Systems, methods, and apparatus for genetic sequence copy number estimation of recombinant protein production for therapeutic applications

Abstract

A method of determining copy number of a genomic region is disclosed. The method can involve isolating and purifying a genomic nucleic acid sample, followed by creating a shotgun library from the sample. The shotgun library is then sequenced and the generated reads are aligned to a reference sequence consisting of a reference genome and vector sequences. The reference genome is divided into a plurality of sizes and the average read coverage of each size is determined until a predetermined number of consecutive average read coverage peaks are formed with equal coverage distance between them. The respective copy number of each of these consecutive peaks is determined. Furthermore, a model is developed to estimate the copy number of a target sequence. The method utilizes a processor to perform the alignment, division, average read coverage determination, copy number determination, and model configuration.

Description

Cross-reference to related applications

[0001] This application claims the benefit and priority of U.S. Provisional Patent Application No. 63 / 551,348, filed February 8, 2024, the entire contents of which are incorporated herein by reference. Technical Background Related fields This disclosure relates to copy number estimation of genetic sequences. More specifically, this disclosure relates to systems, methods, and apparatus for estimating the copy number of genetic sequences used in the generation of recombinant proteins for therapeutic applications.

[0003] Related technical descriptions The production of recombinant proteins, especially those for therapeutic purposes, relies heavily on Chinese hamster ovary (CHO) cells. These cells are favored for their ability to perform complex post-translational protein modifications. However, long-term culture of CHO cells presents several challenges. One issue is the reduction in recombinant protein expression levels. Furthermore, changes in the primary structure of these proteins can also affect their functionality and safety. Therefore, given their therapeutic applications, characterization of the genetic translation sequence is typically required for the generation of recombinant proteins from transgenic sources.

[0004] Given the therapeutic properties of these proteins, the regulatory submission process for these therapeutic proteins involves a safety assessment of CHO-derived recombinant clones. One aspect of this assessment is molecular characterization at the DNA level. This characterization process includes determining the copy number, integrity, and stability of the transgene, as well as elucidating its integration into the host genome.

[0005] Historically, the molecular characterization of these clones primarily relied on a combination of Southern blotting and Sanger sequencing. More modern methods include quantitative PCR (qPCR) and, more recently, digital droplet PCR (ddPCR). While these methods are robust and have been the primary means of clone characterization, they are not without limitations. They tend to be both time-consuming and resource-intensive, which can be a significant drawback in fields where efficiency and cost-effectiveness are increasingly important.

[0006] The advent of next-generation sequencing (NGS) technology marks a significant advancement in the field of molecular characterization. NGS offers a highly sensitive alternative to traditional methods such as Southern blotting, and is both more cost-effective and labor-saving. However, the application of NGS in this context is not without challenges. Current NGS methods, primarily developed for the characterization of transgenic crops, typically rely on PCR-based library preparation workflows. These methods, along with overly simplistic approaches for copy number estimation, can lead to inaccuracies. Specifically, they can cause deviations in the determination of transgenic copy numbers, necessitating the use of additional ddPCR methods for accurate estimation. Summary of the Invention

[0007] The implementation scheme disclosed in this paper includes a single PCR-free whole-genome NGS library preparation and sequencing workflow, combined with advanced bioinformatics analysis. This novel method compares data with reference sequences, enabling accurate determination of copy number, variant identification, detection of vector structural variations, and annotation of vector integration. This approach represents a significant advancement over existing methods, providing more accurate and efficient characterization of CHO-derived recombinant clones.

[0008] Various known techniques can be utilized, such as NGS methods known as targeted locus amplification (TLA) and long-read sequencing technologies, particularly those based on Oxford Nanopore technology. Some embodiments of this disclosure can build upon and enhance these existing methods, providing a more comprehensive and efficient solution for the molecular characterization of CHO-derived recombinant clones.

[0009] One implementation of a method for determining the copy number of a genomic region involves several steps. The method isolates and purifies a genomic nucleic acid sample. A shotgun library is then created from the genomic nucleic acid sample. The shotgun library is sequenced to generate reads. These reads are aligned to a reference sequence using a processor, the reference sequence comprising a reference genome and a vector sequence. The reference sequence may also include one or more auxiliary sequences, such as an *E. coli* genome sequence. The reference genome is then divided into multiple sizes using the processor.

[0010] The processor determines the average read coverage for each size of the partitioned reference genome. This determination is continued until a predetermined number of consecutive average read coverage peaks are formed, with equal coverage distances between each peak. The processor then determines the corresponding copy number associated with each of these consecutive average read coverage peaks.

[0011] The processor can then build a model configured to estimate the copy number of the target sequence. In summary, the method utilizes multiple steps, including the isolation and purification of genomic nucleic acids, the creation of a shotgun library, sequencing, alignment, partitioning of a reference genome, determination of average read coverage and copy number, and the establishment of a model for copy number estimation.

[0012] A method for determining the copy number of a genomic region involves several steps. The method may include isolating and purifying a genomic nucleic acid sample. A shotgun library is then created from the genomic nucleic acid sample. The shotgun library is sequenced to generate reads. These reads are aligned to a reference sequence using a processor, the reference sequence comprising a reference genome and a vector sequence. The reference genome is then partitioned into multiple sizes using the processor.

[0013] The processor determines the average read coverage for each size of the partitioned reference genome. This determination is continued until a predetermined number of consecutive average read coverage peaks are formed, with equal coverage distances between each peak. The processor then determines the corresponding copy number associated with each of these consecutive average read coverage peaks.

[0014] The processor can then build a model configured to estimate the copy number of the target sequence. In summary, the method utilizes multiple steps, including the isolation and purification of genomic nucleic acids, the creation of a shotgun library, sequencing, alignment, partitioning of a reference genome, determination of average read coverage and copy number, and the establishment of a model for copy number estimation.

[0015] The method can estimate the copy number of a target sequence located within a vector sequence. Specifically, after determining the average read coverage and consecutive average read coverage peaks of the partitioned reference genome, the processor can build a model to estimate the copy number of the target sequence. The target sequence refers to the genetic sequence located within the vector sequence, which is used together with the reference genome to construct the reference sequence. By analyzing the correlation between read coverage and copy number on the reference genome fragment, the model can then accurately calculate the predicted copy number of such target sequences found in the vector based on the coverage observed from aligned sequencing reads.

[0016] The method may involve determining the copy number of a vector sequence as a target sequence. Specifically, after isolating and purifying genomic nucleic acids from a sample, creating a shotgun library from the sample, sequencing the library to generate reads, and aligning the reads to a reference sequence containing a reference genome and a vector sequence, the processor divides the reference genome into multiple sizes. The processor then determines the average read coverage for each size of the divided reference genome until a predetermined number of consecutive average read coverage peaks are formed, with equal coverage distances between each peak. Furthermore, the processor determines the corresponding copy number associated with each of these consecutive peaks. Finally, the processor builds a model configured to estimate the copy number of the vector sequence (in this case, the target sequence).

[0017] The method determines the copy number of a specific predetermined gene located within a vector sequence. After isolating and purifying genomic nucleic acids from a sample, a shotgun library is created and sequenced to generate reads. These reads are then aligned to a reference sequence containing a reference genome and the vector sequence using a processor. The processor divides the reference genome into multiple sizes and determines the average read coverage for each size until a predetermined number of consecutive coverage peaks are observed that are equidistant from each other. The processor can then determine the corresponding copy number associated with each coverage peak. Finally, the processor builds a model to estimate the copy number of the target sequence. If the target sequence is a predetermined gene within the vector, the model can specifically determine the copy number of that gene based on its observed read coverage.

[0018] Furthermore, the model developed by the processor for estimating the copy number of a target sequence can be configured to perform this estimation without requiring genomic location parameters of the target sequence. Specifically, in some embodiments, the model does not require prior knowledge of the target sequence's location in the reference genome to calculate the estimated copy number. Instead, the model can rely solely on observed read coverage of the target sequence determined from aligned sequencing reads. This allows the model to provide unbiased copy number predictions for any target sequence based solely on the measured coverage, without needing to know the precise genomic location of the target sequence, in this particular embodiment.

[0019] The reference sequence used in this method can be the DNA genome of the Chinese hamster ovary (CHO) cell line. CHO cell lines are commonly used to produce recombinant proteins via vector-mediated transgene insertion. Determining the copy number of these inserted transgenes is used to evaluate CHO cell lines considered for therapeutic protein production. Using the CHO genome as a reference sequence allows this method to estimate the copy number of transgenes integrated into the CHO genome via this vector-mediated insertion method. The resulting copy number estimate provides information on transgene expression and stability in candidate CHO cell lines for biotherapeutic applications.

[0020] The processor can analyze a reference sequence containing the genome of a mammalian cancer cell line. Specifically, the reference sequence used in the alignment step can be formed from the DNA genome of a mammalian cancer cell line. By utilizing the genome of a mammalian cancer cell line as a reference sequence, this method can determine the copy number of genomic regions in a cancer sample. Copy number estimation of cancer genomic regions can provide insights into the alterations present in the cancer genome relative to the normal genome. This may contribute to a deeper understanding of cancer biology and the development of new cancer therapies.

[0021] The method may involve creating a shotgun library from a genomic nucleic acid sample. To prepare the library, next-generation sequencing (NGS) or NGS technology is utilized. Specifically, a PCR-free workflow is used to create the shotgun library. By employing polymerase chain reaction (PCR) or a PCR-free method, this helps reduce potential bias and errors that may be introduced by PCR amplification. Therefore, the act of creating a shotgun library in this method utilizes a PCR-free workflow when constructing the library using next-generation sequencing or NGS technology. Using a PCR-free method for library preparation may contribute to generating unbiased representativeness of the genomic sample for subsequent analysis.

