A method and system for identifying microbial single-cell genome species based on a known reference genome without the need for assembly.
By employing flow cytometry tandem classification and two-stage alignment techniques based on known reference genomes, the efficiency and accuracy issues of microbial single-cell genome species identification have been resolved. This enables highly efficient species identification without co-assembly, and is suitable for accurate species identification of large-scale samples as well as identification of low-quality samples.
Patent Information
- Application Number
- CN202611425912.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2026-09-15
- Publication Date
- 2026-11-17
- Estimated Expiration
- 2046-09-15
AI Technical Summary
Existing technologies for identifying microbial single-cell genome species suffer from high computational resource requirements, strict sequencing depth requirements, sensitivity to data purity, and difficulties in co-assembly, resulting in low identification efficiency and insufficient accuracy. In particular, it is difficult to identify species efficiently and accurately in known reference genome samples.
Using a method based on known reference genomes, classification is performed through flow tandem sequencing. Barcodes are used to identify splitting results. Combined with double truncation criteria and two-stage alignment technology, efficient and accurate species identification without co-assembly is achieved. This includes the use of single database loading, candidate species union, and local reference genome indexing. Low-quality reference bins are removed, and overall alignment and single reference alignment are performed to determine species.
Without co-assembly, it improves the efficiency and accuracy of species identification for large-scale samples, reduces database loading overhead and the number of comparisons, supports the identification of low-quality samples, and reduces the impact of chimeric and mixed bins, explicitly reporting contamination and uncertainty.
Smart Images

Figure CN122930195B_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of bioinformatics and relates to the analysis of microbial single-cell sequencing data. Specifically, it relates to a method and system for identifying microbial single-cell genome species based on a known reference genome without the need for assembly. Background Technology
[0002] Microbial communities are widely distributed in the human gut, soil, and water environment. Their species composition and functional differences are closely related to host health, disease occurrence, and ecological processes. Understanding the species composition of microbial communities is a prerequisite for understanding their biological functions. Microbe-seq, a microbial single-cell genome sequencing technology (Wenshan Zheng et al., High-throughput, single-microbe genomics with strain resolution, applied to a human gut microbiome, Science 376, eabm1483, 2022), utilizes droplet microfluidics combined with bioinformatics analysis to obtain the genomic information of tens of thousands of single-celled microorganisms from complex microbial communities without culturing. This significantly improves throughput and resolution compared to traditional technologies, providing a new data foundation for species-level analysis of microbial communities.
[0003] In the data analysis workflow of the aforementioned technologies, single-amplified genomes (SAGs) are typically used as the basic analytical unit. Initial bins are first established for each SAG, then the bins are clustered based on sequence similarity. Sequencing sequences from SAGs within the same group are recovered and merged for de novo assembly. The assembled bins are then refined to remove contaminants, and finally, species identification is performed based on the assembly results. However, this technology, centered on co-assembly, still faces several challenges in practical applications.
[0004] First, co-assembly demands significant computational resources. De novo assembly consumes substantial memory and computation time; the memory requirement for assembling a single group can reach tens to hundreds of GB. Co-assembling and bin refinement of thousands of SAGs group by group often takes days to weeks on ordinary computing nodes and is difficult to scale through simple parallelization, becoming a bottleneck for efficiency in large-scale sample analysis. Second, co-assembly has certain requirements for sequencing depth. After whole-genome amplification, the coverage distribution of SAGs is uneven, and low-coverage regions are unlikely to form effective assembly sequences. When the overall sequencing depth of the sample is insufficient, co-assembly cannot generate bins that meet the requirements for completeness and contamination, and these SAGs cannot be used for species identification. Third, co-assembly is sensitive to data purity. Double-cell droplets, incomplete sorting, reagent contamination, or mixing of closely related species can introduce signals from multiple genomes into the sequencing sequences of SAGs. Co-assembly can easily produce chimeric or mixed bins, and bin refinement is difficult to completely remove contamination, thus affecting the accuracy of species identification. On the other hand, with the continuous accumulation of reference genomes in public databases, many microbial species in samples already have high-quality reference genomes available. In such samples, if species identification of SAGs can be performed directly using known reference genomes, the aforementioned difficulties in the co-assembly process can theoretically be bypassed. However, how to efficiently and accurately implement this approach on large-scale SAG data still faces practical challenges: processing each SAG individually would result in a large amount of repetitive database loading and computational overhead; the selection of candidate species needs to balance sensitivity and specificity; and the acquisition and management of reference genomes also need to balance efficiency and reusability.
[0005] Therefore, how to efficiently and accurately identify microbial single-cell genomes based on known reference genomes without co-assembly is a technical problem that urgently needs to be solved in this field. Summary of the Invention
[0006] To address the aforementioned problems in the existing technology, this invention provides a method and system for species identification of microbial single-cell genomes based on a known reference genome and without the need for assembly. Without the need for de novo co-assembly and bin refinement, it automatically, efficiently, and accurately identifies microbial single-cell SAGs, enabling species-level analysis of large-scale single-cell samples.
[0007] The technical solution adopted in this invention is: a method for identifying microbial single-cell genome species, comprising the following steps:
[0008] S1. The sequencing sequence files of each SAG in the sample are streamed in concatenation and input into a single-run classification process. The classification process loads the classification database once and classifies the sequencing sequences of all SAGs. Then, according to the barcode identifier in the sequencing sequence name, the classification results are split into each SAG to obtain a species-level classification report for each SAG.
[0009] S2. Arrange the candidate species in the species-level classification report of each SAG in descending order of the number of covered sequencing sequences, and retain the candidate species according to the double truncation criterion to obtain the candidate species list of each SAG; the double truncation criterion is: when the proportion of the cumulative number of covered sequencing sequences of the retained candidate species to the total number of covered sequencing sequences exceeds the first threshold, and when the proportion of the number of covered sequencing sequences of the current candidate species to the total number of covered sequencing sequences is lower than the second threshold;
[0010] S3. Take the union of the candidate species lists of all SAGs in the sample to obtain the sample-level candidate species set; for each candidate species in the sample-level candidate species set, query the local reference genome index. If a match is found, reuse the local reference genome; if no match is found, obtain the reference genome of the candidate species online and write back the local reference genome index.
[0011] S4. Using the reference genomes of each candidate species as reference bins, perform quality checks on the completeness and contamination of each reference bin, and re-annotate the taxonomy. Remove reference bins that do not meet the quality requirements to obtain a set of reference bins.
