based on the lower respiratory tract pathogenic microorganism reference database, candidate pathogenic microorganisms and their corresponding detection sequences were identified by combining k-mer classification and sequence alignment.
[0143] For each detected sequence, sequence-level feature extraction, hard filtering, and reliability scoring are performed to obtain reliable detected sequences; and
[0144] The number of reliably detected sequences is counted based on the candidate pathogenic microorganisms, and candidate results for report review are generated based on the number of reliably detected sequences and preset reporting rules.
[0145] A computer program product includes a computer program that, when run on a processor, causes the processor to perform steps S2 to S6 in the metagenomic sequencing detection method of claim 1.
[0146] A portfolio of metagenomic sequencing detection products for lower respiratory tract pathogens includes:
[0147] The library construction kit described in any of the preceding claims; and
[0148] Metagenomic sequencing data analysis program used in conjunction with the library construction kit or a computer-readable storage medium storing the metagenomic sequencing data analysis program;
[0149] The metagenomic sequencing data analysis program is configured to receive sequencing data generated from dual-tag sequencing libraries prepared by the library construction kit, and to perform quality control, host sequence removal, pathogen identification, reliability evaluation of detected sequences, statistics of reliable detected sequences, standardization of effective sequence numbers, background correction of template-free blank control, and generation of candidate results for report review on the sequencing data.
[0150] The reliability evaluation of the detected sequences utilizes at least one or more of the following: intra-sample repeatability features, cross-sample complete consistency features, template-free blank control sequence hit features, and high-frequency background features from the same detection batch.
[0151] The beneficial effects of this invention are:
[0152] (1) Achieve full-process integration: Organically integrate library construction with standardized bioinformatics analysis system to achieve full-process detection from lower respiratory tract sample nucleic acid to pathogen detection report.
[0153] (2) Fast analysis speed: The analysis time for a single sample is reduced from 60-120 minutes in the traditional method to 6-10 minutes, which is 6-20 times faster and meets the rapid diagnosis needs of patients with acute and critical illnesses in clinical practice.
[0154] (3) High host removal efficiency: By systematically constructing a high-quality host database, the host sequence removal rate reaches 99.99%, which is 100 times higher than conventional methods in terms of microbial signal enrichment.
[0155] (4) High detection sensitivity: The lowest detection limit is 10 copies / sample, and the detection sensitivity reaches 97.9% under simulated sample (pathogen abundance 10 reads) conditions.
[0156] (5) Low false positive rate: The detection specificity reaches 99.1% under simulated sample (pathogen abundance 10,000 reads) conditions. Through the reliability evaluation of the detected sequence, unreliable detected sequences such as non-specific alignment, sharing of closely related species, repetitive sequences, negative / blank control contamination, and background noise of the same test batch are eliminated, which effectively reduces false positives. Attached Figure Description
[0157] Figure 1: Schematic diagram of the library construction process of this invention
[0158] Figure 2 Schematic diagram of the construction process of the pathogenic microorganism genome reference database of this invention
[0159] Figure 3 : Schematic diagram of the host sequence database construction process of this invention
[0160] Figure 4 : Schematic diagram of the metagenomic data analysis workflow of this invention
[0161] Figure 5 Statistical chart of host sequence removal efficiency evaluation results
[0162] Figure 6 Analysis of statistical results over time Detailed Implementation
[0163] The specific technical solution of the present invention is as follows:
[0164] I. Document Building Module
[0165] This invention employs enzyme digestion and fragmentation, along with in-situ dual-tag PCR amplification using magnetic beads, to construct a library of sample nucleic acids. The library construction process includes adjusting the amount of nucleic acid introduced and adding internal standards, enzyme digestion and fragmentation, end repair and A-addition, universal adapter ligation, binding with 0.6× magnetic beads, in-situ PCR with dual-tag primers added without eluting the magnetically bound nucleic acids, purification with 1.0× magnetic beads, and library quality control.
[0166] II. Construction of Pathogenic Microorganism Genome Reference Database
[0167] This invention constructs a customized reference database for the nucleic acids of lower respiratory tract infection pathogens. This database is built through a quality control process including genome screening, redundancy removal, misannotation correction, and contamination region removal. Based on the screened high-quality genome sequences, Kraken2, Bowtie2, and BLAST indexes are constructed to support the implementation of a joint identification strategy of rapid k-mer classification and accurate sequence alignment.
[0168] III. Construction of Host Sequence Database
[0169] This invention constructs a high-purity host sequence database by integrating multiple versions of human genome data, performing redundancy removal, and removing endogenous viral and microbial contamination sequences through quality control processes. The database size is optimized to 4.2 GB, and the host sequence removal rate reaches 99.99%.
[0170] IV. Metagenomic Data Analysis Workflow
[0171] The metagenomic data analysis workflow includes data quality control, host sequence removal, pathogen identification, reliability assessment of detected sequences, and statistical analysis and report review assistance for reliable detected sequences. The entire analysis process takes 6-10 minutes.
[0172] V. Reliability assessment of detected sequences and statistics of reliable detected sequences
[0173] This invention establishes a method for evaluating the reliability of detected sequences. For each detected sequence output by the pathogen identification module, sequence-level feature extraction, hard filtering, and reliability scoring are performed to eliminate unreliable detected sequences. The number of reliable detected sequences is then counted according to the candidate pathogens. This number of reliable detected sequences is used to assist in report review and to determine the validity of preset reporting rules.
[0174] Some embodiments of this patent include the following technical solutions:
[0175] A product portfolio for detecting nucleic acid metagenomic sequencing of lower respiratory tract pathogens is provided, the product portfolio comprising a library construction kit and accompanying analysis procedures or a computer-readable storage medium storing the accompanying analysis procedures.
[0176] The library construction kit includes fragmentation, end repair and A-addition reaction solution, LRTI ligase buffer, LRTI ligase, universal adapter, amplification enzyme, double-tagged amplification primers, positive control, negative control and internal standard.
[0177] The detection process also includes an NTC (Network Control Unit) that performs sample processing, nucleic acid extraction, library construction, and sequencing in parallel with the sample to be tested. The positive control is used to verify the detection capability for different types of target pathogens; the negative control is used to verify the negative status of the target pathogen and the ability to distinguish closely related species; the internal standard is used to monitor the library construction, sequencing, and analysis process of a single sample; and the NTC is used to monitor batch background introduced by reagents, environment, and experimental operations.
[0178] The accompanying analysis program utilizes the detection results of the test sample, positive control, negative control, internal standard, and NTC to complete experimental quality control, reliability evaluation of detected sequences, statistics of reliable detected sequences, standardization of effective sequence numbers, and NTC background correction.
[0179] The library construction module includes the following steps:
[0180] (1) Input volume and addition of internal standard: Take 10 ng of nucleic acid according to the nucleic acid quantification results and add water to 24 μL; when the nucleic acid concentration is not higher than 0.4 ng / μL, take 24 μL of nucleic acid extraction product; add 1 μL of internal standard to the test sample, positive control and negative control, and do not add internal standard to NTC;
[0181] (2) Enzyme digestion, end repair and A addition: Add 5 μL of fragmentation, end repair and A addition reaction solution to 25 μL of nucleic acid entering the system;
[0182] (3) Universal adapter ligation: Add LRTI ligase buffer, LRTI ligase and universal adapter to 30 μL of the product from the previous step to carry out the adapter ligation reaction;
[0183] (4) 0.6× magnetic bead binding: Add 30 μL of nucleic acid purification magnetic beads to 50 μL of ligation product to bind the adapter ligation product to the magnetic beads. Remove the supernatant after magnetic adsorption, without eluting the nucleic acid.
[0184] (5) In situ PCR with dual-tag magnetic beads: PCR amplification reaction solution and dual-tag amplification primers containing i5 and i7 tag sequences were added directly to the magnetic beads, and 10 cycles of in situ PCR amplification with magnetic beads were performed.
[0185] (6) 1.0× magnetic bead purification: Add 50 μL of nucleic acid purification magnetic beads to the amplification product, wash with ethanol and elute with enzyme-free water to obtain the final sequencing library;
[0186] (7) Library Quantification and Sequencing: The final library concentration was quantified. Libraries with different dual-tag combinations were mixed at equal quality for DNB preparation and combined probe-anchored polymerization sequencing. The sequencing data quality control module used the FASTP tool to perform quality control on the raw sequencing data. The quality control steps included: removing adapter sequences, removing low-complexity sequences, removing PCR repetitive sequences, removing short sequences less than 36 bp, and using the sliding window method to remove low-quality regions. The sliding window size was 5 bp, and regions with an average quality value lower than Q15 were removed, retaining high-quality reads for subsequent analysis.