[0022] One method for creating shotgun libraries involves using synthetic sequencing. Using this method, the shotgun library is sequenced using synthetic sequencing technology. Specifically, synthetic sequencing is performed using a reversible dye terminator, in which fluorescently labeled nucleotides are added one by one in repeating cycles. Imaging is performed after each nucleotide incorporation to identify the added bases and determine the base sequence in each DNA fragment. This cycle is repeated to generate reads representing short segments of genetic material in the shotgun library. Overall, synthetic sequencing allows for the parallel sequencing of millions of DNA fragments from a shotgun library.

[0023] One method for sequencing shotgun libraries is using Illumina sequencing. Illumina sequencing is a high-throughput next-generation sequencing (NGS) method that can generate a large number of DNA reads in parallel. This sequencing method first involves preparing a shotgun library from a genomic nucleic acid sample, as described above. The shotgun library is then sequenced by Illumina, which uses synthetic sequencing methods to determine the sequence of nucleotide bases in fragmented DNA fragments. This generates millions or billions of short sequencing reads representing segments of genomic material present in the original sample. The reads obtained from Illumina sequencing can then be used for downstream alignment and analysis, as further discussed in the summary of the invention.

[0024] One method used in creating shotgun libraries involves DNABSEQ sequencing, a tagging-based method provided by Illumina. With DNABSEQ sequencing, shotgun libraries undergo an enzymatic fragmentation and tag ligation process called tagging. This allows for simultaneous fragmentation and tagging of DNA fragments before cloning amplification and sequencing. Creating shotgun libraries using DNABSEQ sequencing provides an alternative to other sequencing methods that may involve PCR amplification steps (potentially introducing bias and errors).

[0025] The reference genome is divided into multiple sizes ranging from 100 base pairs (bp) to 51,200 bp. Specifically, the processor divides the reference genome into fragment sizes from 100 bp to 51,200 bp. This fragment size range allows for the analysis of read coverage at different lengths of the reference genome. By dividing the reference genome into fragments of different lengths within this size range, the processor can determine trends and relationships between observed read coverage and target sequence genome copy numbers. This size range used to divide the reference genome helps in robust copy number estimation.

[0026] When dividing the reference genome into multiple sizes, the processor analyzes fragment sizes ranging from 50 base pairs (bp) to 51,200 bp, including partition sizes of 50 bp, 100 bp, 200 bp, 400 bp, 800 bp, 1,600 bp, 3,200 bp, 6,400 bp, 12,800 bp, 25,600 bp, and 51,200 bp. By determining the average read coverage for each of these partition sizes, the processor can observe patterns in coverage values ​​across different fragment sizes. This adaptive partitioning allows the processor to adjust fragment sizes to optimize peak resolution when analyzing read coverage distribution.

[0027] The method divides the reference genome into multiple sizes ranging from 50 base pairs (bp) to 51,200 bp. These sizes include 50 bp, 100 bp, 200 bp, 400 bp, 800 bp, 1,600 bp, 3,200 bp, 6,400 bp, 12,800 bp, 25,600 bp, and 51,200 bp. Specifically, the processor uses the partition sizes of 800 bp, 1,600 bp, 3,200 bp, 6,400 bp, 12,800 bp, 25,600 bp, and 51,200 bp to estimate the copy number of the target sequence. The processor determines the average read coverage for each of these partition sizes and analyzes the coverage peaks and inter-peak distances. Based on this analysis, the processor is then able to estimate the copy number associated with the target sequence.

[0028] To estimate the standard deviation of the determined copy number, the processor can utilize partition sizes ranging from 50 base pairs (bp) to 12,800 bp. Specifically, the processor analyzes the average read coverage determined for partition sizes of 50 bp, 100 bp, 200 bp, 400 bp, 800 bp, 1,600 bp, 3,200 bp, 6,400 bp, and 12,800 bp. By utilizing the coverage data obtained at these different partition sizes, the processor can assess the variability and noise present in the coverage measurement. This variability analysis allows the processor to estimate the standard deviation of the copy number, which provides a measure of the reliability and accuracy of the determined copy number.

[0029] When partitioning the reference genome, the processor ensures that all partitions of the partitioned reference genome are non-overlapping. Specifically, in one embodiment, each segment of the analyzed reference genome is included in only a single partition and does not overlap with any other partition. This non-overlapping approach prevents duplicate counting of the same region of the reference genome when determining the average read coverage for each partition size. Analyzing only non-overlapping segments allows for accurate assessment of coverage on the reference genome for each test partition size. Non-overlapping partitions provide reliable coverage data, which the processor can then use to determine the average read coverage for each size until a predetermined number of consecutive average read coverage peaks are identified.

[0030] The method may involve dividing a reference genome into multiple sizes, with at least two divisions overlapping each other. By allowing a degree of overlap between the divisions of the reference genome, read coverage on the reference genome can be analyzed more comprehensively. For example, coverage information of the overlapping region between a first division of 800 base pairs and a second division of 1,600 base pairs can be used to provide additional data points for determining the relationship between observed read coverage and copy number on the reference genome. Therefore, overlapping divisions can help estimate genome copy number more accurately from aligned sequence read data.

[0031] When determining the copy number of genomic regions, the processor divides the reference genome into multiple sizes. In a specific implementation, all partitions of the reference genome can be sequential. That is, when partitioning the reference genome, the processor in this implementation may not create random or non-sequential partitions. Instead, it can partition the reference genome into contiguous segments or fragments of varying lengths. This allows the processor to systematically analyze read coverage across sequential partition sizes of the reference genome. Sequential partitioning enables the processor to methodically determine the average read coverage for each size and identify consecutive peaks of average read coverage indicating different copy numbers. The sequential nature of the partitions helps to accurately identify the relationship between coverage and copy number.

[0032] The method determines the number of consecutive average read coverage peaks to be analyzed. The processor is configured to observe a predetermined number of consecutive average read coverage peaks, each with an equal coverage distance between it. In one exemplary embodiment of the method, the predetermined number of consecutive average read coverage peaks to be analyzed is three peaks. By setting the number of peaks to three, the processor determines the average read coverage for each size of a partitioned reference genome until it observes three consecutive peaks with an equal coverage distance between them. This allows the method to accurately estimate the copy number of the target sequence using patterns in sequencing data represented by three different coverage peaks.

[0033] To determine the copy number associated with each peak, the processor compares the observed equal coverage distance between peaks with the value of a first peak. This first peak is one of a predetermined number of consecutive average read coverage peaks identified in the method. By comparing the equal coverage distance with the value of the first peak, the processor can establish a first corresponding copy number corresponding to the first peak of the consecutive average read coverage peaks. Specifically, the processor analyzes the relationship between the equal coverage distance and the value of the first peak to infer the appropriate copy number to be assigned to that first peak. This process enables the processor to systematically determine the corresponding copy number for each peak based on the relative values ​​of the peaks and the observed equal coverage distances between peaks.

[0034] The processor analyzes the equal coverage distances between consecutive average read coverage peaks. For the first peak of a predetermined number of consecutive average read coverage peaks, if the value of the first peak is approximately equal to the equal coverage distance, the processor determines the first corresponding copy number for that first peak to be one. By comparing the value of the first peak with the determined equal coverage distance between peaks, the processor can infer that if the value of the first peak is approximately equal to that distance, it represents a copy number of one. This process allows the processor to systematically assign a corresponding copy number to each peak based on its value relative to the observed equal intervals between peaks in the coverage data.

[0035] When determining the corresponding copy number associated with each of the consecutive average read coverage peaks, if two peaks exist and they have an equal coverage distance, the processor can determine that the first peak corresponds to copy number one. Furthermore, the processor can determine that the second peak of the two consecutive peaks corresponds to a second corresponding copy number two, whereby the coverage value of the second peak will be approximately twice the equal coverage distance between the two peaks. By comparing the peak coverage values ​​with the equal coverage distance, the processor can infer the corresponding copy number for each peak in the sequence. This process allows the processor to analyze read coverage data and assign copy numbers to represent the copy status reflected by each different coverage peak.

[0036] One step in the method may involve determining the corresponding copy number associated with consecutive average read coverage peaks. The resulting average read coverage peaks have equal coverage distances between them. To determine the corresponding copy number, the processor compares the value of the first peak to the equal coverage distance. If the value of the first peak is approximately twice the equal coverage distance, the processor determines the first corresponding copy number to be two. This indicates that when the coverage value of the first peak is approximately twice the inter-peak distance, the copy number allocated to that first peak will be two.

[0037] The method determines the copy number associated with consecutive average read coverage peaks having equal coverage distances. If the copy number of the first peak is determined to be two based on its coverage value being approximately twice the equal coverage distance, then a second peak of a predetermined number of consecutive average read coverage peaks will correspond to a second corresponding copy number three. This is because if the coverage value of the first peak represents a copy number of two, and there is an equal coverage distance between the peaks, then the coverage value of the second peak will be approximately three times the equal coverage distance. Therefore, the second peak is assigned a copy number of three.

[0038] The method can also utilize a regression model to correlate coverage and copy number. Specifically, the processor establishes a regression model mathematically representing the relationship between observed read coverage and genome copy number. By determining the coverage on a partitioned reference genome and the copy number assigned to consecutive average coverage peaks, the processor can generate a regression model. This regression model uses statistical regression techniques to correlate coverage to copy number. The resulting regression equation summarizes how coverage and copy number change together based on the analyzed data. This regression model is then used by the processor as the established model to estimate the copy number of target sequences determined from alignment reads.

[0039] The method may involve using the processor to partition a reference genome into multiple sizes. Each partitioned reference genome may correspond to at least one gene within the genome. By partitioning the reference genome and analyzing read coverage at different partition sizes, the processor can identify trends and patterns related to the correlation between read coverage and copy number of genomic regions containing one or more genes of interest. This gene-level analysis allows for a close examination of the coverage profiles of specific genes, facilitating more targeted copy number estimations of genes present in the partitioned reference genome segments.