[0012] S5. Perform a two-stage alignment for species identification: In the first stage, the sequences of each reference bin in the reference bin set are merged into a merged reference sequence carrying a bin identifier prefix. The sequencing sequences of each SAG are batch aligned to the merged reference sequence. The number of sequencing sequences aligned to each reference bin for each SAG is counted according to the bin identifier prefix, and the overall alignment rate and dominant bin for each SAG are obtained. The overall alignment rate is the proportion of the number of sequencing sequences aligned to the merged reference sequence for the SAG to the total number of sequencing sequences for that SAG. The dominant bin is the reference bin with the most aligned sequencing sequences for that SAG. In the second stage, the sequencing sequences of each SAG are aligned to the reference genome of its dominant bin to obtain the single reference alignment rate for each SAG. When the overall alignment rate of an SAG is greater than the species identification alignment rate threshold, the SAG is identified as the species corresponding to its dominant bin; otherwise, the SAG is identified as unknown.
[0013] Co-assembly involves reconstructing the genome from sequencing sequences, a process that demands high levels of sequencing depth, data purity, and computational resources. This invention, however, starts from a known reference genome, eliminating the need for genome reconstruction during species identification; instead, it necessitates comparison between the reference genomes of candidate species. Based on this, this invention replaces co-assembly with reference genome alignment, providing a reference-aligned, non-co-assembled microbial single-cell genome species identification method that meets the efficiency and accuracy requirements of large-scale SAG data.
[0014] In the classification stage, unlike existing technologies that run the classification program one SAG at a time, this invention streams and concatenates the sequencing sequences of all SAGs within a sample before inputting them into a single classification run. Classification databases are typically tens of gigabytes in size. Running the database one SAG at a time requires repeated loading, and the loading overhead increases linearly with the number of SAGs. However, after a single loading, this overhead is amortized across all SAGs, reducing the number of loadings from the order of magnitude of the number of SAGs to a constant. To ensure that the merged classification results can still be attributed to their respective SAGs, this invention uses the barcode identifier carried in the sequencing sequence name to split the classification results, thus obtaining a species-level classification report for each SAG with only one classification run. In the candidate species determination stage, this invention employs a double truncation criterion to retain candidate species for each SAG. Single-cell SAGs may contain signals from multiple species due to two-cell separation, incomplete sorting, or reagent contamination. If only the single species with the highest proportion is retained, the true composition of the mixed SAG will be masked; if all species are retained without screening, a large amount of background noise will be introduced. In the dual truncation criterion, the process stops when the cumulative coverage exceeds the first threshold, ensuring that the retained candidates are sufficient to represent the sequencing signal of the SAG; the process stops when the proportion of a single species falls below the second threshold, excluding background noise. These two conditions constrain the candidate set from the perspectives of information integrity and noise removal, respectively, ensuring that each SAG obtains a candidate species list of appropriate size and sufficient coverage. In the reference genome acquisition and alignment process, this invention first merges the candidate species of all SAGs to obtain a sample-level candidate species set, and then organizes the reference genome around this set. The significance of merging is that many SAGs share closely related candidate species, and the number of unique candidate species at the sample level is much smaller than the sum of the number of candidates for each SAG. Therefore, the acquisition, quality control, and alignment index construction of the reference genome only need to be performed once for this smaller set. Specifically, the local reference genome index allows the downloaded reference genome to be reused across samples, avoiding duplicate downloads; quantitative quality control and independent taxonomic re-annotation of the reference genome allow for local verification of the online-acquired reference sequences, identifying and removing poorly quality or poorly annotated reference sequences before alignment. Building upon this foundation, a two-stage alignment process further controls alignment overhead: In the first stage, sequences from all reference bins are concatenated into a merged reference sequence carrying a bin identifier prefix. Each SAG requires only one batch alignment to simultaneously obtain the overall alignment rate and dominant bin, with the number of alignments not increasing with the number of candidate species. In the second stage, precise alignment is performed only on the dominant bin of each SAG, calculating indicators such as single-reference alignment rate, coverage, and alignment purity. Finally, species determination is completed based on whether the overall alignment rate exceeds the species determination alignment rate threshold. SAGs with insufficient alignment signals are conservatively marked as unknown, rather than being forcibly assigned a species.
[0015] In summary, the above steps work synergistically: single-load classification provides efficient species-level signals; the double-truncation criterion transforms these signals into a moderately sized candidate set; candidate union and local indexing enable the acquisition and organization of the reference genome to be completed in a single sample-by-sample manner; and the two-stage alignment completes species determination per SAG with constant alignment overhead. Thus, this invention eliminates the need for co-assembly while balancing the efficiency of large-scale data processing with the accuracy of species identification.
[0016] Further, in step S1, the streaming concatenation includes: writing the paths of the sequencing sequence files of each SAG into a file list, calling the decompression tool file by file in the file list for streaming decompression, and serially outputting the decompressed sequencing sequences to the classification process via a standard input pipe. This process does not generate any merged intermediate sequencing sequence files. File-by-file streaming decompression and standard input pipe transmission avoid the problem of excessively long command-line parameters and also saves disk space occupied by merging intermediate files.
[0017] Further, in step S1, the barcode identifier is extracted from the sequencing sequence name according to the following rules: when the sequencing sequence name contains the separator |, the field after the last | is taken; if the field contains an underscore, the part before the underscore is taken as the barcode identifier; otherwise, the field is used as the barcode identifier. The splitting process performs a single linear scan on the merged classification results to write each sequencing sequence into the corresponding SAG result, while simultaneously accumulating the number of classified sequencing sequences, the number of unclassified sequencing sequences, and the frequency table of classification unit identifiers for each SAG.
[0018] Furthermore, step S1 also includes: calculating the proportion of unclassified sequencing sequences in each SAG, where the proportion of unclassified sequencing sequences is the ratio of the number of unclassified sequencing sequences to the total number of sequencing sequences in that SAG. When the proportion of unclassified sequencing sequences is not greater than 0.7, the SAG is determined to be a valid SAG. SAGs with an excessively high proportion of unclassified sequences usually have poor sequencing quality or contain a large number of sequences from sources outside the database; removing them first can prevent them from interfering with subsequent analysis.
[0019] Further, in step S2, the first threshold is 0.9 and the second threshold is 0.1; the candidate species list is recorded in the form of a list of taxid identifiers separated by semicolons.
[0020] Further, step S2 also includes: calculating the taxonomic purity of each SAG, wherein the taxonomic purity is the proportion of the number of covered sequencing sequences of the candidate species with the highest number of covered sequencing sequences to the total number of covered sequencing sequences of all candidate species in that SAG, and the candidate species with the highest number of covered sequencing sequences is used as the operational taxonomic unit (OTU) of that SAG. Taxonomic purity characterizes the degree of concentration of sequencing sequences in the dominant species and can be used to assess the degree of mixing of SAGs.