[0187] The host sequence removal module uses Bowtie2 to align the quality-controlled sequencing data to the host sequence database, removes successfully aligned human reads, and retains unaligned reads for pathogen identification; the alignment parameters are: alignment similarity ≥ 95%, alignment length ≥ 40 bp.
[0188] The host sequence database is constructed through the following steps:
[0189] (1) Integrate multiple versions of human genome data, the sources of which include: T2T-CHM13v1.1 complete human genome sequence, YH2.0 Chinese population-specific genome sequence, GRCh38.p13 international standard reference genome sequence, NCBI NT library human sequence, human mRNA sequence, fusion gene sequence and mitochondrial sequence;
[0190] (2) The integrated sequence was split into 500 bp segments and the sequences were assembled. After redundancy removal by clustering with a sequence similarity of ≥98%, the database size was optimized to 4.2 GB, which is 94% smaller than the original integrated data;
[0191] (3) Identify and remove endogenous viral integration sequences and microbial contamination sequences from the human genome to obtain a high-purity host sequence database.
[0192] The pathogen identification module uses the following steps to identify pathogens from the host-free sequencing data:
[0193] (1) The Kraken2 k-mer method was used to quickly classify the hostless sequences and obtain preliminary identification results;
[0194] (2) For the sequences that were not classified in step (1), Bowtie2 was used to perform precise alignment and identification based on the reference genome;
[0195] (3) Integrate the identification results of steps (1) and (2) and output the final list of pathogenic microorganisms.
[0196] The customized lower respiratory tract pathogen reference database includes the Kraken2 index, Bowtie2 index, and BLAST index. These three types of indexes are used in combination to balance analysis speed and identification accuracy.
[0197] The reference database is constructed through the following steps:
[0198] (1) Obtain the genome sequence of the target pathogen from the NCBI GenBank database;
[0199] (2) Based on multi-dimensional feature annotation and weighted scoring model, the quality of genome sequences is evaluated and screened to obtain a set of high-quality candidate genome sequences;
[0200] (3) Cluster the high-quality candidate genome sequence set, select representative genome sequences from each category to form a non-redundant representative genome set;
[0201] (4) Verify the species classification accuracy of representative genome sequences through average nucleic acid consistency analysis and remove genome sequences with incorrect species classification;
[0202] (5) The verified genome sequence is subjected to contamination fragment identification and removal to obtain the final high-quality genome sequence;
[0203] (6) Based on the high-quality genome sequence, construct the Kraken2 index, Bowtie2 index and BLAST index respectively.
[0204] The detection sequence reliability determination module extracts sequence-level features for each detected sequence. These features include effective alignment length, alignment similarity, difference between the best and second-best alignment scores, target species-specific k-mer ratio, number of species with the same best hit, LCA taxonomic level, hit ratio of closely related species, genome hit ratio within the target species, source region type, average Phred quality value, Q30 ratio, N base ratio, low complexity ratio, PCR duplicates within the same sample, optical duplicates between samples, occurrence of negative or blank controls in the current batch, and high-frequency background sequences in the current batch.
[0205] The reliability determination module for the detected sequence first performs hard filtering. When the detected sequence appears in the negative control or blank control of the same test batch, belongs to the high-frequency background sequence of the same test batch, belongs to the optical duplicate between samples, originates from the reference blacklist area, or there are multiple species with the same best hit and the LCA classification level cannot support species-level resolution, the detected sequence is directly determined as an unreliable detected sequence and will not be included in the subsequent reliability scoring.
[0206] For detected sequences that do not trigger hard filtering rules, the reliability score of the detected sequence is calculated according to the formula R(q)=Σwi×si×100; where q is the detected sequence, wi is the weight of the i-th sequence-level indicator, si is the score of the i-th sequence-level indicator, and R(q) ranges from 0 to 100. When R(q)≥70, the detected sequence is determined to be a reliable detected sequence; when R(q)<70, the detected sequence is determined to be an unreliable detected sequence and is not included in the number of reliable detected sequences.
[0207] The reliable detection sequence statistics module aggregates and counts reliable detection sequences according to candidate pathogens to obtain the number of reliable detection sequences corresponding to each candidate pathogen; the number of reliable detection sequences is used for the rule judgment of pathogen positivity.
[0208] A kit for nucleic acid metagenomic sequencing detection of lower respiratory tract infection pathogens, the kit comprising library construction reagents;
[0209] The library construction kit comprises the following components: fragmentation, end repair and A-addition reaction solution, LRTI ligase buffer, LRTI ligase, universal adapter, DNA amplification enzyme, dual-tag amplification primers containing i5 and i7 tag sequences, positive control, negative control, and internal standard. This kit is used to prepare sequencing libraries of nucleic acids derived from lower respiratory tract samples.
[0210] The kit is suitable for nucleic acid samples derived from bronchoalveolar lavage fluid.
[0211] A suite of metagenomic sequencing products for detecting lower respiratory tract pathogens includes the aforementioned library construction kit, and metagenomic sequencing data analysis software used in conjunction with the library construction kit or a computer-readable storage medium storing the analysis software.
[0212] The analysis software includes a sequencing data quality control module, an experimental quality control module, a host sequence removal module, a pathogen identification module, a detected sequence reliability determination module, a reliable detected sequence statistics module, an effective sequence number standardization module, an NTC background correction module, and a report review assistance module.
[0213] The experimental quality control module is used to determine whether the detection results of positive controls, negative controls, and internal standards meet the preset validity requirements; the detection sequence reliability determination module is used to identify and remove specific contamination sequences, cross-sample completely identical sequences, and batch high-frequency background sequences; the effective sequence number standardization module is used to standardize the number of reliable detection sequences of candidate pathogens to 20,000,000 effective sequences; the NTC background correction module is used to compare the candidate pathogen signals in the test sample with the background signals in the same batch of NTC. A method for detecting lower respiratory tract infection pathogens through nucleic acid metagenomic sequencing includes the following steps:
[0214] (1) Library construction: Nucleic acid from lower respiratory tract samples is used to construct a library to obtain a sequencing library, and high-throughput sequencing is performed to obtain raw sequencing data;
[0215] (2) Data quality control: The raw sequencing data were quality controlled using the FASTP tool to remove adapter sequences, low-complexity sequences, PCR repetitive sequences and short sequences less than 36 bp in length. The sliding window method was used to remove low-quality regions with an average quality value lower than Q15 to obtain high-quality reads.
[0216] (3) Host sequence removal: High-quality reads are aligned to the host sequence database using Bowtie2, and human reads with alignment similarity ≥95% and alignment length ≥40 bp are removed, while unaligned reads are retained;
[0217] (4) Pathogen identification: Kraken2 is used to perform rapid k-mer classification of the reads after host removal, and Bowtie2 is used to perform precise comparison and identification of the unclassified sequences. The results of the two methods are integrated to output a list of pathogen identification.
[0218] (5) Detection sequence reliability assessment: Sequence-level feature extraction and hard filtering are performed on each detected sequence output by the pathogen identification module; For detected sequences that do not trigger the hard filtering rules, the reliability score R(q) of the detected sequence is calculated according to the preset weight and scoring rules. When R(q) ≥ 70 points, it is determined to be a reliable detected sequence, and when R(q) < 70 points, it is determined to be an unreliable detected sequence.
[0219] (6) Reliable Detection Sequence Statistics and Report Assistance Generation: After removing unreliable detection sequences, reliable detection sequences are collected and statistically analyzed according to candidate pathogens to obtain the number of reliable detection sequences corresponding to each candidate pathogen. The report review results are generated based on the number of reliable detection sequences and preset reporting rules.
[0220] The entire analysis process takes 6-10 minutes.
[0221] The lower respiratory tract sample was obtained from bronchoalveolar lavage fluid.
[0222] A rapid nucleic acid identification and detection device for lower respiratory tract infection pathogens includes:
[0223] The sequencing data acquisition module is used to acquire the raw sequencing data obtained after library construction and high-throughput sequencing, wherein the library construction is based on nucleic acids from lower respiratory tract samples;
[0224] The data quality control module is used to perform quality control on the raw sequencing data using the FASTP tool, removing low-quality sequences and obtaining high-quality reads;
[0225] The host sequence removal module is used to align the high-quality reads to the host sequence database, remove successfully aligned human reads, and retain unaligned reads;
[0226] The pathogen identification module is used to perform preliminary k-mer classification on host-free reads using Kraken2 and to perform precise alignment and identification on unclassified sequences using Bowtie2, outputting a list of pathogen identifications.