[0040] Furthermore, the method may further include normalizing read coverage using a reference coverage value obtained from a control sample. To perform this normalization, the processor first obtains read coverage data from a separate control sample. This control sample coverage represents a baseline reference value, relative to which the sample's coverage can be normalized. The processor then divides the coverage of each read from the sample by the reference coverage value derived from the control. Normalizing coverage in this way takes into account any technical variations between samples, such as differences in the total number of sequencing reads, and allows for more accurate comparisons of coverage levels between genomic regions. This normalization step enhances the method's ability to reliably estimate copy numbers by controlling for technical biases in the sequencing data.

[0041] The method may further include estimating the standard deviation of the estimated copy number based on read coverage and the size of the target sequence. To estimate the standard deviation, the processor analyzes read coverage at different sizes of a partitioned reference genome. Using varying read coverage from different partition sizes and the estimated copy number, the processor calculates the standard deviation for each estimated copy number. This standard deviation takes into account read coverage variability based on the size of the target sequence (e.g., a gene) used to estimate the copy number. The estimated standard deviation, based on the variability observed in read coverage associated with sequences of different lengths, provides a measure of reliability for each copy number prediction.

[0042] In addition to determining the copy number, the method may further include determining a confidence interval for the estimated copy number. This determination is based on the standard deviation of the copy number estimation and the estimated copy number itself. Specifically, the processor first estimates the standard deviation of the copy number based on read coverage and the size of the target sequence, as described in the previous steps. This standard deviation reflects the variability of the copy number estimation. The processor then uses this standard deviation and the actual estimated copy number to calculate a confidence interval for each copy number prediction. The confidence interval provides a measure of the reliability or accuracy of each estimated copy number.

[0043] Alignment of sequencing reads with reference sequences is performed using an accelerated version of the BWA alignment program. Specifically, the processor aligns the generated reads to a reference sequence, which comprises a reference genome tandem with the vector sequence. This alignment is performed using an accelerated version of the BWA alignment program algorithm. BWA stands for Burrows-Wheeler alignment, typically used to map low-divergence sequences to large reference genomes. Using an accelerated version of this algorithm allows for optimized and efficient read alignment to reference sequences. Employing an accelerated alignment program helps to speed up the read alignment process, enabling the alignment of hundreds of millions of reads within a reasonable timeframe for downstream analysis.

[0044] When determining the average read coverage for each size of a partitioned reference genome, the processor can calculate a median coverage value for each size. Specifically, in some embodiments, the processor does not simply calculate the average coverage of all reads mapped to a given size partition. Instead, it calculates the median coverage value for each partition size to minimize the impact of outliers in the coverage data. By using the median coverage instead of a simple average, the influence of very high or very low coverage reads can be reduced. This can lead to a more accurate characterization of coverage central tendency across each genome partition size. Using the median coverage value for each partition size allows the processor to robustly determine the relationship between coverage and copy number, thereby facilitating the downstream generation of effective genome copy number estimation models.

[0045] The method further includes training the model using the average base level coverage of each of the partitioned reference genome sequences and the known copy number within the reference sequence. To train the model, the processor first determines the average base level coverage of different partitions with known copy numbers. These partitions with known copy numbers allow the processor to establish a relationship between measured coverage and actual copy numbers. The processor then uses the coverage measurements and the corresponding known copy numbers to develop and optimize the model. The trained model captures the correlation between the observed coverage level and copy number within the reference sequence itself. This process enables the model to more accurately estimate the copy number of the target sequence based on its coverage level.

[0046] The method may further include performing quality control analysis on read coverage to identify potential sequencing errors and filter out false positive copy number determinations. By analyzing read coverage, the processor can identify errors that may occur during sequencing. This allows the processor to remove read coverage affected by sequencing errors from downstream analyses. Furthermore, the processor filters out any copy number determinations that are incorrectly identified as true positives due to sequencing errors or other technical artifacts. In some specific embodiments, performing this quality control analysis can improve the accuracy of copy number determination by excluding low-quality or erroneous read coverage data and copy number determinations that do not truly represent the genomic sample.

[0047] The method further includes visualizing the estimated copy number and its associated confidence intervals on a Circos plot. A Circos plot is a circular graphical representation used to analyze and depict various genomic features derived from sequence data. Specifically, the Circos plot generated by this method displays a trajectory showing the estimated copy number of the target genomic region, read coverage information, and the confidence intervals corresponding to the copy number predictions. This visual depiction helps in interpreting and examining the results of copy number analysis. It provides insights into the estimated abundance and variability of target sequences within a sample in an intuitive form.

[0048] The method can determine the copy number of one or more specific genes of interest within a vector sequence. By analyzing reads generated from sequencing and subsequently aligned to a reference sequence (including both the reference genome and the vector sequence), the processor can estimate the coverage of the genes of interest within the vector. Using this observed coverage of the genes of interest and a model that correlates coverage with copy number, the processor can then predict the precise copy number of one or more genes of interest within the vector sequence. This provides valuable information about the abundance and representativeness of the genes of interest contained in the genomic material of a sample.

[0049] The method may further include determining the copy number of one or more genes of interest (GOIs) located within the vector sequence. To estimate the copy number of these GOIs, the processor first obtains the observed read coverage for each GOI from the aligned sequencing reads. The processor then applies a model that correlates read coverage with copy number. By inputting the measured GOI read coverage into this model, which incorporates the relationship between coverage and copy number, the processor can calculate the estimated copy number of that particular GOI based on its observed read coverage data and model parameters. This allows the method to accurately determine the copy number of a specific gene within the vector, providing useful information about the abundance and representativeness of target GOI sequences integrated into the genome.

[0050] The method analyzes samples of genetically modified cells containing vector sequences integrated into the genomic DNA. This vector sequence can be a full-size vector or a partial-size vector, depending on the specific genetic modification applied to the cell. A full-size vector may represent the complete, unchanged vector sequence used to modify the cell. A partial-size vector may indicate rearrangements or deletions in the integrated vector sequence compared to the original vector. By determining whether a full-size or partial-size vector has been integrated into the genomic DNA, the method provides insights into the structure and characterization of the transgenic modifications present in the analyzed sample. This information supports comprehensive molecular characterization of the cells for purposes such as selecting desired biotherapy cell lines or evaluating genetic engineering approaches.

[0051] The method may further include identifying and reporting any vector-vector tandem events based on split reads mapped to different parts of the vector sequence. To detect vector tandem events, sequencing reads aligned to the vector reference sequence are extracted from a global alignment file. These extracted reads are then re-aligned to the vector reference sequence using commonly used read alignment software. This re-alignment process aims to identify vector chimeric reads, i.e., reads aligned to different non-homologous parts of the vector sequence. A tandem location is considered valid if supported by multiple different chimeric reads. By analyzing these split reads mapped to different parts of the vector sequence, the method can detect whether any vector-vector tandem events have occurred. This provides a comprehensive characterization of vector structure within the genome sample.

[0052] The method may also involve detecting and reporting any sequence variants in the integration vector. To this end, the processor compares the generated sequencing reads with known vector sequences. By analyzing differences or mismatches between the reads and vector sequences, the processor can identify any single nucleotide mutations, small insertions, or deletions that may be present in the integration vector. These potential sequence variants are then evaluated, as true variants likely represent real mutations that have occurred in the integration vector over time. However, some observed variants may simply be sequencing errors introduced during library preparation or sequencing. The processor applies various screening steps and filtering criteria to identify high-confidence sequence variants and eliminate false positives. Any validated sequence variants identified in this manner are then output and reported by the processor. This allows for the characterization of any changes that may have occurred in the integration vector sequence compared to the original reference sequence.

[0053] The method may further include detecting vector integration sites in a reference genome. This is accomplished by analyzing chimeric reads from aligned reads. Chimeric reads are reads that partially map to the vector sequence and partially to non-vector regions. Additionally, paired reads are examined, where one paired read maps to the vector sequence and the other to a non-vector region. These chimeric and paired reads are then de novo assembled to generate longer contigs. These contigs are aligned to a reference genome, which includes both the host genome and the vector sequence. This alignment process identifies the boundary between the donor vector DNA and the host genome, thereby determining the precise insertion location of the vector within the host genome.

[0054] The method described herein can be used for the molecular characterization of Chinese hamster ovary (CHO) cells, which are commonly used for the production of recombinant proteins. CHO cells are generally the preferred cell type for recombinant protein production because they are capable of performing complex post-translational modifications of proteins. However, long-term culture of CHO cells can present challenges, such as the decline in recombinant protein expression levels over time or changes in protein structure that affect function and safety. This method allows for detailed molecular characterization of CHO cells at the genomic level, facilitating the assessment of copy number stability, transgenic sequence integrity, and any alterations within the host CHO genome. This comprehensive molecular analysis supports the selection and validation of CHO cell lines for the safe and effective production of recombinant therapeutic proteins.

[0055] This method for determining the copy number of genomic regions can be used to characterize a reference genome at the molecular level. Specifically, after isolating and purifying genomic nucleic acids from a sample, a shotgun library is created and sequenced to generate reads. These reads are then aligned to a reference sequence, which includes both the reference genome and a vector sequence, using a processor. The processor divides the reference genome into multiple sizes and determines the average read coverage for each size. Once consecutive peaks of average read coverage with equal distances between them are identified, the processor determines the corresponding copy number associated with each peak. Finally, the processor builds a model to estimate the copy number of the target sequence. This process, by analyzing read coverage and determining copy numbers across the genome, allows for the molecular characterization of the reference genome.

[0056] The method described herein can be used for the molecular characterization of mammalian cancer cell lines. Specifically, it allows for the characterization of the copy number and genomic content of integrated vectors or transgenes within cancer cells. By applying this method to cancer cell samples, researchers can gain valuable insights into genetic alterations in cancer. For example, this method can be used to analyze transgene integration patterns and copy number variations in human cancer cell lines used in studies. This molecular characterization of transgenic cancer cell lines can provide important information for further scientific understanding of cancer genetics.