[0021] Further, in step S3, based on the mapping table between taxid identifiers (taxid) and nuccore accession numbers in the classification database, the taxid identifiers of candidate species are mapped to nuccore accession numbers. The entries in the local reference genome index include Assembly number, Assembly name, species name, taxid identifier, local reference genome file path, and a flag indicating whether the reference genome was successfully acquired. The online acquisition includes: obtaining the corresponding Assembly metadata from the nuccore accession number via a public query interface; extracting the Assembly number and Assembly name from the Assembly metadata; constructing a hierarchical directory by grouping the Assembly number (after removing the prefix and version number) into sets of three digits; concatenating these groups to generate the download address for the reference genome; and downloading the reference genome. The write-back process uses a temporary file write followed by atomic replacement, and employs file locks to prevent concurrent write conflicts. The resulting local reference genome database acquires the reference genome as needed through a process of querying the local index, online querying and downloading, and writing back the index. Download records are auditable, and the index file is not corrupted during network interruptions or process anomalies.
[0022] Further, in step S4, the quantitative quality inspection includes: calculating the integrity and contamination of each reference bin using a genome quality assessment tool, classifying each reference bin into high-quality, medium-quality, low-quality, and very low-quality levels according to preset thresholds, and removing reference bins of very low quality; the independent taxonomic re-annotation includes: annotating each reference bin with a taxonomic annotation tool independent of the reference genome source annotation, and merging the re-annotation results with the reference bin set; if different reference bins are re-annotated as the same species, only the reference bin with the best assembly quality is retained, and the rest are removed. The species annotations inherent in the reference genomes of public databases may occasionally have biases, and the reference sequences themselves may also be of poor quality. Local independent verification before use ensures that both stages of alignment are based on high-quality references. The main function of re-annotation is to remove redundancy: when the reference genomes of multiple candidate species are determined to be the same species after independent annotation, only the reference bin with the highest integrity and lowest contamination is retained as the representative of that species, avoiding ambiguity in species identification caused by reference sequence redundancy in subsequent alignment stages.
[0023] Further, in the first stage of step S5, the name of each sequence in the merged reference sequence is formed by concatenating the identifier of its reference bin with the original sequence name; the SAGs are batch-aligned in batches, with each batch containing 50-200 SAGs; the overall alignment rate multi_align_rate and the dominant bin dominant_bin are calculated using the following formulas:
[0024] ;
[0025] Where s is SAG, R s A represents the total number of sequencing sequences of SAG s. s b is the number of sequencing sequences that SAG s were aligned to the merged reference sequence. j For the j-th reference bin, A s,j This represents the number of sequencing sequences from SAG s that have aligned to the j-th reference bin. Since the name of each sequence carries the identifier of its reference bin, parsing a single alignment result allows us to count the number of sequencing sequences from each SAG that have aligned to each reference bin by prefix, thus simultaneously determining whether an alignment is successful and which species it most likely originates from in a single alignment.
[0026] Furthermore, in the second stage of step S5, the formula for calculating the single-reference alignment rate (single_align_rate) is as follows:
[0027] ;
[0028] Among them, A' sR represents the number of sequencing sequences of SAG s aligned to the reference genome of its dominant bin. s This represents the total number of sequencing sequences for SAGs.
[0029] Furthermore, the second stage of step S5 also includes: calculating the nX coverage_nX of each SAG on the reference genome of its dominant bin, and calculating the alignment purity of each SAG:
[0030] ;
[0031] ;
[0032] Wherein, nX coverage (coverage_nX) refers to the proportion of base sites on the reference genome that have been detected at least n times out of the total length of the reference genome, C n X L represents the number of sites covered by a sequencing depth of at least n times, where n can be 1, 5, or 10, and L is the reference genome length of the dominant bin. Alignment purity reflects the concentration of SAG sequencing signals in the dominant species: an alignment purity close to 1 indicates that the sequencing sequence is almost entirely from the dominant species, while an alignment purity significantly less than 1 indicates that signals from other genomes are mixed into the sequencing sequence.
[0033] Furthermore, the range of the alignment rate threshold is 0.2~0.5; step S5 also includes: when the alignment purity of the SAG is greater than the purity threshold, the SAG is labeled as a pure SAG, otherwise it is labeled as a contaminated SAG; the purity threshold is 0. Since the reference genome itself can be regarded as a pure sequence, the purity threshold is 0 in the no-assembly mode, that is, no purity penalty is applied, and the purity index is only used as a reference for downstream analysis.
[0034] Furthermore, the method also includes a downstream quality control step after species identification: SAGs are arranged in descending order of alignment purity; the purity threshold is iteratively lowered from an initial purity threshold by a preset step size, so that the proportion of SAGs with alignment purity higher than the purity threshold reaches the target allocation ratio; SAGs with alignment purity lower than the lowered purity threshold are labeled as contaminated SAGs; the initial purity threshold is 0.9, and the preset step size is 0.05; and / or, SAGs with 1x coverage lower than the lower coverage limit are removed. Downstream quality control only adjusts the labeling type of SAGs or filters SAGs entering downstream analysis, without affecting the species identification results themselves.
[0035] Secondly, the present invention also provides a microbial single-cell genome species identification system, comprising:
[0036] The SAG classification module is used to stream and concatenate the sequencing sequence files of each SAG in the sample and input them into a single-run classification process. The classification process loads the classification database and classifies the sequencing sequences of all SAGs in a single run. Then, based on the barcode identifier in the sequencing sequence name, the classification results are split into each SAG to obtain a species-level classification report for each SAG.
[0037] The candidate species determination module is used to sort the candidate species in the species-level classification report of each SAG in descending order by the number of covered sequencing sequences, retain the candidate species according to the double truncation criterion, and obtain the candidate species list of each SAG. The double truncation criterion is: when the proportion of the cumulative number of covered sequencing sequences of the retained candidate species to the total number of covered sequencing sequences exceeds a first threshold, and when the proportion of the number of covered sequencing sequences of the current candidate species to the total number of covered sequencing sequences is lower than a second threshold, the retention is stopped.
[0038] The local reference genome acquisition module is used to obtain the union of the candidate species lists of all SAGs in the sample to obtain a sample-level candidate species set; for each candidate species in the sample-level candidate species set, the local reference genome index is queried. If a match is found, the local reference genome is reused; if no match is found, the reference genome of the candidate species is obtained online and the local reference genome index is written back.
[0039] The reference bin construction and re-annotation module is used to take the reference genome of each candidate species as a reference bin, perform quality checks on the completeness and contamination of each reference bin, perform taxonomic re-annotation, remove reference bins that do not meet the quality requirements, and obtain a set of reference bins.