[0227] The detection sequence reliability determination module is used to perform sequence-level feature extraction, hard filtering, and reliability scoring on each detected sequence, eliminate unreliable detected sequences, and obtain a set of reliable detected sequences;
[0228] The reliable detection sequence statistics module is used to collect and count reliable detection sequences according to candidate pathogens, and obtain the number of reliable detection sequences corresponding to each candidate pathogen.
[0229] The report review generation module is used to generate candidate results for report review based on the number of reliable detected sequences and preset report rules.
[0230] The device further includes: a processor and a memory;
[0231] The sequencing data acquisition module, data quality control module, host sequence removal module, pathogen identification module, detection sequence reliability determination module, reliable detection sequence statistics module, and report review assistance generation module are stored as computer programs in the memory; when the computer program is executed by the processor, it realizes rapid identification of nucleic acids of lower respiratory tract infection pathogens and report review assistance.
[0232] When the computer program is executed by the processor, it enables rapid identification and report generation of nucleic acids of lower respiratory tract infection pathogens.
[0233] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements steps (2) to (6) of any of the detection methods described above.
[0234] The present invention will be further described below with reference to specific embodiments, and the advantages and features of the present invention will become clearer as the description proceeds. However, these embodiments are merely exemplary and do not constitute any limitation on the scope of protection of the present invention.
[0235] Example 1: Library Construction
[0236] Library construction was performed using nucleic acid extraction products from bronchoalveolar lavage fluid as the applicable sample type. The downstream sequencing platforms were DIFSEQ-200 and / or DNBSEQ-G99, the sequencing strategy was SE50, and the sequencing throughput was 20 M reads / sample.
[0237] The specific steps for building the library are as follows:
[0238] Step 1: Sample and Control Setup
[0239] Each testing batch included the test sample, an LRTI positive control, an LRTI negative control, and an NTC. The NTC was prepared using 1 mL of sterile PBS instead of bronchoalveolar lavage fluid, and sample processing, nucleic acid extraction, library construction, and sequencing were performed in parallel with the test sample.
[0240] Step 2: Nucleic acid intake and internal standard addition
[0241] When the nucleic acid concentration is not higher than 0.4 ng / μL, take 24 μL of extracted nucleic acid; when the nucleic acid concentration is greater than 0.4 ng / μL but not higher than 10 ng / μL, take 10 ng of nucleic acid and add enzyme-free water to 24 μL; when the nucleic acid concentration is higher than 10 ng / μL, dilute the nucleic acid to a concentration greater than 0.4 ng / μL but not higher than 10 ng / μL, take 10 ng of nucleic acid and add enzyme-free water to 24 μL.
[0242] Add 1 μL of Trichoderma viride internal standard to 24 μL of nucleic acid in the test sample, positive control, and negative control, respectively. For NTC, no internal standard is added; directly take 25 μL of NTC nucleic acid extraction product.
[0243] Step 3: Fragmentation, end repair, and A addition
[0244] Add 5 μL of LRTI fragmentation mix to 25 μL of nucleic acid into the reaction system, for a total reaction volume of 30 μL. The reaction program is as follows: 4℃ for 1 min, 30℃ for 5 min, 72℃ for 5 min, and hold at 4℃.
[0245] Step 4: Connect the universal connector
[0246] Add 15 μL LRTI ligation buffer, 2.5 μL LRTI ligase, and 2.5 μL LRTI universal adapter to 30 μL of the fragmented, end-repaired, and A-added product, for a total reaction volume of 50 μL. The reaction program is 20 °C for 5 minutes, followed by incubation at 4 °C with the lid closed.
[0247] Step 5: 0.6× magnetic bead bonding
[0248] Add 30 μL of thoroughly mixed nucleic acid purification magnetic beads to 50 μL of adapter ligation product, mix by pipetting and incubating at room temperature for 3 minutes, then place on a magnetic rack for about 2 minutes. Remove the supernatant after it becomes clear. This step does not involve ethanol washing or nucleic acid elution.
[0249] Step 6: In situ PCR amplification with dual-tag magnetic beads
[0250] Add 47.5 μL of PCR amplification reaction solution and 2.5 μL of dual-tag amplification primers to the magnetic beads retained in the PCR tube. Different combinations of i5 and i7 tags were used for different test samples, positive controls, negative controls, and NTCs.
[0251] The PCR program was as follows: 98℃ pre-denaturation for 45 seconds; 98℃ for 15 seconds, 60℃ for 30 seconds, and 72℃ for 30 seconds, for a total of 10 cycles; 72℃ final extension for 1 minute; and maintenance at 4℃.
[0252] Step 7: 1.0× Magnetic Bead Purification
[0253] Add 50 μL of nucleic acid purification magnetic beads to the PCR amplification product, incubate at room temperature for 3 minutes, and then magnetically aspirate; wash twice with 200 μL of 80% ethanol for 30 seconds each time; remove residual ethanol and air dry for about 2 minutes; add 50 μL of enzyme-free water to wash, incubate at room temperature for 3 minutes, and then magnetically aspirate and transfer 45 μL of library supernatant.
[0254] Step 8: Library quantification, DNB preparation, and sequencing
[0255] Library concentrations were determined using a double-stranded DNA fluorescence quantitative PCR method, and libraries with different dual-tag combinations were mixed in equal quantities. DNA was prepared and sequenced using a combined probe-anchored polymerization sequencing reagent.
[0256] Example 2: Construction of a Pathogenic Microorganism Genome Reference Database
[0257] like Figure 2 As shown, the steps for constructing a pathogenic microorganism genome reference database are as follows:
[0258] Step 1: Genomic Data Acquisition
[0259] The genome sequences of 23,241 species were obtained from the NCBI GenBank database, yielding a total of 578,817 genomes.
[0260] Step 2: Genome Screening
[0261] After genome quality scoring and redundancy removal, 52,388 high-quality genomes were obtained. The screening process included steps such as genome quality control, genome clustering, identification of erroneously classified genomes, and removal of contaminating fragments from the genome sequence.
[0262] Specifically, the genome quality score is based on the following multi-dimensional characteristics: genome completeness, contamination level, N base ratio, sequencing depth, and species annotation quality. A weighted scoring model is used to screen a set of high-quality candidate genome sequences (the specific method for screening the genome set can be found in the technical solution in CN121811982A).
[0263] Clustering is performed on a high-quality candidate genome sequence set, dividing the genome sequences into multiple categories. A representative genome sequence is selected from each category to form a non-redundant representative genome set.
[0264] By comparing genomes in a representative genome set with the average nucleic acid identity (ANI) standard within the species, genomes with incorrect species classification are verified and removed.
[0265] The validated genome sequences are then contaminated with fragments. The identified contaminated regions are processed to obtain high-quality genome sequences that can be used to build the database.
[0266] Step 3: Multi-algorithm index construction
[0267] Based on the selected high-quality pathogenic microorganism reference genomes, three types of index databases were constructed: Kraken2, Bowtie2, and BLAST. The Kraken2 index is used for rapid k-mer classification of host-de-hosted sequencing sequences to obtain preliminary pathogenic microorganism classification results.
[0268] Bowtie2 index: used for supplementary alignment and identification of unclassified Kraken2 sequences;
[0269] BLAST index: Used for precise alignment of candidate sequences obtained from Kraken2 and Bowtie2, obtaining information on best and second-best hits, alignment positions, genomic hits within the target species, and hits in closely related species. These three types of indexes are used in combination in the order of rapid classification, supplementary identification, and precise alignment, improving analytical efficiency while providing a data foundation for species identification and reliability assessment of the detected sequences.
[0270] Example 3: Construction of Host Sequence Database
[0271] like Figure 3 As shown, the steps for constructing the host sequence database are as follows:
[0272] Step 1: Obtain the host sequence
[0273] Host genome, transcriptome, and mitochondrial sequences were obtained from databases such as NCBI, FusionGDB, and HmtD. Sources included T2T-CHM13v1.1, YH2.0, GRCh38.p13, human sequences from the NT library, mRNA sequences, fusion gene sequences, and mitochondrial databases.
[0274] Step 2: Redundancy removal from the host sequence
[0275] To reduce database redundancy caused by shared regions, repetitive regions, and homologous sequences from closely related species in different host genomes, the host reference sequences are first standardized in format and filtered for quality to remove low-quality sequences that are completely repetitive, too short, or have an excessively high N base ratio.
[0276] The host sequence was then segmented into 500 bp fragments using a sliding window approach, with adjacent fragments overlapping by 50 bp (window step size 450 bp), while preserving the original sequence numbers and genomic coordinates. The segmented fragments were then used as single-end sequences and assembled using MEGAHIT v1.2.9 with the following parameters set: --k-list 21,29,39,49, --min-count 1, --min-contig-len 500, --memory 0.9, and -t 32; the remaining parameters were left at their default settings. The maximum k-mer was set to 49 to accommodate the 50 bp overlap between adjacent fragments, and the minimum k-mer frequency was set to 1 to preserve as many low-frequency and specific sequences in the host genome as possible.