[0057] The method further includes calculating the coverage variability for each size of the partitioned reference genome. This is done to estimate a reliability parameter for the estimated copy number. Specifically, after partitioning the reference genome into different sizes and determining the average read coverage for each size, the processor calculates the variability of read coverage across the different sizes. This coverage variability provides an indication of the reliability of the estimated copy number, with lower variability corresponding to a more reliable estimate. By taking the variability factor into account, the processor can account for noise and uncertainty in coverage measurements when determining the confidence and accuracy of the copy number prediction.

[0058] In determining the average read coverage for each size of a partitioned reference genome, the processor calculates the average base level coverage for each partition by taking into account the size and copy number of each fragment. Specifically, when calculating the average base level coverage over a given partition, the processor considers both the length of each partitioned reference genome fragment and the expected copy number based on previous analysis. This allows the processor to more accurately determine the relationship between coverage and copy number by normalizing for differences in fragment size and predicted copy status when calculating the average coverage value. Taking into account size and estimated copy number provides a more refined assessment of coverage trends and their correlation with copy number.

[0059] The method may also involve identifying coverage anomalies within genomic regions. To identify any coverage anomalies, the processor analyzes the average read coverage across partition sizes of the reference genome. By examining the coverage values ​​for each partition size, the processor can identify any irregular or anomalous coverage readings that deviate from the expected pattern. Such anomalous coverage results may be caused by local sequencing errors or read error mapping due to duplications or structural variations in the genome. Identifying any coverage anomalies allows the processor to filter out erroneous coverage data from the determination of average read coverage and copy number estimation. In some specific implementations, only coverage patterns considered regular are used to establish the relationship between coverage and copy number to generate an accurate copy number estimation model.

[0060] The method may further include identifying homologous sequence regions within the vector sequence. To this end, after sequencing the shotgun library and aligning the resulting reads to a reference sequence, the processor analyzes the vector sequence portion of the reference sequence. The processor can search for identical or very similar sequences within the vector. These homologous sequences represent regions within the vector that are highly similar in their nucleotide composition. By locating homologous regions within the vector, the method provides additional characterization of the vector's internal structure and composition, beyond copy number determination. This additional step enhances the molecular characterization of the integrated vector sequence.

[0061] The method may further include comparing the estimated copy number of the target sequence with a predetermined threshold to estimate the copy number stability and integrity of the target sequence. By comparing the estimated copy number of the target sequence (e.g., a gene of interest) with a predetermined threshold, insights into copy number stability and integrity can be obtained. If the estimated copy number is within the threshold range, it indicates that the copy number is stable and intact. However, if the estimated copy number falls outside the threshold range, it may indicate copy number instability or impaired integrity. Therefore, comparing the estimated copy number with a predetermined threshold provides a means of assessing the copy number stability and consistency of the target sequence.

[0062] The method further includes validating the accuracy of the copy number determination using molecular characterization techniques. This molecular characterization can be derived from one of several established methods, including Southern blotting analysis, digital droplet PCR (ddPCR), or quantitative PCR (qPCR). Southern blotting analysis is a conventional technique that uses gel electrophoresis and DNA hybridization to visualize DNA fragments and validate the estimated copy number. Furthermore, ddPCR and qPCR are highly sensitive methods capable of accurate validation by directly measuring DNA copy number. The method validates the accuracy of the copy number determination by comparing the estimated copy number from partitioning and model-based methods with the results of Southern blotting, ddPCR, or qPCR analysis. This validation step helps confirm the reliability of the copy number estimates calculated through this novel genome partitioning and modeling workflow.

[0063] To improve the accuracy of variant detection and integration site annotation, the method utilizes long-read sequencing technology to generate reads. Compared to short-read techniques, long-read sequencing produces significantly longer reads. These longer reads provide additional sequence information, enhancing the accuracy of identifying variants present in the vector sequence when compared to a reference sequence. They also allow for more precise determination of vector integration sites by facilitating the assembly of split reads and paired reads across the vector-host junction. Therefore, employing long-read sequencing for read generation within this method can be beneficial for downstream analyses of variant detection and vector integration site annotation.

[0064] Additional analysis can be performed on sequencing reads to detect any rearrangements or structural variations that may be present within the reference sequence. The method analyzes the alignment patterns of sequencing reads aligned to a reference sequence (including a reference genome and vector sequence). By examining the alignment patterns of the reads, the processor can identify any potential rearrangements or structural variations that differ from the reference sequence. This may involve analyzing various aspects of the read alignment, such as alignment position, orientation, and pairing information. Detecting any rearrangements or structural variations provides further characterization of the sample's genomic sequence beyond copy number determination. This additional analysis enhances the comprehensive molecular profiling of the sample achieved through this method.

[0065] Furthermore, the method may involve generating a molecular characterization report. This report summarizes the analysis results. It contains information such as the estimated copy number of the target sequence determined by the model. The report also includes details about any variants detected in the sequence. Specifically, it outlines the variant profile, providing data on identified single nucleotide variants and small insertions or deletions. Additionally, the report specifies the integration sites of the target sequence into the reference genome. Locating these integration sites is achieved by analyzing chimeric and paired reads mapped to vector and non-vector regions. In summary, the molecular characterization report summarizes findings regarding the estimated abundance, variants, and genomic location of the target sequence within the reference genome.

[0066] The method may further include determining the ploidy distribution of a reference genome to optimize the estimated copy number of the target sequence. To determine the ploidy distribution, the processor analyzes the observed coverage of fragments of different sizes from the reference genome. By examining coverage at different fragment sizes, the processor can identify data patterns associated with the ploidy or copy number state of the reference genome region. The position and width of peaks in the ploidy distribution provide insights into the true copy number spectrum. For example, the dominant peak typically corresponds to the major copy number. Considering the ploidy distribution can help improve the accuracy of copy number estimations for the target sequence.

[0067] The accuracy of the estimated copy number determined by the method is further validated using at least one of FISH analysis, ddPCR analysis, or qPCR. Experimental validation techniques, such as fluorescence in situ hybridization (FISH), digital droplet PCR (ddPCR), or quantitative PCR (qPCR), can be applied to biological samples to obtain independent copy number measurements for comparison with the estimates calculated using this method. Comparing the estimated copy number obtained by this method with values ​​obtained from well-established validated methods helps confirm the reliability and accuracy of the copy number determination achieved through the computational process.

[0068] To further optimize copy number estimation, the processor can acquire coverage data for different fragments of varying sizes from the reference genome. These fragment sizes include 50 base pairs (bp), 100 bp, 200 bp, 400 bp, 800 bp, 1,600 bp, 3,200 bp, 6,400 bp, and 12,800 bp. The processor extracts coverage information for each of these fragment sizes from aligned reads. Analyzing coverage at different fragment sizes allows the processor to better understand the relationship between coverage and copy number. This relationship can be used to develop accurate copy number estimation models. Smaller fragment sizes help assess coverage variability and noise, while larger fragments are more useful for identifying copy number differences. Therefore, acquiring coverage data for fragments ranging from 50 bp to 12,800 bp helps generate accurate models for estimating genome copy number.

[0069] Furthermore, the method may involve normalizing read coverage. The processor determines the largest peak among consecutive average read coverage peaks identified during the determination of average read coverage for each size of the partitioned reference genome. The processor then normalizes the coverage at each point by dividing the coverage at each point by the coverage of the largest peak. This normalization process provides a normalized coverage value that can be used for downstream analysis by the processor. Normalizing coverage based on the largest peak helps to explain possible differences in the total number of reads between samples or runs, thus allowing for more efficient comparison of coverage data from multiple experiments.

[0070] The method may further include normalizing the occurrence values ​​determined by the analysis. Specifically, in some embodiments, the processor can normalize the occurrence count by the total occurrence count. This is done by dividing the occurrence count for each data point by the sum of the occurrence counts for all data points. This calculation provides normalized occurrence values, allowing comparisons of the frequency of different coverage measures. Normalizing in this way helps to account for possible differences in total occurrence counts between samples or experiments. It corrects the values ​​to make them directly comparable, which can improve the accuracy and reliability of subsequent analytical steps, such as determining copy number or detecting anomalies. In summary, normalization of occurrence values ​​enables more robust statistical analysis of sequencing data.

[0071] To further optimize copy number estimation, the processor can extract the average read coverage for each size of the partitioned reference genome with known copy numbers. The processor then calculates the standard deviation for each size of the partitioned reference genome. Calculating the standard deviation assesses the variation in coverage between fragments of different sizes. Using the standard deviation, the processor can better understand the reliability of the estimated copy number based on the observed variation in coverage between fragments of different sizes. Assessing the variation in fragment coverage using the standard deviation helps optimize the estimated copy number determined by this method.

[0072] The method may also involve identifying tandem events in vector sequences. To this end, the processor analyzes aligned reads to identify potential vector tandem linking reads. A vector tandem linking read is a read containing two segments aligned to different parts of the vector sequence, linking vector sequence segments together within the read. By identifying these linking reads, the processor can determine whether any tandem events have occurred between vector sequences integrated into the host genome. Detecting vector tandem events provides further insight into the structural characteristics and organization of vector sequences present in a sample.

[0073] The method may further include identifying connective reads to detect tandem events in the vector sequence. A connective read refers to a read containing two segments aligned to different parts of the vector sequence. Specifically, the method analyzes aligned sequencing reads to identify connective reads having a first segment mapped to one region of the vector sequence and a second segment mapped to another non-homologous region of the vector sequence. These connective reads indicate that two regions of the vector sequence are joined together within a single read. Detecting connective reads in this manner allows the method to determine the presence of vector tandem events, where independent parts of the vector sequence have fused together. This provides insights into the structural characterization and integrity of the integrated vector.