[0040] A two-stage alignment and species identification module is used to perform two-stage alignment and species identification: In the first stage, the sequences of each reference bin in the reference bin set are merged into a merged reference sequence carrying a bin identifier prefix. The sequencing sequences of each SAG are batch aligned to the merged reference sequence. The number of sequencing sequences aligned to each reference bin for each SAG is counted according to the bin identifier prefix, and the overall alignment rate and dominant bin for each SAG are obtained. The overall alignment rate is the proportion of the number of sequencing sequences aligned to the merged reference sequence for the SAG to the total number of sequencing sequences for that SAG. The dominant bin is the reference bin with the most aligned sequencing sequences for that SAG. In the second stage, the sequencing sequences of each SAG are aligned to the reference genome of its dominant bin to obtain the single reference alignment rate for each SAG. When the overall alignment rate of an SAG is greater than the species identification alignment rate threshold, the SAG is identified as the species corresponding to its dominant bin; otherwise, the SAG is identified as unknown.
[0041] The beneficial effects of this invention are:
[0042] Compared to existing species identification pathways centered on de novo co-assembly, this invention 1) eliminates the need for co-assembly and bin refinement, completing species identification based on alignment with a known reference genome. Even SAGs with insufficient sequencing depth or low data purity can be identified, unaffected by chimeric or mixed bins; 2) by merging sequencing sequences for single-load classification, the number of database loading times is reduced from the SAG level to a constant level. For 2000 SAGs, loading overhead is reduced to approximately 1 / 2000, and the database loading time percentage is reduced from approximately 60% to approximately 1.5%; 3) through candidate species union and two-stage alignment, the number of alignments is reduced from the product of the number of SAGs and candidate species to the SAG level, eliminating the candidate species number factor; 4) by using a local reference genome index, cross-sample reuse of the reference genome is achieved, requiring only initial download, significantly reducing network overhead; 5) the entire streaming process does not generate merged intermediate sequencing sequence files, saving disk space and supporting breakpoint resume and incremental processing. Real-world sample validation shows that this invention can reproduce the vast majority of known species identified by the co-assembly process and additionally detect a large number of known species that were not identified by co-assembly. On low-quality samples, this invention does not force species identification and can explicitly report contamination and uncertainty, avoiding species misclassification caused by co-assembly incorrectly incorporating contaminating SAGs into the bin. Attached Figure Description
[0043] Figure 1 This is a reference flowchart for the method of identifying microbial single-cell genome species with non-parameter assembly in an embodiment of the present invention;
[0044] Figure 2 This is a complete flowchart of the method for identifying microbial single-cell genome species with non-assembly parameters in an embodiment of the present invention;
[0045] Figure 3 This is an overlay diagram of known species with no assembly and co-assembly in ZC samples from Example 1 of this invention;
[0046] Figure 4 This is a per-SAG consistency rate diagram of ZC samples co-assembled and unassembled at each classification level in Embodiment 1 of the present invention;
[0047] Figure 5 This is a distribution diagram of the no-assembly determination results for IMU-A031 sample in Embodiment 2 of the present invention;
[0048] Figure 6 This is a quality control grading chart of the unassembled purity of the IMU-A031 sample in Example 2 of the present invention. Detailed Implementation
[0049] The following specific embodiments illustrate the implementation of the present invention. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and various details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that, unless otherwise specified, the following embodiments and features described therein can be combined with each other.
[0050] Combination Figures 1-2 The overall process shown in this invention, the non-assembled microbial single-cell genome species identification method, starts from the sequencing sequences (FASTQ files) after each SAG preprocessing, and proceeds through single loading and classification of merged sequencing sequences, double truncation annotation of candidate species, acquisition of local reference genomes, construction and re-annotation of reference bins, and two-stage alignment for species identification. Finally, it outputs a SAG species identification result table (including SAG-species mapping, coverage, alignment purity, and annotation type) and a species abundance table. The specific implementation of each step is as follows.
[0051] Step S1: Single loading and classification of merged sequencing sequences
[0052] The FASTQ file paths of all SAGs within the sample are written to a file list. A multi-threaded decompression tool is called file by file in the list for streaming decompression (if a multi-threaded decompression tool is unavailable, it reverts to single-threaded decompression). The decompressed sequencing sequences are serially output to the standard input pipeline. The pipeline terminal is a single-run classification process, reading all sequencing sequences from the standard input and loading the classification database only once to complete the classification of all SAGs. No merged intermediate FASTQ files are generated throughout the process. In this embodiment, the classification database is a custom database built based on the GTDB_v226 genome, with species as the classification granularity. The database contains species name records (the correspondence between taxid and scientific name) and classification hierarchy records (the correspondence between taxid and classification hierarchy), and memory mapping can be enabled to reduce physical memory usage.
[0053] The classification process outputs classification results for each sequence (each line contains the sequence name and classification taxid) and a classification report aggregated by classification level. The sequence name (read_id) carries a barcode identifier, which is extracted according to the following rules: when the sequence name contains the separator |, the field after the last | is taken; if the field contains an underscore, the part before the underscore is taken as the barcode identifier; otherwise, the entire field is used as the barcode identifier.
[0054]
[0055] Based on this, a single linear scan is performed on the merged classification results. The classification result of each sequencing sequence is written into the result file of the corresponding SAG, and the number of classified sequencing sequences, the number of unclassified sequencing sequences, and the taxid frequency table are accumulated for that SAG. For each SAG, a species-level classification report is generated based on its taxid frequency table and combined with the species name record and taxonomic level record in the database, retaining only entries at the species level. The proportion of unclassified sequencing sequences is calculated for each SAG:
[0056] ;
[0057] Where, N unclassified N represents the number of unclassified sequencing sequences of this SAG. total_reads This represents the total number of sequencing sequences for the SAG (each pair of sequences in paired-end data is counted as one sequence). When the unclassified_ratio is not greater than 0.7, the SAG is considered taxonomically valid and proceeds to subsequent analysis.
[0058] Step S2: Double truncation annotation of candidate species
[0059] Let the candidate species in the SAG s species-level taxonomy report, arranged in descending order of the number of sequenced covers, be {(t1,c1), (t2, c2), …, (t m , c m )}, where m is the total number of candidate species, t j Let c be the taxid of the j-th candidate species. j Let c1 be the number of sequencing sequences covered by this candidate species, satisfying c1 ≥ c2 ≥ … ≥ c m Total coverage C total The sum of the number of sequencing sequences covering all candidate species. Candidate species are retained sequentially according to the following double truncation criterion: retention is stopped when the cumulative coverage of the first i retained candidate species exceeds 0.9% of the total coverage; retention is stopped when the current candidate species t... i When the coverage of a candidate species falls below 0.1% of the total coverage, retention is stopped. Retained candidate species are recorded as a semicolon-separated list of taxids, denoted as major_species(s) = t1;t2;…;t k , where k is the number of candidate species retained. Each candidate taxid is mapped to a nuccore accession using a mapping table between taxid and nuccore accession number, for use in obtaining the reference genome. Simultaneously, the taxonomic purity of each SAG is calculated:
[0060] ;
[0061] Where c1 is the number of covered sequencing sequences of the candidate species (dominant species) with the highest number of covered sequencing sequences, Σ j c j The total number of sequencing sequences covering all candidate species was determined, and the taxid of the dominant species was used as the operational taxonomic unit (OTU) for this SAG. A taxonomic purity close to 1 indicates that the sequencing sequences are almost entirely from a single species, while a taxonomic purity significantly less than 1 indicates the presence of mixed signals.