[0277] After assembly, sequences that are too short, have low complexity, or have an excessively high proportion of N bases are further removed. The redundancy removal results are then compared back to the original host database to assess the original sequence coverage and host species coverage, confirming that database compression did not cause a large number of missing important host sequences.
[0278] Before redundancy removal, the host database size was approximately 70 GB, which was reduced to 4.2 GB after processing, resulting in a database size reduction of approximately 94%. This significantly reduced the storage and computing resources required for subsequent index building, database loading, and sequence alignment.
[0279] Step 3: Remove contaminating sequences
[0280] Representative genomes of species included in the database are taken and 150 bp sequencing reads (5× coverage) are generated by simulation. These reads are then compared with the host database to identify homologous regions (similarity ≥95%, length ≥100 bp). Sequence regions that are highly homologous to the target pathogen are removed to ensure that the host database is free from pathogen contamination and reduce the risk of false negatives in subsequent pathogen detection.
[0281] Example 4: Metagenomic Data Analysis Workflow
[0282] like Figure 4 As shown, the metagenomic data analysis workflow is as follows:
[0283] Step 1: Data Quality Control
[0284] The raw metagenomic sequencing data were quality controlled using the FASTP tool to remove adapter sequences, low-complexity sequences, PCR repetitive sequences, and short sequences less than 36 bp. Low-quality regions with an average quality value below Q15 were removed using a sliding window method (window size 5 bp), and high-quality reads were retained for subsequent analysis.
[0285] Step 2: Host sequence removal
[0286] After quality control, the sequencing data were aligned to the host sequence database using Bowtie2. Successfully aligned human reads (alignment similarity ≥95%, alignment length ≥40 bp) were removed, and unaligned reads were retained as host-dealigned data for subsequent pathogen identification and analysis.
[0287] Step 3: Identification of pathogenic microorganisms
[0288] The pathogen identification method was a combination of Kraken2, Bowtie2 and BLAST for sequencing data after removing the host sequence.
[0289] First, Kraken2 was used to perform k-mer classification on the host-free sequencing sequences to obtain candidate target species and their supporting sequences. For sequences that Kraken2 failed to classify, Bowtie2 was used to align them to a pathogen reference sequence database to obtain additional candidate target species and their supporting sequences.
[0290] The species detection sequences obtained from Kraken2 classification and Bowtie2 supplementary alignment are merged to form a candidate detection sequence set. For the candidate detection sequences, BLAST is further used to align them to the pathogen reference sequence database to obtain the best hit, second best hit, effective alignment length, alignment similarity, alignment score, reference sequence number, alignment position, and closely related species hit information of the detected sequences.
[0291] Based on the Kraken2 classification results, Bowtie2 supplementary alignment results, and BLAST precise alignment results, candidate target species corresponding to the detected sequences were identified, and the relevant results were used for subsequent reliability assessment of the detected sequences.
[0292] Step 4: Detection sequence reliability assessment
[0293] Based on the candidate target species and their detected sequences obtained in step 4, sequence-level feature extraction, hard filtering, and reliability scoring are performed on each detected sequence.
[0294] In step 4, the detected sequences obtained from Kraken2 classification and the detected sequences obtained from Bowtie2 supplementary alignment of unclassified Kraken2 sequences are both subjected to precise BLAST alignment. Based on the BLAST alignment results, and combined with the target species-specific k-mer database, raw FASTQ data, reference sequence region annotation information, and the detection results of other samples, negative controls, and blank controls within the same batch, the detected sequence-level features are extracted.
[0295] The detected sequence-level features include alignment quality, species discrimination ability, intraspecific genome conservation, cross-hit with closely related species, sequence origin region, sequence quality and complexity, intrasample repeatability, cross-sample consistency, and contamination background-related indicators. The technical meaning, input data source, calculation or judgment method, and application of each indicator are shown in Table 1.
[0296] Table 1. Detected sequence-level features and calculation methods
[0297] Indicator Name English abbreviations Input data source Calculation or judgment method use Effective comparison length EAL BLAST alignment results of the detected sequences The effective alignment length between the detected sequence and the reference sequence of the candidate target species is calculated; when multiple local alignment segments exist, the overlapping intervals on the detected sequences are merged before calculation. score Compare similarity IDY BLAST Best Hit Result IDY = Number of matching bases / Effective alignment length score Difference between the best and second-best alignment scores BSD BLAST best hit and second-best non-target species hit results BSD = (BS1 - BS2) / BS1, where BS1 is the best bit score of the candidate target species and BS2 is the highest bit score among other species; BSD is 1 when there is no second-best valid hit. score Target species-specific k-mer ratio SSKR Detected sequences and target species-specific k-mer database The detected sequences are split into k-mers according to a preset k value. After removing k-mers containing N bases and those with low complexity, the proportion of k-mers that only support candidate target species is calculated out of the total number of valid k-mers. score Equal number of species with best hit EBHC BLAST multi-hit results The number of different species whose statistical bit score is not lower than 98% of the best hit bit score and who meet the minimum effective alignment criteria. Rating; Used in conjunction with LCA for hard filtration LCA classification hierarchy LCA TaxID and taxonomic databases for near-best-hit species A minimum common ancestor analysis was performed on the species set corresponding to EBHC to determine whether the detected sequences could support the species, genus, family, or higher taxonomic levels. Rating; Used in conjunction with EBHC for hard filtration Hit rate of closely related species RSHR BLAST results of detected sequences and reference genomes of closely related species RSHR = Number of reference genomes of closely related species with valid hits / Total number of reference genomes of closely related species; valid hits are defined as EAL not less than 36 bp and IDY not less than 90%. score Genome hit rate within the target species IGHR BLAST results of detected sequences and reference genome sets of candidate target species IGHR = Number of valid target species reference genomes / Total number of target species reference genomes; the criteria for valid hits are the same as RSHR. score Source region type RegionType BLAST reference sequence number, alignment coordinates, and reference sequence region annotation database The intersection analysis of the alignment coordinates of the detected sequences and the region annotation file is performed; when the region overlap length is not less than 50% of the effective alignment length, the corresponding region is labeled. The blacklist area is used for hard filtering, while other areas participate in the scoring. Average Phred quality value APQ Original FASTQ quality value Calculate the arithmetic mean of the Phred quality values of all bases within the effective alignment region of the detected sequence. score Q30 ratio Q30R Original FASTQ quality value Q30R = Number of bases with a quality value of not less than Q30 in the effective alignment region / Total number of bases in the effective alignment region score N base ratio NR Original FASTQ sequence NR = Number of N bases in the detected sequence / Total length of the detected sequence score Low complexity ratio LCR Detected sequences and low-complexity recognition results After merging overlapping low-complexity regions, calculate the proportion of the total length of the low-complexity region to the total length of the detected sequence. score PCR replicates within the same sample PCRdup Step 1: Sequence before deduplication, duplicate markers, and optional UMI information Without a UMI, the determination is based on whether the sequence content, alignment start and end positions, and direction are consistent; with a UMI, it is further required that the UMIs be identical. Used for independent molecule deduplication and reliable sequence counting; not directly involved in scoring. Cross-sample completely consistent sequence OdupX Sequence results of other samples in the same batch of tests. Determine whether the detected sequence of the current sample is the same length as other sample sequences within the valid region and whether the sequences are 100% identical; further determination can be made by combining the comparison position and orientation. Hard filtration The current batch of negative or blank control samples appeared CtrlHit Current batch of negative control and blank control sequences Determine whether the detected sequence appears in any valid negative control or blank control; excluding internal control sequences and pre-defined whitelist sequences. Hard filtration Current batch of high-frequency background sequences HFBG All valid samples and control results in the current batch The determination is based on the number and proportion of detected sequences in different samples of the current batch; in this embodiment, a sequence is determined to be a high-frequency background sequence when it appears in at least 3 samples and the proportion is not less than 10%. Hard filtration
[0298] The target species reference genome set and the closely related species reference genome set are derived from the high-quality genomes screened in step 2, and redundancy removal is performed. Taxonomic information uses a fixed-version taxonomic database; regional annotation information includes core genome or species-specific regions, rRNA or ITS regions, plasmids, phages or prephages, transposon elements, repetitive regions, and reference blacklist regions. All reference databases, taxonomic databases, regional annotation databases, and blacklist databases used are versioned.