[0074] The method may further include identifying chimeric reads from aligned reads. Chimeric reads are sequencing reads that partially map to the vector sequence and partially to a non-vector region of the reference genome. By examining these chimeric reads, which contain segments aligned to both vector and non-vector genomic regions, the method can obtain insights into the genomic location of vector integration. Analyzing chimeric reads can help determine the location of vector sequence integration within the host genome, providing information about the insertion site and the boundary between vector and non-vector genomic regions. This additional analysis of chimeric reads enhances the characterization of vector integration and sequence structure obtained through the copy number determination process.

[0075] To detect vector tandem events and identify the precise insertion location of the vector into the host CHO genome, the method analyzes ligature reads in sequencing reads. A ligature read is a read containing two segments aligned to different parts of the vector sequence. The method extracts sequencing reads aligned to the vector reference sequence from a global alignment file. These extracted reads are then realigned to the vector reference sequence using common read alignment software (e.g., consed / phrap). This realignment process aims to identify chimeric reads, i.e., reads aligned to different non-homologous parts of the vector sequence. A ligature location is considered valid if supported by multiple different chimeric reads. By determining vector-vector ligatures based on the identified chimeric reads, the method can thus determine the precise insertion location of the vector within the reference genome (including both the host genome and the vector sequence).

[0076] The method further includes de novo assembly of the identified chimeric reads and their respective paired reads to generate longer, continuous sequences, also known as contigs. These contigs represent continuous extensions of the DNA sequence. The de novo assembly process joins sequencing reads that provide only short segments of the complete sequence to reconstruct longer fragments. By assembling short reads that are partially mapped to the vector sequence and partially mapped to non-vector regions, and by assembling a read in which one paired read is mapped to the vector and the other to a non-vector region, the method aims to generate longer contigs that span the boundary between the vector insertion and the host genome sequence. These assembled contigs provide information to help identify the precise connection points where the vector has been integrated into the reference genome.

[0077] The method further includes aligning any assembled contigs to a reference genome, which consists of both the host genome and the vector sequence. This alignment process aims to identify the boundary between the reference genome and the vector sequence. Specifically, assembled contigs generated from chimeric reads are aligned to the reference genome. This reference genome includes sequences from the host genome and the vector sequence. By aligning the contigs to this combined reference genome, the locations where the reference genome sequences and the vector sequences are joined can be identified. This process thus delineates the boundary between the genomic regions of the reference genome and the vector sequence. Furthermore, an alignment file is created from this alignment of the assembled contigs to the reference genome. This newly formed alignment file captures the determined boundary between the reference genome and the vector sequence.

[0078] The method can utilize a system for determining copy number. This system includes a processor configured to perform the method. The processor is capable of performing any of the steps described in the method claims. Specifically, the processor can isolate and purify genomic samples, create a shotgun library, sequence the library to generate reads, align the reads to a reference sequence, partition the reference genome into multiple sizes, determine the average read coverage for each partition size, determine the copy number of each consecutive coverage peak, and build a model to estimate the copy number of the target sequence. The processor is used to perform various operations of the method, including sequence alignment, genome partitioning, calculating coverage statistics, allocating copy numbers, and developing copy number estimation models. By utilizing this system with a configured processor, the method can determine the genome copy number in an automated, computational manner.

[0079] The method can be implemented by a device containing a set of processor-executable instructions operable. This set of processor-executable instructions is configured to cause a processor to perform one or more operations of the method described herein. Specifically, the processor-executable instructions allow the processor to align reads to a reference sequence, divide a reference genome into multiple sizes, determine the average read coverage for each size of the divided reference genome until a predetermined number of consecutive average read coverage peaks are formed, with equal coverage distances between each peak, determine the corresponding copy number associated with each of these consecutive peaks, and build a model configured to estimate the copy number of a target sequence. By utilizing the processor-executable instructions, the method can be implemented using a device containing a processor to automate the steps of determining the copy number of a genomic region.

[0080] The method may involve using a computer-readable medium storing instructions that, when executed by a processor, cause the processor to perform multiple steps. First, the processor is instructed to align generated reads from sequencing to a reference sequence. This reference sequence consists of a reference genome and a vector sequence. Next, the processor is instructed to divide the reference genome into multiple different sizes. Then, the processor determines the average read coverage for each size of the divided reference genome. It continues this determination until a predetermined number of consecutive average read coverage peaks are formed, with equal coverage distances between each peak. After this, the processor determines the corresponding copy number associated with each of the consecutive average read coverage peaks. Finally, the processor is instructed to build a model configured to estimate the copy number of the target sequence. In this way, the computer-readable medium can store instructions that guide the processor to perform methods for aligning reads, dividing the genome, determining coverage, allocating copy numbers, and creating a copy number estimation model.

[0081] The method can also utilize a computer-readable medium configured to cause a processor to perform additional steps of the method. Specifically, the computer-readable medium can instruct the processor to perform various aspects of the method described herein. These additional implementations may include isolating and purifying samples, creating shotgun libraries using different sequencing technologies, dividing a reference genome into sizes ranging from 50 to 51,200 base pairs, determining copy number by comparing peak distances or using regression analysis, developing models to estimate copy number without requiring genomic position parameters, and applying the method to characterize CHO cells, mammalian cancer cell lines, or specific genes of interest. Thus, the computer-readable medium allows the processor the flexibility to perform any combination of these additional steps to determine genome copy number.

[0082] One embodiment of the computer-readable medium used in the method is non-transitory. Furthermore, the computer-readable medium can cause the processor to divide the reference genome into multiple sizes, determine the average read coverage of each size until a predetermined number of consecutive coverage peaks are formed and are equidistant from each other, and determine the corresponding copy number for each peak. The computer-readable medium storing the instructions can be non-transitory, meaning it is not transient but a physical medium, such as a hard disk drive, USB drive, solid-state drive, or other fixed storage medium, rather than a transient medium such as a signal or carrier wave.

[0083] The method further includes extracting genomic DNA (gDNA) from mammalian cell lines to obtain a genomic nucleic acid sample. The mass of the gDNA sample is in the range of 300 ng to 2000 ng. The gDNA sample is then tagged to form labeled gDNA fragments. First, the sample is cleaned using sample preparation beads. Then, an index is ligated to the labeled gDNA fragments. Finally, the sample is cleaned a second time using sample preparation beads to obtain a library for next-generation sequencing (NGS).

[0084] The method described herein can be used for the molecular characterization of Chinese hamster ovary (CHO) cells or Homo sapiens cell lines commonly used for the production of recombinant proteins. CHO cells and certain Homo sapiens cell lines are often preferred cell types for recombinant protein production because they are capable of performing complex post-translational modifications of proteins. However, long-term culture of these cell types can present challenges, such as the decline in recombinant protein expression levels over time or changes in protein structure that affect function and safety. This method allows for the molecular characterization of CHO cells or related Homo sapiens cell lines at the genomic level, facilitating the assessment of copy number stability, transgenic sequence integrity, and any alterations within the host genome. This molecular analysis supports the selection and validation of optimal cell lines for the safe and effective production of recombinant therapeutic proteins.

[0085] The method may involve isolating and purifying genomic nucleic acid samples from biological samples. This step provides raw genomic material for subsequent analysis. The sample may contain homogeneous cell types or heterogeneous mixtures of multiple cell types. To isolate nucleic acids, the sample first undergoes cell lysis to break the cell membrane and release its contents. This allows for the extraction of genomic DNA. Preferably, the extracted genomic DNA has a high molecular weight to avoid shearing and fragmentation in subsequent processing steps. High molecular weight genomic DNA helps maintain DNA integrity, which can be used to create an unbiased library representative of the genomic sample and obtain accurate sequencing results.

[0086] The method utilizes a PCR-free library preparation workflow to reduce potential bias and errors that may be introduced by PCR amplification. Specifically, when creating shotgun libraries from purified genomic nucleic acid samples, a tag-based method is used that does not include a polymerase chain reaction (PCR) amplification step. By employing a PCR-free method, this helps to reduce the number of mutations, insertions, or deletions that occur compared to library preparation methods that use PCR amplification steps. Therefore, the act of creating shotgun libraries in this method uses a PCR-free workflow when constructing libraries using next-generation sequencing technologies. Utilizing a PCR-free method for library preparation helps to generate unbiased representativeness of the genomic sample for subsequent analysis.

[0087] The method may further include adjusting the volume of the sample preparation beads used during the first or second purification step to alter the median size of fragments within the library. By changing the volume of the sample preparation beads used in the initial or subsequent purification stages, the median length of the fragments comprising the sequencing library can be altered. This allows for optimization of fragment size distribution based on the intended sequencing application or downstream analysis.

[0088] The method modifies the median fragment size of the library by adjusting the volume of the sample preparation beads during a first or second purification step. In one embodiment, the method adjusts the volume such that the median fragment size is greater than 350 base pairs. Adjusting the volume of the sample preparation beads used in the purification step allows for regulation of the final median fragment size of the DNA insert in the library. A median fragment size greater than 350 base pairs allows for more robust detection of integration sites compared to a smaller median fragment size.

[0089] The method may involve performing a first purification step after tying the genomic DNA sample and before indexing. This first purification step uses sample preparation beads to purify the sample. Specifically, this first purification step includes using sample preparation beads with a volume selected from the range of 30 to 50 microliters. By selecting a sample preparation bead volume of 30 to 50 microliters in the first purification step, the method can modify parameters such as the median fragment size of the library produced by the downstream library preparation process.

[0090] The method may involve adjusting parameters in the library preparation step to modify the characteristics of the resulting DNA library. Specifically, the second purification step in the shotgun library preparation process from purified genomic nucleic acid samples includes using sample preparation beads with volumes selected from the range of 10 μL to 20 μL. By selecting a sample preparation bead volume within this specified range for the second purification step, characteristics such as the size distribution profile of fragments in the resulting sequencing library can be adjusted. The size distribution of DNA fragments can affect downstream steps, such as the clustering of template molecules and sequencing accuracy; therefore, modifying this parameter provides a means to optimize the quality of sequencing data output.