[0062] Step S3: Local reference genome acquisition
[0063] The union of the major species of all SAGs within the sample is taken to obtain the sample-level candidate species set T = {t1,…, t}. K} where K is the number of unique candidate species after the union. Since many SAGs share closely related candidate species, K is much smaller than the sum of the number of candidates for each SAG. For each candidate species in set T, the corresponding nucleic acid sequence accession number is obtained through the mapping table between taxid and nucleic acid sequence accession number. Then, the reference genome is obtained by following the process of looking up the local index, querying and downloading online, and writing back to the index:
[0064] (1) Query the local reference genome index. The local reference genome index is an incremental index, and its entries include Assembly number, Assembly name, species name, taxid, local reference genome file path, and a flag indicating whether the reference genome was successfully obtained. If the accession number of the nucleic acid sequence to be queried matches the index entry and the flag indicates that it has been obtained, the corresponding local reference genome file is directly reused; otherwise, online retrieval is initiated.
[0065] (2) Online Acquisition. The corresponding Assembly identifier is obtained from the NCBI public query interface using the nucleic acid sequence accession number. The metadata summary of the Assembly is then obtained, from which the Assembly number (e.g., GCF_000006745.1, where GCF_ is prefixed with RefSeq and GCA_ with GenBank), Assembly name, species name, and taxid are extracted. The 9-digit Assembly number, after removing the prefix and version number, is grouped into hierarchical directories in sets of 3 (e.g., GCF_000006745.1 corresponds to 000 / 006 / 745). This hierarchy is then concatenated with the Assembly prefix, the Assembly number with the version number, and a fixed genome file naming suffix to generate a download address for the reference genome compressed file, which is then downloaded. In this embodiment, the batch query size is 10, the download retries are 5, and the download timeout is 300 seconds.
[0066]
[0067] Here, `download_link` is the complete FTP download address for the compressed FASTA reference genome (genomic.fna.gz); `prefix` is the Assembly prefix (GCF_=RefSeq, GCA_=GenBank); `AAA / BBB / CCC` are the three-level subdirectory groupings of the Assembly number: the remaining 9 digits after removing the prefix and version number from the `accession` are grouped into sets of 3 digits (e.g., GCF_000006745.1→000006745→000 / 006 / 745), corresponding to the NCBI FTP storage hierarchy; `accession_name` is the Assembly number with version number (e.g., GCF_000006745.1); `genomic.fna.gz` in the filename is a fixed NCBI naming suffix. The first-level root directory `https: / / ftp.ncbi.nlm.nih.gov / genomes / al1 / ` is the NCBI Genome FTP root path.
[0068] (3) Write back the index. Write the newly acquired reference genome information to the local reference genome index. When writing, first write to a temporary file and then replace it atomically. File locks are used to prevent concurrent write conflicts.
[0069] Thus, the local reference genome database consists of three parts: a reference genome storage directory, an incremental index, and a taxid mapping table. The reference genome only needs to be downloaded the first time it is used, and can be reused across samples thereafter.
[0070] Step S4: Refer to bin build and re-annotation
[0071] For each candidate species in the sample-level candidate species set T, the reference genome obtained in step S3 is normalized into a reference bin, forming a reference bin set. The specific species to which each SAG belongs is not assigned during the construction phase, but determined by the two-stage alignment in step S5. Since the reference genome comes from a public database and includes online species annotations, the reference bins are re-annotated as follows to independently verify their quality and classification: the integrity and contamination of each reference bin are calculated using a genome quality assessment tool (such as CheckM), and they are classified into high-quality, medium-quality, low-quality, and very low-quality levels according to preset thresholds; independent taxonomic annotation tools (such as GTDBTK) are used to independently annotate the reference bins, and the re-annotation results are merged with the reference bin list; if different reference bins are re-annotated to the same species, only the reference bin with the best assembly quality is retained as the alignment reference, and the remaining low-quality reference bins are discarded. The main purpose of re-annotation is to remove redundancy: when reference genomes of multiple candidate species are determined to belong to the same species after independent annotation, only the reference bin with the highest integrity and lowest contamination is retained as the representative of that species, avoiding ambiguity in species identification caused by reference sequence redundancy in subsequent alignment stages. This process involves local independent verification of the annotations downloaded online, and reference genomes with annotation bias or poor quality are identified and filtered before alignment.
[0072] Step S5: Two-stage comparison and seed determination
[0073] Phase 1 (Batch Comparison): Let the reference bin set retained after re-annotation be B = {b1, …, b}. B}, where B is the total number of reference bins. The sequences from each reference bin are concatenated into a combined reference sequence `combined_bin`. After concatenation, the name of each sequence retains the identifier prefix of its respective reference bin; that is, the sequence name is formed by concatenating the bin identifier with the original sequence name. All sequencing sequences from each SAG (approximately 100 SAGs per batch, which can be parallelized according to the number of computed cores) are aligned to the alignment index of the combined reference sequence at once. Short-read or long-read aligners can be used, and high-sensitivity local alignment is employed. The alignment results are analyzed, and the sequencing sequences of each SAG are counted by barcode, and the number of sequencing sequences in each reference bin is counted by the bin identifier prefix in the sequence name. For SAG s, the overall alignment rate `multi_align_rate` is calculated, and the dominant bin `dominant_bin` is determined.
[0074] ;
[0075] Among them, R s A represents the total number of sequencing sequences of SAG s. s b is the number of sequencing sequences that SAG s were aligned to the merged reference sequence.j For the j-th reference bin, A s,j The number of sequences aligned to the j-th reference bin for SAG s. `multi_align_rate(s)` is the overall alignment rate (the proportion of successfully aligned sequences to the total number of sequences), and `dominant_bin(s)` is the dominant bin with the most aligned sequences (even if A...). s,j The largest b j The comparison results of the first stage are persistently stored, supporting breakpoint resume and incremental processing.
[0076] Phase 2 (Single SAG Alignment): For each SAG, an index is built using the reference genome of its dominant bin as a reference, and precise alignment is performed. The single-reference alignment rate (single_align_rate) is calculated.