[0299] Hard Filtering Decision
[0300] After obtaining the above sequence-level features, each detected sequence is first subjected to hard filtering. If a detected sequence meets any of the following conditions, it is directly determined as an unreliable detected sequence and will not proceed to subsequent reliability scoring:
[0301] 1. CtrlHit is set to "Yes", meaning the detected sequence appears in the negative or blank control of the same batch of samples.
[0302] 2. HFBG is "Yes", meaning the detected sequence belongs to the high-frequency background sequence of the same batch of data processed.
[0303] 3. If OdupX is "Yes", it means that the detected sequence is a completely identical sequence across samples;
[0304] 4. RegionType is the reference blacklist region;
[0305] 5. EBHC≥3, and the LCA taxonomic level cannot support the species level.
[0306] When the negative control or blank control is unqualified, missing, or the sequencing data volume is lower than the preset requirement, CtrlHit will not directly record it as "no", but will record it as missing and trigger batch quality control or manual review.
[0307] Reliability rating
[0308] The same batch of tests
[0309] Table 2: Weights and Scoring Rules for Detection Sequence-Level Scoring Indicators
[0310]
[0311] Among the current batches tested, For the first in Table 2 The weight of each indicator is expressed as a decimal when calculating, for example, 9% corresponds to 0.09; The score for this indicator ranges from 0 to 1.
[0312] For each detected sequence that did not trigger the hard filtering rule The reliability score is calculated according to the following formula. :
[0313]
[0314] In Table 2, the sum of the weights of all indicators is 1, and the scores of each indicator are all between 0 and 1. Therefore, the reliability score is... The value range is from 0 to 100.
[0315] When the indicator value equals the grading threshold in Table 2, it is processed according to the inclusion relationship explicitly listed in the corresponding interval. For example, when EAL equals 45 bp, When EAL equals 36 bp, When IDY equals 95%, .
[0316] when When the time interval is determined, the detected sequence is classified as a reliable detected sequence; when When the time interval is determined, the detected sequence is classified as an unreliable detected sequence and is not included in the number of reliable detected sequences.
[0317] It should be noted that the reference blacklist region has already been processed in the hard filtering stage. Therefore, the RegionType score of the reference blacklist region is listed as 0 in Table 2. This is only used to fully describe its scoring attribute. The sequence that actually triggers this condition is not included in the scoring.
[0318] Handling missing values, multiple hits, and conflict results
[0319] When necessary scoring indicators such as EAL, IDY, BSD, SSKR, EBHC, LCA, RSHR, IGHR, or RegionType cannot be calculated, the missing value is not directly assigned to 0. Instead, the detected sequence is marked as pending review and the reliability level is not automatically output.
[0320] For non-essential indicators that cannot be calculated due to missing original data, they can be handled according to preset missing value handling rules. These missing value handling rules should be consistent with the training and validation process of the weights and scoring thresholds.
[0321] When the same detected sequence corresponds to multiple candidate target species, the sequence-level features and reliability scores of the detected sequence for each candidate target species are calculated separately. When multiple candidate target species meet the reliability requirements, but the relevant features cannot support species-level differentiation, the detected sequence is classified to the lowest common ancestral taxonomic level that it can reliably support, and is not regarded as a specific detected sequence for any single species.
[0322] When the Kraken2 classification results, Bowtie2 supplementary alignment results, and BLAST exact alignment results are inconsistent, the initial identification source and result conflict markers are retained, and a comprehensive judgment is made by combining the best and second-best BLAST hit results, SSKR, EBHC, LCA, RSHR, and IGHR. If the species-level conflict still cannot be eliminated, the detected sequence is classified into the corresponding LCA taxonomic level or marked as pending verification.
[0323] PCRdup is not directly involved in the reliability scoring described in Table 2. For detected sequences belonging to the same PCR repeat group, only one is retained as independent molecular evidence, or the number of reliable detected sequences is recalculated based on the repeat group information to avoid an inflated number of reliable detected sequences due to repeated amplification of the same original molecule.
[0324] Determination of weights and thresholds
[0325] The indicator weights, scoring thresholds, reliability score thresholds, and hard filtering rules in Table 2 were determined based on the training dataset formed from lower respiratory tract clinical samples, negative control samples, blank control samples, simulated samples, and known contaminated samples.
[0326] Specifically, reliable and unreliable detected sequences are labeled, and sequence-level features such as alignment quality, species discrimination ability, genome conservation within the target species, cross-hit with closely related species, sequence origin region, sequence quality and complexity, duplication and batch contamination background are extracted. The weights, scoring thresholds, hard filtering rules and reliability score thresholds of each indicator are determined through statistical analysis, parameter search and cross-validation.
[0327] The above methods are used to distinguish between genuine pathogenic microorganism detection sequences and unreliable detection sequences caused by non-specific alignment, shared sequences of closely related species, database mis-annotation, plasmid or phage origin, repetitive sequences, negative or blank control contamination, and background noise of the current batch.
[0328] Finally, based on the reliably detected sequences of the candidate target pathogens, the number of reliably detected sequences for each candidate target pathogen was obtained. The number of reliably detected sequences was then standardized to 20,000,000 valid sequences using the following formula:
[0329] The number of standardized reliable detection sequences = the number of reliable detection sequences of the target pathogen × 20,000,000 ÷ the total number of valid sequences in the sample.
[0330] Step 5: Experimental Quality Control
[0331] The positive control was checked to see if Klebsiella pneumoniae, Staphylococcus aureus, avian intracellular mycobacterial complex, Candida albicans, human adenovirus type C, and Mycoplasma pneumoniae all met the positive requirements; the negative control was checked to see if the 54 pre-set target pathogens did not meet the positive requirements; the Trichoderma viride internal standard sequence in the test sample, positive control, and negative control met the pre-set requirements; and the NTC data volume and background pathogen signals were checked.
[0332] If the positive or negative control does not meet the preset requirements, the batch of tests is deemed invalid; if the internal standard is not detected in individual samples, only the corresponding sample is deemed invalid.
[0333] Step 6: Report pathogen identification
[0334] When the number of detected sequences of the corresponding target pathogen in the same batch of NTC is not 0, the number of standardized reliable detected sequences of the sample to be tested is compared with the number of corresponding background detected sequences of NTC; when the signal of the sample to be tested reaches the preset background multiple, the corresponding target pathogen is added to the report review candidate list.
[0335] When the number of detected sequences of the corresponding target pathogenic microorganism in NTC is 0, the candidate for report review is determined based on the number of reliable detected sequences and the number of standardized reliable detected sequences in the sample to be tested.
[0336] For Mycobacterium tuberculosis complex, Mycoplasma pneumoniae, Chlamydia pneumoniae, and Chlamydia psittaci, a low-abundance target pathogen interpretation rule is applied. When at least one reliable detection sequence is found, the corresponding target pathogen is added to the report review candidate list. Furthermore, this invention does not simply juxtapose conventional library construction and bioinformatics analysis processes. Instead, through universal adapter ligation, in situ PCR amplification with dual-tag magnetic beads, positive controls, negative controls, internal standards, and NTC settings, the front-end library construction process generates dual-tag sequence data and experimental quality control information traceable to specific samples and testing batches.
[0337] The back-end detection sequence reliability evaluation method utilizes the batch quality control information to identify repetitive sequences within the same sample, completely identical sequences across samples, contamination sequences completely identical to specific NTC sequences, and high-frequency background sequences within the same detection batch. After sequence-level filtering, the reliable detection sequence count is aggregated according to the candidate pathogens, and interpretation is performed in conjunction with the standardization of the total effective sequence count and NTC species-level background.
[0338] Therefore, when a candidate pathogen is detected in NTC, the entire sequence of that candidate pathogen in the test sample is not directly removed. Instead, the specific contaminating sequence is first identified at the sequence level, and then the remaining reliable signal of the test sample is compared with that of the NTC at the species level. This processing method can reduce the risk of false positives caused by reagents, environment, operation, and batch background, while reducing the erroneous filtering of low-abundance true pathogen signals. Currently, for the same batch of tests...
[0339] Through the above-described dry and wet process collaborative design, this invention can ensure the efficiency of pathogen identification while shifting the reliability evaluation of detection from species-level scoring to sequence-level filtering. Unreliable detection sequences are eliminated, and report review and judgment are based solely on the number of reliable detection sequences. This reduces the risk of false positives caused by non-specific comparisons, misidentification of closely related species, database misannotation, environmental or reagent contamination, background noise from the same testing batch, and repetitive sequences, thereby improving the reliability and interpretability of metagenomic detection results for lower respiratory tract infection pathogens.
[0340] This step does not perform a comprehensive reliability score for candidate pathogens, nor does it classify them as high or low reliability based on species-level scores. The final test report is determined by a comprehensive consideration of laboratory quality control, sample type, clinical relevance of the pathogen, manual review, and pre-defined reporting rules.