[0091] The method may involve isolating and purifying a genomic nucleic acid sample. A shotgun library is then created from this sample. The shotgun library is sequenced to generate reads, and a processor aligns these reads to a reference sequence containing a reference genome and a vector sequence. The processor divides the reference genome into multiple sizes. The processor determines the average read coverage for each size of the divided reference genome and continues this determination until a predetermined number of consecutive average read coverage peaks are formed, with equal coverage distances between each peak. The processor then determines the corresponding copy number associated with each of these consecutive peaks. The processor can then build a model configured to estimate the copy number of the target sequence.

[0092] Furthermore, in steps involving adjusting the volume of sample preparation beads during the first and second purification steps to change the median fragment size of the library, if the method uses a volume of 33 μL of sample preparation beads in the first purification step and a volume of 15 μL of sample preparation beads in the second purification step, it can further change the median fragment size of the library.

[0093] The method may involve adjusting the volume of sample preparation beads used during the first and second purification steps to alter the median fragment size in the library. Specifically, when the volume of sample preparation beads used in the first purification step is 45 μL and the volume used in the second purification step is 12 μL, it allows researchers to control the size of the fragments obtained in the DNA library created for next-generation sequencing. Controlling fragment size in this way helps researchers optimize the library for downstream sequencing applications and data analysis.

[0094] The method can also involve reducing sequencing errors that occur during high-throughput next-generation sequencing (NGS). This is achieved by performing high-throughput NGS on a DNA library prepared according to the following steps: extracting genomic DNA from a mammalian cell line to obtain a sample; tagging the sample to form labeled DNA fragments; performing a first purification of the sample using sample preparation beads; ligating an index to the labeled fragments; and performing a second purification of the sample using sample preparation beads to obtain the final library. Sequencing this library, produced with fewer mutations, insertions, or deletions, helps minimize errors introduced during sequencing compared to methods using PCR amplification.

[0095] The method described herein can also involve, in some specific embodiments, preparing shotgun libraries with an average fragment length of approximately 1500 base pairs, and sequencing them using Nanopore sequencing technology. The genomic nucleic acids are isolated from a clonal cell line derived from Chinese hamster ovaries (CHO), specifically comprising genomic DNA (gDNA). The gDNA is fragmented to a size of approximately 0.3 to 2 kilobases, and then the fragmented DNA is converted into a sequencing-compatible library. Furthermore, the resulting DNA library can optionally be quantified using a Qubit fluorometer to ensure target sequencing performance. Brief description of the attached diagram These and other aspects will become more apparent from the following detailed description of various embodiments of the present disclosure, taken in conjunction with the accompanying drawings, in which: Figure 1 This is a block diagram illustration of a system configured to estimate the copy number of a target sequence by generating and applying a model, according to one embodiment of this disclosure; Figure 2 This is a block diagram illustration of a computing device configured to estimate the copy number of a target sequence by generating and applying a model, according to one embodiment of the present disclosure. Figure 3 This is a schematic diagram illustrating the alignment of a sequencing read with a reference sequence containing a reference genome and a vector sequence according to one embodiment of the present disclosure. Figure 4 This is a schematic diagram illustrating, according to one embodiment of the present disclosure, the division of a reference genome into segments of different sizes and the determination of segment coverage; Figure 5 This is a graph illustrating the correlation between coverage and copy number of a copy number estimation model according to one embodiment of the present disclosure. Figure 6 The illustration shows a graph of gene density and read coverage on a known copy number CHO gene, used to create a correlation curve between coverage and copy number, according to one embodiment of the present disclosure. Figure 7 This is a graph illustrating an embodiment of the present disclosure, showing how copy number estimation is optimized based on the ploidy distribution of genomic regions of different sizes. Figure 8 A graph showing the standard deviation as a function of target fragment size and its copy number estimation according to one embodiment of this disclosure is illustrated; and Figure 9 This is a normalized graph illustrating one embodiment of this disclosure; Figure 10 A flowchart of a method 1000 for estimating the copy number of a genetic sequence (e.g., a transgene integrated in a CHO cell line for recombinant protein production) according to one embodiment of the present disclosure is shown. Figure 11 A comparison of insert fragment size distributions generated by different library preparation schemes from several implementations is provided for use with... Figure 1 The systems are used together to compare the insert length distributions of several PCR-free library preparation protocols according to one embodiment of the present disclosure; Figure 12 An embodiment according to this disclosure is shown, using Illumina ® DNA PCR-free library preparation workflow generated for use in Figure 1 Uniformity of coverage generated by the system's NGS library; Figure 13 According to one embodiment of this disclosure, a bioanalyzer trace for the fragment size distribution of cleaved DNA used in Nanopore library preparation confirms the effectiveness of the fragmentation process; Figures 14A-14C A comparative analysis of sequencing coverage between Nanopore and Illumina sequencing methods is provided according to one embodiment of this disclosure, showing the correlation and consistency of coverage on vector sequences; Figure 15 A comparison of copy number predictions for genes of interest (GOIs) derived from Nanopore and Illumina sequencing, according to one embodiment of this disclosure, is presented; and Figure 16 The image shows a screenshot of the Integrative Genomics Viewer (IGV) with full coverage of GOI via Nanopore sequencing, according to one embodiment of this disclosure. Invention Details Figure 1A block diagram illustration of a cloud-based system 100 for estimating the copy number of a target sequence (e.g., a target genetic sequence) by creating and utilizing models, according to one embodiment of this disclosure, is shown. System 100 includes a cloud service provider 102, one or more personal computers 104, a genetic sequencer 103, and a mobile device 106. System 100 also includes a genetic sequence analyzer component 112. The genetic sequence analyzer component 112 can be used to generate models that can be used to estimate the copy number of the target sequence, as described in more detail below. The copy number estimation of genetic material from sample 101 can be used to accept or reject the use of cell lines for therapeutic purposes (e.g., using Chinese hamster ovary (CHO) cells to insert transgenes into the CHO genome via a vector for recombinant protein synthesis). In other embodiments, the sample may be one of the following: CHO DG44, CHO-S, C0101, F435, CHO DXB11, CHO K1 SF, CHO K1 PF, CHO K1 ATCC, CHO K1 ECACC, CHOZn, human embryonic kidney 293 (HEK-293), Per.C6 cells, CAP-T, NSO, Sf9, and Sf21 insects.

[0098] Cloud service provider 102 can be configured to facilitate the analysis of genomic data and the generation of models by providing on-demand high-performance computing capabilities. In some embodiments of this disclosure, cloud service provider 102 may be a managed service, such as a company providing cloud computing services to businesses and individuals, enabling cloud service provider 102 to provide the infrastructure, software, and platform required to host, manage, and deliver cloud-based services. In some embodiments, cloud service provider 102 may provide Infrastructure as a Service, Platform as a Service, and / or Software as a Service. Cloud service provider 102 can be configured to scale its computing resources up or down based on the user's needs at a given time.

[0099] The overall reference system 100, personal computer 104, genetic sequencer 103, and mobile device 106 can communicate with each other via network 108. Network 108 can be Wi-Fi, Ethernet, Bluetooth, etc., and can utilize the Internet and related protocols such as TCP / IP. Network 108 can be a local area network, wide area network, physical bus (e.g., Universal Serial Bus), Internet, or some combination thereof. Computer 104 and / or mobile device 106 can be used to communicate, control, or engage with sequencer 103 and / or genetic sequencer analyzer component 112 via network 108.

[0100] Sample 101 is a biological sample containing DNA that requires sequencing to obtain detailed genetic information. Sample 101 may include clonal cell lines with specific genomic regions containing an integrative vector carrying a transgene. These transgenes can be designed for various purposes, such as the production of recombinant proteins for therapeutic purposes or cellular genetic modification for specific research goals. Vectors carrying transgenes can vary in size, complexity, and diversity depending on the specific target of genetic modification. Typically, vectors carry transgenes that include DNA sequences encoding proteins or hereditary traits of interest, as well as regulatory elements, such as promoters, enhancers, or terminators, to ensure proper expression and regulation of the transgene. Therefore, Sample 101 may include Chinese hamster ovary (CHO) cells with transgenes inserted via vectors to create recombinant proteins for therapeutic purposes.

[0101] Genetically modified organisms (GMOs) can be integrated into the host genome through various methods, including plasmid-based transfection, viral vector-mediated gene delivery, or other gene-editing technologies such as CRISPR / Cas9. These technologies allow for targeted insertion of GMOs into specific genomic sites or random integration into the genome. Once integrated, the GMO becomes a permanent part of the host genome and is passed on to daughter cells during cell division. The copy number of the GMO can vary in different cells within a cloned cell line or a specific genomic region.

[0102] To sequence sample 101, a genetic sequencer 103 was used. The genetic sequencer 103 is a high-throughput sequencing instrument capable of generating a large number of DNA reads 105 in parallel. The sequencing process can begin with the preparation of a PCR-free shotgun sequencing library from sample 101. The library preparation workflow can follow a PCR-free method, which helps reduce potential PCR amplification bias and errors. Various commercial kits, such as those from Illumina, are available. ® DNA PCR-free Prep, Tagging Kit, Truseq DNA PCR-free Kit, Perkin Elmer NextFlex Rapid XT Kit, and Qiagen QIAseq FX PCR-free DNA Library Kit can be used to prepare PCR-free libraries.

[0103] Once the library is prepared, it is loaded onto a genetic sequencer 103, which performs the sequencing process. The genetic sequencer 103 uses high-throughput next-generation sequencing technologies, such as sequencing by synthesis (SBS), to generate millions or billions of short reads 105 from the DNA fragments in the library. The resulting reads 105 represent short segments of the genetic material present in sample 101. These reads 105 typically have a fixed length of 150 bases, but can also range from 50 to 600 bases or longer depending on the specific sequencing platform and technology used.