[0077] ;
[0078] Among them, A' s R represents the number of sequencing sequences of SAG s aligned to the reference genome of its dominant bin. s This represents the total number of sequencing sequences for SAGs.
[0079] Calculate the coverage_nX of n X, and calculate the alignment purity of each SAG:
[0080] ;
[0081] ;
[0082] Among them, C n X The number of sites covered by sequencing depth of at least n times, where n is 1, 5, or 10, and L is the reference genome length of the dominant bin. Alignment purity is the proportion of aligned sequences from the dominant bin to all successfully aligned sequences.
[0083] Species identification: When the overall alignment rate (multi_align_rate(s)) of a SAG is greater than the species identification alignment rate threshold (min_align_rate), the SAG is identified as the species corresponding to its dominant bin; otherwise, it is identified as unknown. The species identification alignment rate threshold is specified by the user and ranges from 0.2 to 0.5. When the alignment purity of an SAG is greater than the purity threshold, it is labeled as a pure SAG (Pure_SAG); otherwise, it is labeled as a contaminated SAG (Contaminated_SAG). In the no-assembly mode, the purity threshold is 0, meaning the reference genome is considered pure and no purity penalty is applied.
[0084] Optional downstream quality control: After species identification, SAGs can be further quality controlled based on alignment purity or coverage. This only affects the labeling type or the selection of SAGs for downstream analysis, and does not affect species identification results. Batch purity quality control is performed as follows: SAGs are arranged in descending order of alignment purity. Starting from an initial purity threshold of 0.9, the purity threshold is iteratively lowered in steps of 0.05 until the proportion of SAGs with alignment purity higher than the purity threshold reaches the target allocation ratio. SAGs with alignment purity lower than the lowered purity threshold are labeled as contaminant SAGs. Coverage quality control is performed as follows: SAGs with 1x coverage below the lower coverage limit are directly removed to ensure that the sequencing depth of SAGs entering downstream analysis meets the standard.
[0085] Example 1: Species identification of real sample ZC
[0086] ZC samples (fecal gut microbiome samples, 46,164 valid SAGs) were used to compare the unassembled mode of this invention with the existing co-assembly process. Both modes were loaded once into the GTDB_v226 species-level taxonomy database and classified by flow cytometry. Subsequently, the unassembled mode of this invention performed steps S2 to S5 above to complete species identification. Known species were predominant in this sample. During the comparison, SAGs that were unidentified by both methods and placeholder species were removed, and only named known species were counted.
[0087] Table 1 shows a comparison of the known species identified by the two methods.
[0088] Table 1. Comparison of the set of known species identified by co-assembly and non-assembly in ZC samples
[0089] index Co-assembly No assembly Identify the number of known species 27 300 Reproduction of co-assembled known species — 25 / 27(92.6%) Additional detection of known species — 275
[0090] Figure 3 This is a diagram showing the overlap of known species in unassembled and coassembled ZC samples. (Based on Table 1 and...) Figure 3 As can be seen, in terms of species coverage, the unassembled model reproduced 25 of the 27 known species identified by co-assembly (92.6%), with only 2 low-abundance members not detected. This means that the vast majority of known species identified by co-assembly are covered by the unassembled model. In terms of species expansion, the unassembled model detected an additional 275 known species not identified by co-assembly, bringing the total number of identified species to 11.1 times that of co-assembly. This is because co-assembly merges multiple closely related SAGs into a few bins, losing interspecific differences, while the unassembled model, through direct alignment with the reference genome, can restore a more granular species classification.
[0091] The consistency rate of each SAG at each taxonomic level was statistically analyzed for 32,761 SAGs that were successfully identified using both methods. The results are shown in Table 2.
[0092] Table 2. Consistency rates of SAGs for each classification level between co-assembled and unassembled ZC samples
[0093] Category hierarchy Consistency Phylum 92.6% Outline (class) 87.4% order 73.0% Family 70.9% genus 65.0% Species 45.2%
[0094] Figure 4 The SAG consistency plots for co-assembled and unassembled ZC samples at each classification level are shown in Table 2. Figure 4 It is evident that the two methods are highly consistent at higher taxonomic levels such as phylum and class. The differences in SAGs at the species level mainly stem from species drift caused by contamination or chimerism in the co-assembly bins, while the unassembled pattern still covers 92.6% of known co-assembled species at the species ensemble level. Due to insufficient alignment rate or purity, the unassembled pattern conservatively labels some SAGs as "Unknown," resulting in a successful species identification rate of 62.4%, lower than the 92.6% for co-assembled SAGs. However, in fecal samples dominated by known species, removing the "Unknown" pattern completely reproduces the results at the known species level and significantly expands upon the co-assembly results.
[0095] Example 2: Species identification of low-quality sample IMU-A031
[0096] IMU-A031 is a low-quality, difficult-to-assemble sample (6874 SAGs, many of which are multi-species contamination), directly demonstrating the advantages of the assembly-free process for low-quality data.
[0097] The co-assembly process assigned 83.6% of the SAGs (5749) to 41 known species. However, a large number of SAGs in this sample were actually contaminated by multiple species. Co-assembly would incorrectly include contaminated SAGs in the bin, leading to misclassification. The non-assembly mode of this invention, after two-stage comparison and purity quality control, only completed species identification for 86 SAGs (1.3%) (of which 85 SAGs were assigned to 10 nameable known species), and identified 5263 SAGs (76.6%) as contaminated (containing multiple significant candidate species in the candidate species list) and 1525 SAGs (22.2%) as Unknown. The distribution of the identification results is shown in Table 3.
[0098] Table 3. Distribution of IMU-A031 sample no-assembly determination results
[0099] Determination type SAG number percentage Established species 86 1.3% pollute 5263 76.6% Unknown 1525 22.2%
[0100] Figure 5 This is a distribution chart of the no-assembly determination results for IMU-A031 samples. Figure 6 This is a quality control grading chart for the unassembled purity of sample IMU-A031. (Source: [Insert chart here]) Figure 6As can be seen, the average alignment purity of identified SAGs was 0.91, significantly higher than that of contaminated SAGs (0.63) and Unknown SAGs (0.08), indicating that alignment purity can effectively distinguish between SAGs with sequencing signals concentrated in a single species and SAGs with mixed signals or insufficient alignment signals. For SAGs successfully identified by both methods, the species-level concordance rate reached 91.8%. This example demonstrates that, on low-quality data, the assembly-free mode of this invention does not force species identification but explicitly reports contamination and uncertainty, avoiding species misclassification caused by the incorrect inclusion of contaminated SAGs in the bin during co-assembly, and its species identification results are highly consistent with those of co-assembly.