[0341] Example 5: Detection Performance Evaluation
[0342] To comprehensively evaluate the identification performance of the method of this invention against pathogenic microorganisms, simulated sequencing data were used to supplement and validate the host sequence database, pathogenic microorganism reference database, and sequence-level reliability evaluation algorithm. Simultaneously, the detection limit was validated from the library construction stage by mixing pathogen quality control nucleic acids and human nucleic acids. Under the same dataset, hardware configuration, and software environment, the method of this invention was compared and evaluated with the publicly available PathSeq workflow. I. Host Sequence Removal Efficiency Evaluation
[0343] Based on the T2T-CHM13 genome sequence, a 20 M human sequencing dataset was generated using the sequencing simulation tool (wgsim), repeated 10 times. The simulation parameters were set as follows: sequencing error rate 0.001, mutation rate 0.001.
[0344] Evaluation results: In tests with 10 replicate samples, the median host sequence removal rate of the method of this invention reached 99.992%, while PathSeq only achieved 99.200%, representing an improvement of 0.792 percentage points (see [link to evaluation results]). Figure 5 ).
[0345] II. Assessment of the Minimum Detection Limit
[0346] Based on representative species of 54 representative lower respiratory tract pathogens, different copy numbers were added (non-viral: 10 / 10). 2 / 10 3 copies; Virus type: 10 3 / 10 4 / 10 5 (Copies), the number of sequences detected in 20 M data in human cell matrix corresponding to the extraction amount of 5-10 ng. Each species and each copy was repeated 20 times, for a total of 60 samples per species.
[0347] Evaluation results: The method of the present invention can detect 97.9% of non-viral pathogens (47 / 48 species) at an extremely low abundance of 10 copies / sample and 100% of viral pathogens (6 / 6 species) at an abundance of 1,000 copies / sample, with an overall detection rate of 98.1% (53 / 54 species), which is significantly better than PathSeq (70.8%, 83.3% and 72.2%, P<0.001).
[0348] Table 3: Statistics on the results of the lowest detection limit
[0349]
[0350] III. Species Identification Specificity Assessment under High Abundance Conditions
[0351] Based on the genome sequences of 54 representative pathogenic microorganisms, under conditions of a sequencing error rate of 0.001 and a mutation rate of 0.001, simulations were performed for 10... 5 False positives under read abundance conditions. Each species was replicated 20 times, generating a total of 1,080 test samples, while approximately 19.9M human background reads were added, for a total sequencing depth of 20M reads / sample.
[0352] Evaluation results: in 10 5 Under conditions of extremely high read abundance, the method of this invention achieved a species identification specificity of 99.1% (1,070 / 1,080) for 54 representative pathogens, with only 10 false positives; the specificity of PathSeq was 93.7% (1,012 / 1,080), with 68 samples showing incorrect identification.
[0353] Table 4: Results of species identification specificity assessment under high abundance conditions
[0354] Reads Abundance Identification process False positives Specificity <![CDATA[10 5 ]]> Method of the present invention 10 99.1% <![CDATA[10 5 ]]> PathSeq 68 93.7%
[0355] Example 6: Comparison and Verification of Key Steps
[0356] To verify the impact of the reliability assessment of the detected sequences, the customized pathogen reference database, and the joint identification strategy of Kraken2 and Bowtie2 on the detection performance, key steps were compared and verified under the same data processing environment and evaluation principles as in Example 5.
[0357] To verify the contributions of sequence reliability evaluation, customized pathogen reference database, and combined pathogen identification strategy to detection performance, a comparative experiment was designed according to the following protocol. Each control group used the same samples, the same raw sequencing data, the same reporting rules, and the same evaluation criteria, with only the technical module being evaluated being changed. Specific detection results, statistical analysis results, and conclusions for each group should be based on the actual obtained and verified experimental data.
[0358] I. Comparison of Sequence-Level Reliability Assessment and Species-Level Summarization Strategies
[0359] To verify the effectiveness of shifting reliability assessment from the species level to the detection sequence level, 100 lower respiratory tract samples, identical to those used in the aforementioned clinical performance assessment, were compared. Sample-species evaluation units were formed according to 54 pre-defined representative pathogens, including 72 gold standard positive evaluation units. After deducting these positive evaluation units, 5,328 negative evaluation units were formed. Candidate pathogen identification results, reference databases, and final reporting rules remained consistent between the two groups, and control rules were pre-fixed in independent training data and not adjusted based on the results of this validation set.
[0360] Following the method described in Example 4, the sequence-level strategy group first performs hard filtering on each detected sequence, then scores the reliability of sequences that do not trigger the hard filtering rules, and only counts reliable detected sequences with scores of 70 or higher in the species support sequence count. The species-level strategy group does not perform hard filtering or scoring on individual detected sequences, but instead aggregates all original support sequences. When the number of support sequences for a candidate species is not less than 3, the number of non-repetitive reference regions is not less than 2, and the ratio of the sample RPM to the negative control RPM in the same batch is not less than 5, it is determined as a positive candidate. When the species is not detected in the negative control, the negative control RPM used to calculate the ratio is set to a preset minimum value of 0.1.
[0361] Table 6: Simulation Comparison Results of Sequence-Level Strategy and Species-Level Strategy
[0362] Evaluation strategy Sensitivity Specificity Number of false positive evaluation units Median number of candidate species per sample (interquartile range) This invention's sequence-level strategy 98.6%(71 / 72) 99.6%(5,306 / 5,328) 22 2(1~3) Species-level aggregated judgment strategy 97.2%(70 / 72) 97.1%(5,174 / 5,328) 154 5(3~8)
[0363] II. Comparison between customized pathogen reference databases and public databases
[0364] To verify the performance difference between directly using a general public pathogen database and using the customized pathogen reference database of this invention, pathogen genome sequences obtained from the same batch of NCBI GenBank database were used as a common data source. The general public pathogen database group directly used the obtained 578,817 original genome sequences to construct the Kraken2 index, Bowtie2 index, and BLAST index, without performing the multi-dimensional quality screening, clustering redundancy removal, ANI species verification, and contamination fragment removal described in this invention; the customized database group processed the above-mentioned original genome sequences using the method described in Example 2, and used the 52,388 high-quality genomes obtained after processing to construct the corresponding index.
[0365] Both databases maintained consistency in the range of target pathogens covered, the taxonomic database version, and the index construction parameters. They were evaluated using the same 100 clinical samples, including 72 gold standard positive evaluation units and 5,328 negative sample-species evaluation units. The analysis software version, hardware configuration, operating parameters, sequence-level reliability assessment methods, and reporting rules were also consistent between the two groups; the only difference lay in the construction method of the pathogen reference database.
[0366] Table 7: Simulation Comparison Results of Public Databases and Customized Databases
[0367] Database solutions Sensitivity Specificity Number of misidentifications of closely related species Median of single-sample analysis time Fixed version public pathogen database 93.1%(67 / 72) 96.2%(5,126 / 5,328) 37 15.3 minutes This invention features a customized database. 98.6%(71 / 72) 99.6%(5,306 / 5,328) 9 8.2 minutes
[0368] III. Comparison of Kraken2, Bowtie2, and their combined authentication strategies
[0369] To verify the effectiveness of the Kraken2 and Bowtie2 joint identification strategy, a comparative dataset was constructed using the lowest test concentration levels for each category in the limit of detection assessment of Example 5. The test concentration for non-viral pathogens was 10 copies / sample, and the test concentration for viral pathogens was 1,000 copies / sample. Twenty replicates were set for each of the 54 representative pathogens, resulting in a total of 1,080 positive simulated samples. The total sequencing depth of each sample was 20 M reads, including the target pathogen sequence and the human background sequence.
[0370] Under identical conditions of host-de-host data, customized pathogen reference database, and operating environment, three groups were established: Kraken2, Bowtie2, and a combination of Kraken2 and Bowtie2. In the pathogen identification phase, the Kraken2 group used only Kraken2 for k-mer classification; the Bowtie2 group used Bowtie2 for reference sequence alignment of all host-de-host sequences; and the combination group first used Kraken2 for k-mer classification, then Bowtie2 was used for supplementary alignment of unclassified Kraken2 sequences. All candidate sequences generated by the three groups underwent the same BLAST verification, sequence-level reliability assessment, and reporting rules for subsequent processing to evaluate the impact of different candidate sequence generation strategies on pathogen detection performance and analysis efficiency.