[0104] Read data 105 is transmitted via network 108 to genetic sequencer component 112. Upon arrival at genetic sequencer component 112, read data 105 is received and processed by sequence read ingestor 123. Sequence read ingestor 123 performs the initial steps of ingesting read data 123 into the analysis pipeline of genetic sequencer component 112. Read ingestor 123 performs various tasks to process and prepare read data for subsequent analysis steps. This may include data validation, quality control checks, and formatting to ensure compatibility with downstream analysis processes. The reads are then stored as read sequences 151 in database 132.

[0105] Genetic sequence analyzer component 112 also includes alignment component 114, which aligns read data 105 to a reference sequence, which may include a reference genome (e.g., CHO), vector sequence (e.g., ... Figure 3 (as shown) and other sequences. Alignment component 114 uses algorithms and bioinformatics techniques to identify the corresponding positions of read sequences within the reference genome.

[0106] Refer again Figure 1 and Figure 3 The alignment component 114 processes the incoming reads from read sequence 151 and performs sequence alignment. During alignment, the alignment component 114 uses alignment algorithms (such as the Burrows-Wheeler alignment program (BWA) or other industry-standard solutions) to compare each read sequence of read sequence 151 with the reference sequence 142. These algorithms efficiently match and / or identify the best-fit alignment positions of the numerous reads 105 generated by the genetic sequencer 103 with the reference sequence.

[0107] The alignment component 114 not only aligns the reads to the reference sequence 142, but also takes into account specific parameters and variations relevant to the PCR-free library preparation workflow. This specialized alignment strategy ensures accurate alignment and reduces potential biases or errors that may arise from the library preparation process.

[0108] The alignment component 114 generates output data, including the alignment read position, alignment quality score, and other relevant information. This alignment data serves as the basis for subsequent copy number estimation, sequence variation detection, and structural variation analysis.

[0109] Following the alignment step, the copy number estimator 124 in the genetic sequence analyzer component 112 generates model parameters 146 to create a model for estimating the copy number of the target sequence. The copy number estimator 124 uses the aligned read positions and alignment quality scores from the alignment component 114 as input.

[0110] In generating the model, copy number estimator 124 employs various statistical and machine learning techniques. It analyzes the distribution of aligned read coverage on the reference sequence to identify patterns and correlations associated with copy number variations. In some implementations, copy number estimator 124 assumes that the actual copy number of any genomic region in the sample is proportional to the observed coverage. Since vectors or genes of interest (GOIs) can be sequenced along with the host genome (e.g., ...), ... Figure 3 As shown in the diagram, the copy number estimator can estimate the copy number of the former by utilizing the correlation between the latter's coverage and copy number. For this purpose, it can be done as follows: Figure 4 The figure shows the estimated coverage of host genes and vector bases.

[0111] Through data processing and analysis, copy number estimator 124 derives model parameters that effectively capture the relationship between read coverage and copy number. Therefore, model parameters 146 may include regression coefficients, intercepts, and other factors characterizing the relationship between read coverage and copy number. These model parameters 146 can be derived through an optimization algorithm that minimizes the error or difference between the predicted copy number and the known copy number obtained from a validation method. The optimization process aims to create a model that provides accurate and reliable copy number estimation based on read coverage. Figure 5 A model is shown where the median coverage corresponding to copies 1, 2, and 3 (derived from coverage analysis of thousands of genes with known copy numbers) Figure 6 Linear regression was used to correlate with copy number.

[0112] refer to Figure 1 and Figure 4 The copy number estimator 124 first divides a portion of the reference sequence 142 into partitioned sequences. These partitioned sequences 142 are then aligned to the reference sequence 142 using alignment component 114, and this data is subsequently stored in the partitioned sequence 144 in database 132. The copy number estimator 124 then obtains the coverage of different segments of varying sizes from the partitioned sequence 144, ranging from 50 bp to 12800 bp. These segments are used to plot the ploidy distribution, such as... Figure 7 As shown, create scatter and line charts with coverage as the x-axis and occurrence count as the y-axis. Figure 7 This is a graph illustrating the ploidy distribution based on genomic region coverage of different sizes to optimize copy number estimation (e.g., 12800 bp). Fragment sizes were fine-tuned to optimize peak visibility and resolution, such as... Figure 8 As shown.

[0113] In some additional implementations, some cell lines can use custom fragment size selection to obtain clear peaks. Figure 9 Specifically, we will examine peak visibility and resolution at different fragment sizes, from 1kb to 100kb. Figure 7 For each fragment size, partitions on the host genome are sequentially acquired to obtain and utilize the largest data points. It is worth noting that, using this method, fragment partitions of different sizes may overlap (e.g., the 800bp partition chr1:1-800 overlaps with the 1600bp partition chr1:1-1600).

[0114] In one specific implementation, after determining the optimal fragment size, we will locate the dominant peak in the ploidy distribution and scan its left and right flank regions using sliding windows (20-200 data points per window). Within each window, we will fit a trend line using multinomial regression (degree less than 10) and calculate the derivative at each location. Finally, after collecting data from all windows, we will calculate the mean derivative at each location. Starting from the dominant peak, the first negative mean derivative block (block size greater than 5 data points) will indicate the location of the subpeak.

[0115] In some additional implementations, a convolutional neural network (CNN) is used to predict cell lines based on ploidy distribution. Coverage on the x-axis is normalized by the location of the dominant peak (see...). Figure 9 The occurrence count on the y-axis is normalized by the total occurrence count across all coverage areas. The normalized values ​​are used as 1D signals for prediction to train the CNN, taking into account parameters such as the maxima of the main peak and the peaks used for peak identification.

[0116] Then, copy number estimator 124 determines the correlation between coverage and copy number at the optimized fragment size to generate model parameters 146 (see example). Figure 5 and Figure 9 The median coverage of each qualified peak is considered as the expected value. The gap between major and sub-major ploidy peaks defines 1× coverage. The algorithm employs regression methods, such as least squares, to assess correlation and aims to obtain a high Pearson correlation coefficient close to 0.999.

[0117] Based on such Figure 5 The trend line derived from the correlation analysis shown (and therefore the model parameter 146) Figure 1 The copy number estimator 124 can estimate the most probable copy number of a target sequence (e.g., a gene of interest) found in the read sequence 151. Therefore, the copy number estimator 124 can estimate the specific copy number of one or more genes of interest.

[0118] To calculate the 95% confidence interval for predicted copies of the gene of interest (GOI), the following steps can be performed. First, scan each major host chromosome and obtain large segments of different ploidies, ensuring continuity and consistent ploidy, which in some specific implementations is 20 units or 5 Mb (see [link to implementation details]). Figure 4Next, the median coverage of each segment is divided by 1 × coverage to calculate the expected copy number for each segment. For segments with known copy numbers, coverage is extracted for each of the nine segment sizes, and the standard deviation (std or σ) for each size is calculated (see [link to relevant documentation]). Figure 6 and Figure 8 ).

[0119] A generative equation is used to represent the negative linear correlation between log2(fragment size) and σ, thus allowing prediction of σ for fragment sizes out of nine standard fragment sizes, with log2(fragment size) having two decimal places of precision (see...). Figure 8 Subsequently, in the specific implementation plan, a data table is created containing the predicted σ values ​​for all segments of the major chromosome. Then, a positive correlation between copy number and σ is identified, specifically a linear positive correlation between log2(copy number) and exp2(σ).

[0120] Using this equation and data table, the fragment size σ is predicted based on the existing copy number, with log2(copy number) having two decimal places of precision. This prediction is achieved through a two-dimensional lookup table, allowing estimation of σ for any given log2(fragment size) (two decimal places of precision) and log2(copy number) (two decimal places of precision) via table lookup. Using the coverage σ in the reference table, the ±95% confidence interval for each GOI prediction is calculated using log2(GOI copy number) and log2(GOI size).

[0121] Then, the trend line established in step six is ​​used to convert the coverage σ to the copy number σ, which represents the linear correlation between coverage and copy number. If the predicted copy number of GOI is less than 1, then log2(one copy of GOI) and log2(GOI size) are used for prediction. Otherwise, log2(predicted copy number of GOI) and log2(GOI size) are used for prediction.

[0122] Refer again Figure 1 Once model parameters 146 are generated, copy number estimator 124 applies the model to estimate the copy number of the target sequence based on read coverage stored in the read sequences. It utilizes alignment data obtained from alignment component 114, including read positions and coverage information, to predict the copy number of the target sequence. The observed read coverage is converted into a copy number estimate using the model parameters, taking into account the specific characteristics of the PCR-free library preparation workflow. Copy number estimation involves applying a model equation that incorporates regression coefficients, intercepts, and other relevant factors determined during model generation. By inputting read coverage from the alignment data into the model equation, copy number estimator 124 calculates the estimated copy number of the target sequence.

[0123] The resulting copy number estimates provide valuable information about the abundance and representativeness of target sequences in the sample. They help characterize genomic content, such as the presence of integrated transgenes or specific genetic regions. Copy number estimates obtained from the model can be used for a variety of purposes, including therapeutic cell line selection, gene expression analysis (e.g., gene expression or transcriptomics may be correlated with DNA copy number, and / or the availability of regulatory sequences may be correlated with gene expression levels), and genetic engineering applications.

[0124] It should be understood that the specific implementation of the copy number estimator 124 and modeling techniques described herein can vary. Different statistical, mathematical, or machine learning methods can be used to generate model parameters 146 and estimate the copy number based on the alignment read coverage. The configuration, algorithm, and methods used by the copy number estimator 124 can be adjusted or customized according to specific applications, technological advancements, and improvements in data analysis techniques.