[0101] In summary, the method of this invention, by eliminating the need for de novo co-assembly and bin refinement, achieves automated species identification of large-scale microbial single-cell samples with constant-level database loading times and comparison overhead that does not expand with the number of candidate species; it can reproduce and significantly extend the species identification results of the co-assembly process on conventional samples, and can explicitly report contamination and uncertainty on low-quality samples, avoiding species misidentification.
[0102] In one specific embodiment, a microbial single-cell genome species identification system is also provided, comprising:
[0103] The SAG classification module is used to stream and concatenate the sequencing sequence files of each SAG in the sample and input them into a single-run classification process. The classification process loads the classification database and classifies the sequencing sequences of all SAGs in a single run. Then, based on the barcode identifier in the sequencing sequence name, the classification results are split into each SAG to obtain a species-level classification report for each SAG.
[0104] The candidate species determination module is used to sort the candidate species in the species-level classification report of each SAG in descending order by the number of covered sequencing sequences, retain the candidate species according to the double truncation criterion, and obtain the candidate species list of each SAG. The double truncation criterion is: when the proportion of the cumulative number of covered sequencing sequences of the retained candidate species to the total number of covered sequencing sequences exceeds a first threshold, and when the proportion of the number of covered sequencing sequences of the current candidate species to the total number of covered sequencing sequences is lower than a second threshold, the retention is stopped.
[0105] The local reference genome acquisition module is used to obtain the union of the candidate species lists of all SAGs in the sample to obtain a sample-level candidate species set; for each candidate species in the sample-level candidate species set, the local reference genome index is queried. If a match is found, the local reference genome is reused; if no match is found, the reference genome of the candidate species is obtained online and the local reference genome index is written back.
[0106] The reference bin construction and re-annotation module is used to take the reference genome of each candidate species as a reference bin, perform quality checks on the completeness and contamination of each reference bin, perform taxonomic re-annotation, remove reference bins that do not meet the quality requirements, and obtain a set of reference bins.
[0107] A two-stage alignment and species identification module is used to perform two-stage alignment and species identification: In the first stage, the sequences of each reference bin in the reference bin set are spliced into a merged reference sequence carrying a bin identifier prefix. The sequencing sequences of each SAG are batch aligned to the merged reference sequence. The number of sequencing sequences aligned to each reference bin for each SAG is counted according to the bin identifier prefix, and the overall alignment rate and dominant bin of each SAG are obtained. The overall alignment rate is the proportion of the number of sequencing sequences aligned to the merged reference sequence of the SAG to the total number of sequencing sequences of the SAG. The dominant bin is the reference bin with the most aligned sequencing sequences for the SAG. In the second stage, the sequencing sequences of each SAG are aligned to the reference genome of its dominant bin to obtain the single reference alignment rate of each SAG. When the overall alignment rate of an SAG is greater than the species identification alignment rate threshold, the SAG is identified as the species corresponding to its dominant bin; otherwise, the SAG is identified as unknown.
[0108] In one specific embodiment, a microbial single-cell genome species identification device is also provided, including a processor and a memory; the processor and the memory are connected via a communication bus; wherein, the processor is used to call and execute a program stored in the memory; the memory is used to store the program, which is at least used to execute the microbial single-cell genome species identification method.
[0109] In one embodiment, a computer-readable storage medium is also provided, on which a computer program is stored, which, when executed by a processor, implements the steps of the microbial single-cell genome species identification method as described above.
[0110] The embodiments described above are merely preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Various modifications and improvements made by those skilled in the art to the technical solutions of the present invention without departing from the spirit of the present invention should fall within the protection scope of the present invention.
Claims
1. A method for identifying microbial single-cell genome species, characterized in that, Includes the following steps: S1. The sequencing sequence files of each SAG in the sample are streamed in concatenation and input into a single-run classification process. The classification process loads the classification database once and classifies the sequencing sequences of all SAGs. Then, according to the barcode identifier in the sequencing sequence name, the classification results are split into each SAG to obtain a species-level classification report for each SAG. S2. Arrange the candidate species in the species-level classification report of each SAG in descending order of the number of covered sequencing sequences, and retain the candidate species according to the double truncation criterion to obtain the candidate species list of each SAG; the double truncation criterion is: when the proportion of the cumulative number of covered sequencing sequences of the retained candidate species to the total number of covered sequencing sequences exceeds the first threshold, and when the proportion of the number of covered sequencing sequences of the current candidate species to the total number of covered sequencing sequences is lower than the second threshold; S3. Take the union of the candidate species lists of all SAGs in the sample to obtain the sample-level candidate species set; for each candidate species in the sample-level candidate species set, query the local reference genome index. If a match is found, reuse the local reference genome; if no match is found, obtain the reference genome of the candidate species online and write back the local reference genome index. S4. Using the reference genomes of each candidate species as reference bins, perform quality checks on the completeness and contamination of each reference bin, and re-annotate the taxonomy. Remove reference bins that do not meet the quality requirements to obtain a set of reference bins. S5. Perform a two-stage alignment for species identification: In the first stage, the sequences of each reference bin in the reference bin set are merged into a merged reference sequence carrying a bin identifier prefix. The sequencing sequences of each SAG are batch aligned to the merged reference sequence. The number of sequencing sequences aligned to each reference bin for each SAG is counted according to the bin identifier prefix, and the overall alignment rate and dominant bin for each SAG are obtained. The overall alignment rate is the proportion of the number of sequencing sequences aligned to the merged reference sequence for the SAG to the total number of sequencing sequences for that SAG. The dominant bin is the reference bin with the most aligned sequencing sequences for that SAG. In the second stage, the sequencing sequences of each SAG are aligned to the reference genome of its dominant bin to obtain the single reference alignment rate for each SAG. When the overall alignment rate of an SAG is greater than the species identification alignment rate threshold, the SAG is identified as the species corresponding to its dominant bin; otherwise, the SAG is identified as unknown.
2. The method according to claim 1, characterized in that, In step S1: the streaming concatenation includes: writing the paths of the sequencing sequence files of each SAG into a file list, decompressing each file in the file list in a streaming manner, and serially outputting the decompressed sequencing sequences to the classification process through a standard input pipe; the barcode identifier is extracted from the sequencing sequence name according to the following rules: when the sequencing sequence name contains the separator |, take the field after the last separator |; if the field contains an underscore, take the part before the underscore as the barcode identifier; otherwise, use the field as the barcode identifier. Step S1 further includes: calculating the proportion of unclassified sequencing sequences in each SAG, wherein the proportion of unclassified sequencing sequences is the ratio of the number of unclassified sequencing sequences to the total number of sequencing sequences in the SAG, and when the proportion of unclassified sequencing sequences is not greater than 0.7, the SAG is determined to be a valid SAG.