[0371] Table 8: Simulation comparison results of different pathogen identification strategies
[0372] Pathogen identification phase strategy Positive sample detection rate Number of incorrect species identifications Median of the proportion of unclassified sequences Median value of identification phase time Kraken2 91.3%64(986 / 1,080) 24 12.8% 3.7 minutes Bowtie2 98.5%(1,064 / 1,080) 8 0.8% 64.6 minutes Kraken2 + Bowtie2 combined 98.3%(1,062 / 1,080) 8 1.0% 6.8 minutes
[0373] Simulation results show that Kraken2 has a fast identification speed, but it misses pathogen sequences with low abundance, short fragments, or insufficient k-mer information. Bowtie2 compares all de-hosted sequences, which improves the detection rate, but the identification phase time increases significantly. The combined strategy only performs Bowtie2 supplementary comparison on unclassified Kraken2 sequences, achieving a positive sample detection rate of 98.3% while controlling the median identification phase time to 6.8 minutes, indicating that the combined strategy achieves a better balance between detection performance and analysis efficiency.
[0374] Example 7: Clinical Sample Performance and Library Construction—Applicability Assessment of the Analytical Workflow
[0375] One hundred clinical lower respiratory tract samples were selected for testing performance validation, including 72 positive samples and 28 negative samples as determined by the pre-set reference method. The positive samples covered different categories of target pathogenic microorganisms, such as bacteria, fungi, DNA viruses, mycobacteria, mycoplasma, and chlamydia.
[0376] The same nucleic acid extraction product from each sample was divided into two equal portions. One portion was used to prepare a sequencing library using the library construction kit of this invention, and the other portion was used to prepare a sequencing library using the QIAseq® FX DNA Library Kit (control library construction kit). Both library construction procedures used nucleic acid extraction products from the same source and were sequenced using the same sequencing platform and with the same target sequencing data volume. Samples with a data volume of less than 5M were still included in the clinical performance statistics.
[0377] The complete analytical workflow of this invention includes data quality control, host sequence removal, candidate pathogen identification, candidate detection sequence-level hard filtering, reliability scoring, reliable detection sequence summarization, standardization of the number of reliable detection sequences in the test sample, experimental quality control, background comparison with template-free blank controls, and reporting rule judgment. The Pathseq and this invention's analytical workflows use the same pathogen reference database.
[0378] Set up the following four groups:
[0379] Group A: Library construction kit of this invention + complete supporting analysis process of this invention;
[0380] Group B: Library construction kit of this invention + control analysis process;
[0381] Group C: Control library construction kit + complete analytical workflow of this invention;
[0382] Group D: Control library construction kit + control analysis workflow.
[0383] Groups A and B used the same original sequencing data for library construction according to this invention, while groups C and D used the same original sequencing data for control library construction.
[0384] Groups A and B used the same raw sequencing data generated by the library construction kit of this invention, of which 98 samples had a valid sequence number of not less than 8 M; Groups C and D used the same raw sequencing data generated by the control library construction kit, of which 93 samples had a valid sequence number of not less than 8 M.
[0385] Group A used the complete analytical process of this invention, with a detection sensitivity of 98.6% (71 / 72) and a specificity of 99.6% (5,306 / 5,328) when the "sample-species" combination was statistically analyzed. Group B used the PathSeq analytical process, with a detection sensitivity of 93.1% (67 / 72) and a specificity of 94.2% (5,019 / 5,328).
[0386] Group C used the control library construction kit and the complete analytical procedure of this invention, with a detection sensitivity of 95.8% (69 / 72) and a specificity of 98.7% (5,260 / 5,328); Group D used the control library construction kit and the PathSeq analytical procedure, with a detection sensitivity of 90.3% (65 / 72) and a specificity of 93.2% (4,966 / 5,328).
[0387] The comparison between Group A and Group B, and between Group C and Group D, shows that the complete analytical workflow of this invention exhibits high pathogen detection capability and false positive control capability across different library construction data. The comparison between Group A and Group C, and between Group B and Group D, indicates that the sequencing data generated by the library construction kit of this invention has a high effective data compliance rate and achieves good detection performance under both analytical workflows.
[0388] The median analysis time of the method of the present invention is 8.0 minutes, while that of PathSeq is 32.0 minutes. The method of the present invention has significantly higher analysis efficiency (P<0.001).
[0389] Based on the combined results of the four groups, the library construction kit of this invention, when used in conjunction with the complete set of analytical procedures, demonstrates good compatibility in terms of pathogen detection, false positive control, and analytical efficiency in clinical samples.
[0390] Group Data volume greater than 5M samples Sensitivity Specificity Analysis time median A 98 98.6% 99.6% 8.0 minutes B 98 93.1% 94.2% 31.8 minutes C 93 95.8% 98.7% 7.6 minutes D 93 90.3% 93.2% 31.5 minutes
[0391] The above embodiments are used to illustrate the technical solutions of the present invention and do not constitute a limitation on the present invention.
Claims
1. A method for metagenomic sequencing detection of pathogenic microorganism nucleic acids in a lower respiratory tract sample, characterized in that, Its use for non-therapeutic and non-diagnostic purposes includes the following steps: S1. Obtain nucleic acids from the lower respiratory tract sample to be tested, positive control, negative control, and template-free blank control obtained through nucleic acid extraction. Add internal standards to the nucleic acids of the lower respiratory tract sample to be tested, positive control, and negative control. No internal standards are added to the template-free blank control. Fragment each nucleic acid, perform end repair, add A, ligate adapters, and perform PCR amplification using dual-tag amplification primers to obtain a sequencing library containing sample tag information. Perform high-throughput sequencing on the sample library to be tested, positive control library, negative control library, and template-free blank control library in the same testing batch to obtain raw sequencing data that can be distinguished according to samples and controls. The sample tag information is used to determine the sample or control source corresponding to the sequencing sequence. S2. Perform quality control on the raw sequencing data, remove sequencing sequences that do not meet the preset quality requirements, and obtain high-quality sequencing sequences; S3. The high-quality sequencing sequences are compared with the host sequence database, and sequencing sequences identified as originating from the host are removed to obtain host-free sequencing sequences. S4. Based on the lower respiratory tract pathogenic microorganism reference database, the host-de-host sequencing sequences are first preliminarily classified using the k-mer classification method. Then, the sequencing sequences that are not classified by the k-mer classification method are aligned with the reference sequence. The results of the preliminary classification and the results of the reference sequence alignment are integrated to obtain candidate pathogenic microorganisms and the set of detection sequences corresponding to the candidate pathogenic microorganisms. S5. Extract sequence-level features for each detected sequence in the detected sequence set, perform hard filtering on each detected sequence based on the sequence-level features, and score the reliability of detected sequences that do not trigger the hard filtering rules; determine the detected sequences that reach the preset reliability score threshold as reliable detected sequences, and determine the detected sequences that trigger the hard filtering rules or do not reach the preset reliability score threshold as unreliable detected sequences. S6. The reliable detection sequences are aggregated according to the candidate pathogens to obtain the number of reliable detection sequences corresponding to each candidate pathogen. The number of reliable detection sequences is standardized to a preset sequencing data volume based on the total number of valid sequences in the sample. The standardized number of reliable detection sequences of candidate pathogens in the lower respiratory tract sample to be tested is compared with the background detection sequence number of the corresponding candidate pathogen in the template-free blank control of the same batch. Based on the preset reporting rules, it is determined whether the candidate pathogen is included in the report review candidate list, and the report review candidate result is generated.
2. The metagenomic sequencing detection method according to claim 1, characterized in that, In step S2, the raw sequencing data is subjected to quality control using the FastP tool. This quality control includes: Remove connector sequences and low-complexity sequences; A sliding window with a window size of 4-6bp is used to remove low-quality regions within the window whose average Phred quality value is lower than Q15. Remove short sequences with a length of less than 30-34 bp after processing; Identify PCR repeat sequences, group sequencing sequences belonging to the same PCR repeat group, retain only one representative sequencing sequence for each PCR repeat group, and retain the PCR repeat identifier of the representative sequencing sequence and the corresponding PCR repeat group information for use in sequence-level feature extraction in step S5 and reliable detection sequence counting in step S6.
3. The metagenomic sequencing detection method according to claim 1, characterized in that, In step S3, the high-quality sequencing sequence is compared with the host sequence database, and the sequencing sequences that are determined to be of host origin are removed, thus obtaining host-free sequencing sequences. The host sequence database is constructed through the following steps: S31. Obtain and integrate human genome sequences, human transcript sequences, and human mitochondrial sequences from multiple sources; S32. The integrated sequence is split into sequence segments of a preset length, and the sequence segments are assembled and clustered to remove redundancy. S33. Identify and remove endogenous viral integration sequences, microbial contamination sequences, and sequence regions highly homologous to the target pathogen from the database to obtain a host sequence database for host sequence removal. In step S33, the sequence regions highly homologous to the target pathogenic microorganism are identified and removed in the following manner: A representative genome of the pathogenic microorganism was selected, and a sequencing sequence with a length of 150 bp and a coverage of 5× was generated based on the representative genome. The simulated sequencing sequences are aligned to the host sequence database to be processed. The host database regions with an alignment similarity of not less than 95% and an alignment length of not less than 100 bp with the simulated sequencing sequences are identified as homologous regions of pathogenic microorganisms, and these homologous regions are removed from the host sequence database.
4. The metagenomic sequencing detection method according to claim 1, characterized in that, The lower respiratory tract pathogen reference database is constructed through the following steps: S41. Obtain the genome sequence of the target pathogen from a public genome database; S42. Based on the multi-dimensional quality characteristics and weighted scoring model of the genome sequence, the genome sequence is evaluated and screened to obtain a set of high-quality candidate genome sequences; S43. Cluster the high-quality candidate genome sequence set and select representative genome sequences from each cluster category to form a non-redundant representative genome set; S44. Verify the species classification accuracy of the representative genome sequences by average nucleic acid consistency analysis, and remove genome sequences with incorrect species classification; S45. The verified genome sequence is subjected to contamination fragment identification and removal to obtain a high-quality genome sequence for constructing the lower respiratory tract pathogen reference database; S46. Construct an index based on the high-quality genome sequence for k-mer classification, reference sequence alignment, and similarity search.
5. The metagenomic sequencing detection method according to claim 1, characterized in that, In step S41, the genome sequence is screened through steps S42 to S45 to obtain a high-quality genome; the multi-dimensional quality characteristics include genome integrity, contamination level, N base ratio, sequencing depth and species annotation quality; In step S46, the Kraken2 index, Bowtie2 index, and BLAST index are constructed respectively. The Kraken2 index is used for rapid k-mer classification and pre-screening of pathogenic microorganisms, the Bowtie2 index is used for accurate sequence alignment and species identification, and the BLAST index is used for similarity search and differentiation of closely related species. In step S4, Kraken2 is used to perform k-mer rapid classification of the hostless sequencing sequences based on the Kraken2 index, and Bowtie2 is used to perform reference sequence alignment of the sequencing sequences that were not classified by Kraken2 based on the Bowtie2 index. The Kraken2 classification results and Bowtie2 alignment results are then integrated.
6. The metagenomic sequencing detection method according to claim 1, characterized in that, The sequence-level features extracted in step S5 include: Effective alignment length (EAL), alignment similarity (IDY), difference between best and second-best alignment score (BSD), target species-specific k-mer ratio (SSKR), number of species with the same best alignment (EBHC), lowest common ancestor classification level (LCA), ratio of closely related species (RSHR), ratio of genomes within the target species (IGHR), origin region type (RegionType), average Phred quality value (APQ), Q30 base ratio (Q30R), N base ratio (NR), low complexity ratio (LCR), PCR repeat identifier (PCRdup) and corresponding PCR repeat information within the same sample, cross-sample completely identical sequence identifier (OdupX), negative control or template-free blank control hit identifier (CtrlHit) within the same test batch, and high-frequency background sequence identifier (HFBG) within the same test batch. A detected sequence is directly classified as an unreliable detected sequence and no reliability score is applied if it meets any of the following conditions: The detected sequence appeared in the negative control or blank control of the same batch of samples. The detected sequences belong to the same batch of high-frequency background sequences. The detected sequence is an optically repeatable sequence between samples; The detected sequence is derived from a reference blacklist region; or The number of equivalent best-hit species of the detected sequence is not less than 3, and its lowest common ancestor classification level cannot support species-level resolution.
7. The metagenomic sequencing detection method according to claim 1, characterized in that, For a detected sequence q that does not trigger the hard filtering rule, the reliability score R(q) is calculated according to the following formula: ; in, Let be the weight of the i-th sequence-level scoring indicator. Assign a score to the i-th sequence-level scoring indicator; When R(q) is not lower than 70 points, the detected sequence q is determined to be a reliable detected sequence; when R(q) is lower than 70 points, the detected sequence q is determined to be an unreliable detected sequence and is not counted in the number of reliable detected sequences. The hard filtering rules, the weights of the sequence-level scoring metrics, the scoring thresholds, and the preset reliability scoring thresholds are determined in the following ways: Using preset library construction conditions, libraries were constructed and sequenced for lower respiratory tract clinical samples, positive controls, negative controls, template-free blank controls, simulated samples, and known contaminated samples to form a training dataset. The preset library construction conditions include at least universal adapter ligation, magnetic bead in situ dual-tag PCR amplification, internal standard addition, and template-free blank control from the same batch. Identify the PCR repeat sequences in the training dataset and group them according to the method described in step S2. For each PCR repeat group, retain only one representative sequencing sequence, and retain the PCRdup of the representative sequencing sequence and the corresponding PCR repeat group information. The detected sequences in the training dataset are labeled as reliable or unreliable detected sequences, and the corresponding sequence-level features are extracted. Through statistical analysis, parameter search, and cross-validation, the weights, scoring thresholds, and preset reliability scoring thresholds of the hard filtering rules and sequence-level scoring indicators are determined.
8. The metagenomic sequencing detection method according to claim 1, characterized in that, In step S6, only the reliable detection sequence is counted in the number of reliable detection sequences of the candidate pathogenic microorganism; When the number of reliable detected sequences of a candidate pathogen reaches the requirements specified in the preset reporting rules, the candidate pathogen is added to the report review candidate list. If the number of reliable detected sequences of a candidate pathogenic microorganism does not meet the requirements specified in the preset reporting rules, the candidate pathogenic microorganism will not be added to the report review candidate list. The number of standardized, reliable detectable sequences of candidate pathogenic microorganisms is calculated using the following formula: Standardized reliable detection sequence count = Reliable detection sequence count of candidate pathogenic microorganisms × 20,000,000 ÷ Total effective sequence count of the sample; The total number of valid sequences in the sample refers to the total number of sequences from the original sequencing data that were used for pathogen comparison and analysis after quality control.
9. The metagenomic sequencing detection device for pathogenic microbial nucleic acids in lower respiratory tract samples according to claim 8, characterized in that, include: The sequencing module is used to acquire nucleic acids extracted from the lower respiratory tract sample to be tested, the positive control, the negative control, and the template-free blank control. Internal standards are added to the nucleic acids of the lower respiratory tract sample to be tested, the positive control, and the negative control; no internal standards are added to the template-free blank control. Each nucleic acid is fragmented, end-repaired, A-added, ligated with adapters, and amplified by PCR using dual-tag primers to obtain a sequencing library containing sample tag information. The sample library to be tested, the positive control library, the negative control library, and the template-free blank control library are subjected to high-throughput sequencing in the same testing batch to obtain raw sequencing data that can be distinguished according to samples and controls. The sample tag information is used to determine the sample or control source corresponding to the sequencing sequence. The quality control module is used to perform quality control on the raw sequencing data, remove sequencing sequences that do not meet the preset quality requirements, and obtain high-quality sequencing sequences. The host-free sequencing sequence acquisition module is used to compare the high-quality sequencing sequence with the host sequence database, remove sequencing sequences identified as originating from the host, and obtain host-free sequencing sequences. The detection sequence set acquisition module is used to first classify the host-de-host sequencing sequences based on the lower respiratory tract pathogen reference database using the k-mer classification method, then perform reference sequence alignment on the sequencing sequences that were not classified by the k-mer classification method, and integrate the results of the preliminary classification and the results of the reference sequence alignment to obtain candidate pathogens and the detection sequence set corresponding to the candidate pathogens. A reliable detection sequence determination module is used to extract sequence-level features for each detected sequence in the detected sequence set, perform hard filtering on each detected sequence based on the sequence-level features, and score the reliability of detected sequences that do not trigger the hard filtering rules; detect sequences that reach a preset reliability score threshold are determined as reliable detected sequences, and detect sequences that trigger the hard filtering rules or do not reach the preset reliability score threshold are determined as unreliable detected sequences. The report generation module is used to aggregate the reliable detection sequences according to the candidate pathogens, obtain the number of reliable detection sequences corresponding to each candidate pathogen, standardize the number of reliable detection sequences to a preset sequencing data volume based on the total number of valid sequences in the sample, compare the standardized number of reliable detection sequences of candidate pathogens in the lower respiratory tract sample to be tested with the background detection sequence number of the corresponding candidate pathogens in the template-free blank control of the same batch, and determine whether the candidate pathogens should be included in the report review candidate list based on preset reporting rules, and generate the report review candidate result.
10. A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method of any one of claims 1-8.
Citation Information
Patent Citations
Respiratory pathogenic microorganism detection database construction method
CN121811982A