3. The method according to claim 1, characterized in that, In step S2: the first threshold is 0.9, the second threshold is 0.1; the candidate species list is recorded in the form of a list of taxonomic unit identifiers separated by semicolons; Step S2 further includes: calculating the classification purity of each SAG, wherein the classification purity is the proportion of the number of covered sequencing sequences of the candidate species with the highest number of covered sequencing sequences to the total number of covered sequencing sequences of all candidate species in the SAG, and the candidate species with the highest number of covered sequencing sequences is used as the operational classification unit of the SAG.
4. The method according to claim 1, characterized in that, Step S3 also includes: Based on the mapping table between taxonomic unit identifiers and nucleic acid sequence accession numbers in the classification database, the taxonomic unit identifiers of candidate species are mapped to nucleic acid sequence accession numbers; The entries in the local reference genome index include Assembly number, Assembly name, species name, taxonomic unit identifier, local reference genome file path, and a flag indicating whether the reference genome was successfully obtained. The online acquisition includes: obtaining the corresponding Assembly metadata from the nucleic acid sequence accession number through a public query interface; extracting the Assembly number and Assembly name from the Assembly metadata; constructing a hierarchical directory by dividing the Assembly number after removing the prefix and version number into groups of 3 digits; splicing the numbers together to generate the download address of the reference genome; and downloading the reference genome.
5. The method according to claim 1, characterized in that, In step S4: The quality inspection includes: calculating the integrity and contamination of each reference bin using a genome quality assessment tool, classifying each reference bin into high quality, medium quality, low quality, and very low quality levels according to a preset threshold, and removing reference bins of very low quality. The taxonomic re-annotation includes: annotating each reference bin with a taxonomic annotation tool that is independent of the reference genome source annotation, and merging the re-annotation results with the reference bin set; if different reference bins are re-annotated as the same species, only the reference bin with the best assembly quality is retained.
6. The method according to claim 1, characterized in that, In the first stage of step S5, the name of each sequence in the merged reference sequence is formed by concatenating the identifier of its reference bin with the original sequence name; each SAG is compared in batches, with each batch containing 50 to 200 SAGs; The formulas for calculating the overall alignment rate (multi_align_rate) and the dominant bin (dominant_bin) are as follows: ; Where s is SAG, R s A represents the total number of sequencing sequences of SAG s. s b is the number of sequencing sequences that SAG s were aligned to the merged reference sequence. j For the j-th reference bin, A s,j The number of sequencing sequences that SAG s aligns to the j-th reference bin.
7. The method according to claim 1, characterized in that, In the second stage of step S5, the formula for calculating the single-reference alignment rate (single_align_rate) is as follows: ; Among them, A' s R represents the number of sequencing sequences of SAG s aligned to the reference genome of its dominant bin. s The total number of sequencing sequences for SAGs; The second stage of step S5 also includes: calculating the nX coverage of each SAG on the reference genome of its dominant bin, and calculating the alignment purity of each SAG. ; ; Among them, C n X The number of sites covered by sequencing depth of at least n times, where n is 1, 5, or 10, and L is the reference genome length of the dominant bin.
8. The method according to claim 7, characterized in that, The threshold value for the fixed-species comparison rate ranges from 0.2 to 0.
5. Step S5 further includes: when the comparison purity of the SAG is greater than the purity threshold, the SAG is labeled as a pure SAG; otherwise, it is labeled as a contaminated SAG; the purity threshold is 0.
9. The method according to claim 7, characterized in that, It also includes downstream quality control steps after seed determination: SAGs are sorted in descending order of comparison purity; the purity threshold is iteratively lowered from the initial purity threshold by a preset step size, so that the proportion of SAGs with comparison purity higher than the purity threshold reaches the target allocation ratio; SAGs with comparison purity lower than the lowered purity threshold are marked as contaminated SAGs. The initial purity threshold is 0.9, and the preset step size is 0.05; and / or, SAGs with 1x coverage below the coverage threshold were removed.
10. A microbial single-cell genome species identification system, characterized in that, include: The SAG classification module is used to stream and concatenate the sequencing sequence files of each SAG in the sample and input them into a single-run classification process. The classification process loads the classification database and classifies the sequencing sequences of all SAGs in a single run. Then, based on the barcode identifier in the sequencing sequence name, the classification results are split into each SAG to obtain a species-level classification report for each SAG. The candidate species determination module is used to sort the candidate species in the species-level classification report of each SAG in descending order by the number of covered sequencing sequences, retain the candidate species according to the double truncation criterion, and obtain the candidate species list of each SAG. The double truncation criterion is: when the cumulative number of covered sequencing sequences of the retained candidate species exceeds the proportion of the total number of covered sequencing sequences, retention is stopped; and when the proportion of the number of covered sequencing sequences of the current candidate species is lower than the proportion of the total number of covered sequencing sequences, retention is stopped. The local reference genome acquisition module is used to obtain the union of the candidate species lists of all SAGs in the sample to obtain a sample-level candidate species set; for each candidate species in the sample-level candidate species set, the local reference genome index is queried. If a match is found, the local reference genome is reused; if no match is found, the reference genome of the candidate species is obtained online and the local reference genome index is written back. The reference bin construction and re-annotation module is used to take the reference genome of each candidate species as a reference bin, perform quality checks on the completeness and contamination of each reference bin, perform taxonomic re-annotation, remove reference bins that do not meet the quality requirements, and obtain a set of reference bins. A two-stage alignment and species identification module is used to perform two-stage alignment and species identification: In the first stage, the sequences of each reference bin in the reference bin set are merged into a merged reference sequence carrying a bin identifier prefix. The sequencing sequences of each SAG are batch aligned to the merged reference sequence. The number of sequencing sequences aligned to each reference bin for each SAG is counted according to the bin identifier prefix, and the overall alignment rate and dominant bin for each SAG are obtained. The overall alignment rate is the proportion of the number of sequencing sequences aligned to the merged reference sequence for the SAG to the total number of sequencing sequences for that SAG. The dominant bin is the reference bin with the most aligned sequencing sequences for that SAG. In the second stage, the sequencing sequences of each SAG are aligned to the reference genome of its dominant bin to obtain the single reference alignment rate for each SAG. When the overall alignment rate of an SAG is greater than the species identification alignment rate threshold, the SAG is identified as the species corresponding to its dominant bin; otherwise, the SAG is identified as unknown.
Citation Information
Patent Citations
Amplifier sequence demultiplexing method and device, electronic equipment and storage medium
CN118800329A
Wading rigid bridge cofferdam stability real-time monitoring method based on multi-source sensing
CN121007599A