[0125] In addition to copy number prediction, further analysis is performed on sequencing read 105. Read 105 is analyzed to detect the vector integration site in the host genome. Furthermore, the method can determine the precise insertion location of the vector within the host CHO genome by analyzing a globally aligned BAM file. Partially mapped reads (called chimeric reads) and their paired reads, as well as reads where one paired read is mapped to the vector and the other to a non-vector region, are examined. These reads can be de novo assembled to generate longer contigs. The assembled contigs are then aligned to a reference genome (including both the host genome and the vector sequence) and other sequences of interest to create a BAM file. The “SA” tag in the BAM file can be used to identify the host-vector boundary. In some implementations, the vector portion of the assembled contigs is masked to create a second set of contigs (masked groups). Both unmasked and masked contigs are then mapped to the reference genome. If both unmasked and masked versions of the contigs align to the same host genome location, the integration site in the host genome is considered valid.

[0126] Chimeric reads that partially map to the vector and partially to non-vector regions, as well as reads in which one paired read maps to the vector and the other to a non-vector region, are examined. These reads are de novo assembled to generate longer contigs, which are then aligned to a reference genome to identify donor DNA-host boundaries and insertion sites within the host genome. Repeat masking annotation is used to improve the accuracy of insertion site identification in repeating regions. The method may involve detecting vector tandem events and identifying the precise location of vector insertion in the host CHO genome.

[0127] In other words, to detect vector tandem events, ligation reads in sequencing data are analyzed—reads containing two segments aligned to different parts of the vector. In some implementations, sequencing reads aligned to the vector reference sequence are extracted from the global alignment file. These extracted reads are then re-aligned to the vector reference sequence using consed / phrap software or other commonly used read alignment software. This re-alignment process aims to identify vector chimeric reads, i.e., reads aligned to different non-homologous parts of the vector sequence. A ligation site is considered valid if it is supported by four or more different chimeric reads.

[0128] Sequence variants in the integration vector were also detected by analyzing sequencing reads 105. Variants were collected at the read level using the pysam package, and a series of screening steps were applied to filter out low-quality variants. The filtering criteria considered variant frequency, quality score, read support from both the positive and negative strands, and distance from potential linkage sites. This filtering process helps eliminate false positive variant determinations and ensures the identification of high-quality variants. NGS reads generated during sequencing may contain variations, such as single nucleotide mutations, small insertions, or deletions, compared to the reference vector sequence. These variations may be genuine mutations or sequencing errors introduced during library preparation (especially if PCR amplification is involved). To identify these variants, the method uses the pysam package to collect variants at the read level. Subsequently, a series of screening steps are implemented to identify true sequence variants. First, reads containing approximately 5% or more of the bases that are variants are excluded. Then, in this specific embodiment, only sequence variants with a Phred quality score greater than 15 are considered to ensure a high confidence level in variant determination. Furthermore, the final variant may be supported by reads from both the positive and negative strands, originating from different read pairs, and located approximately 8 or more base pairs from the nearest junction or structural variation. These filtering criteria, tailored to PCR-free library preparation workflows, help eliminate false-positive variant detection and improve variant identification accuracy.

[0129] Finally, a Circos diagram can be generated to visualize and analyze various features of the integration vector for display on the GUI component 118. This diagram includes the gene of interest (GOI) location, read coverage, repetitive regions, split read junctions indicating potential vector tandem events, and the trajectory of variants. This visual representation aids in the interpretation and analysis of sequence data, providing insights into the structure and characteristics of the integration vector and its relationship to the GOI. A Circos diagram is employed to visualize and analyze various features of the integrated donor DNA, such as read coverage anomalies, vector repetitive regions, tandem vector sequence regions, vector sequence variants, and the integration location in the host genome. In one implementation, the Circos diagram is created using Python and R scripts, enabling comprehensive analysis and visualization of the integration vector sequence within the CHO genome.

[0130] In summary, the sequencing process for sample 101 involved preparing a PCR-free shotgun sequencing library, performing high-throughput sequencing using a genetic sequencer 103, and generating short reads 105 representing the genetic material in the sample. These reads 105 were then subjected to bioinformatics analysis, including alignment to a reference genome, copy number determination, detection of vector integration sites, identification of sequence variants, and generation of visualizations such as Circos diagrams. These analyses provide detailed information about the integrated transgenic clone, its copy number, structural variations, and sequence variants, contributing to a comprehensive understanding of the sample's genetic characterization.

[0131] Refer again Figure 1 In some implementations, a dedicated application for engaging with the genetic sequencer component 112 may be used, such as a mobile application on mobile device 106 or a desktop application on personal computer 104. Communication may include transmitting data in HTML, XML, JSON, YAML, or any data format. The genetic sequencer component 112 may provide users with user-level accounts via typical login mechanisms. The genetic sequencer component 112 may be a web application, web server, web service, etc., and may utilize one or more protocols to communicate data.

[0132] Cloud service provider 102 can provide genetic sequence analyzer component 112 as a web page, a web application, or a program that can be downloaded and executed on computer 104 or mobile device 106. Genetic sequence analyzer component 112 includes copy number estimator 124, communication component 116, GUI component 118, and process model executor 120.

[0133] As previously described, in another embodiment, the copy number estimator 124 can use Monte Carlo simulation data to allocate confidence scores. In one embodiment, the frequentist confidence score can be derived using frequentist statistics. For example, the confidence score can use distributed sample data, hypothesis testing, p-values, significance tests, confidence intervals, etc. In an additional embodiment, a confidence score is calculated for each (or group of) using the posterior probability in the Bayesian estimate, representing an updated belief about the copy number estimate. Thus, the confidence score can be, for example, a posterior distribution or a confidence interval of the Bayesian estimate.

[0134] Process model executor 120 can execute one or more stored models to determine the copy number of one or more sub-components and / or execute alignment component 114. Process model executor 120 can be executable code configured to, for example, use virtual processor 124 to execute, interpret, or utilize the model to report the copy number to genetic sequence analyzer component 112.

[0135] The genetic sequence analyzer component 112 also includes a communication component 116. The communication component 116 facilitates seamless communication and data exchange between multiple software applications, devices, and systems. Specifically, the communication component 116 may include protocol processing, message formatting, data serialization, encryption, and authentication to facilitate communication with the computer 104 and / or mobile device 106. The communication component 116 may utilize message formatting mechanisms to format messages into formats such as XML, JSON, binary formats, and / or proprietary message formats. The communication component 116 may utilize various encryption algorithms, such as RSA, AES, ECC, symmetric encryption, and asymmetric encryption, to achieve secure communication between the genetic sequence analyzer component 112 and the computer 104 and / or mobile device 106.

[0136] The genetic sequencer component 112 also includes a real-time data ingestor 123. The real-time data ingestor 123, which can be a real-time or near-real-time ingestor, is configured to collect and process read sequences 151 received via the sequencer 103. Read sequences 114 can be stored in a database 132. A user account 150 can be associated with a reference sequence 142, partitioned sequences, model parameters, and / or read sequences 114.

[0137] GUI component 118 can render a display for computer 104 and / or mobile device 106. GUI component 118 can be a web-based provider, such as Flask, an HTML server, a web framework, etc. GUI component 118 can provide widgets, information, buttons, options, and menus to facilitate user interaction with genetic sequence analyzer component 112.

[0138] GUI component 118 can be used to log in user account 150 so that the user can create, save, or retrieve data from database 132, or otherwise integrate with any account functionality. Additionally or alternatively, GUI component 118 can save favorites, select default parameters, or adjust drawing parameters. GUI component 118 can instruct other components to execute instructions based on user-initiated workflows. That is, GUI component 118 can receive events such as mouse clicks, button presses, or GUI widget interactions to initiate routines, a series of steps, or a series of actions. For example, GUI component 118 can guide the user step-by-step on how to set up and use process model 144 within database 132.

[0139] The GUI component 118 can also be used to visualize the results of the copy number estimator 124, in a summary, simulation manner, and / or to provide various visualization tools for data analysis. Therefore, each user can log in to their user account 150 to visualize their analysis results.

[0140] Resource scheduler 110 can schedule requests for actions to one or more virtual servers 122, each virtual server having a virtual processor 124, virtual memory 126, and virtual disk space 128. Virtual servers 122 can execute on one or more servers 121 in server cluster 119, and are scheduled and activated by resource scheduler 110.

[0141] Figure 2 A block diagram illustration of a computing device 200 for estimating the copy number of a target sequence (e.g., a target genetic sequence) by creating and utilizing a model, according to one embodiment of the present disclosure, is shown. Figure 2 The computing device 200 can be Figure 1 The computing device 200 includes a computer 104 or mobile device 106. The computing device 200 includes an I / O interface 210 for communication therein. The computing device 200 includes a data storage device 204, a processor 206, a network interface 208, a memory 225, and a user I / O device 226. The data storage device 204 stores data and may be a hard disk drive, flash drive, thumb drive, volatile memory, non-volatile memory, semi-volatile memory, etc. The processor 206 can execute one or more processor-executable instructions 212, which may be stored in the data storage device 204 and / or the memory 225. For example, the processor 206 can execute processor-executable instructions 212 stored in the memory 225, which are retrieved from the data storage device 204. The memory 225 also includes program data 214, which may include information related to the processor-executable instructions 212. The computing device 200 may include a user I / O device 226, such as a cursor device 230 (e.g., a touchscreen or mouse), a keyboard 232 (virtual or physical), and / or a display 228 (which may be a touchscreen). The computing device 200 communicates with the network 202 via the network interface 208.

[0142] although Figure 2 The computing device 200 can be used as Figure 1 It is part of System 100, but in some implementations, the environmental impact minimization function can reside entirely within... Figure 2 Within 200 computing devices. For example. Figure 1 The genetic sequence analyzer component 112 can reside as the genetic sequence analyzer component 242. Figure 2 The processor can execute instructions within 212. For example, Figure 2 The comparison component 234, communication component 236, GUI component 238, process model executor 240, sequence segment ingestor 241, and copy number estimator 243 can each have the same characteristics as... Figure 1The comparison component 114, communication component 116, GUI component 118, process model executor 120, sequence segment ingestor 123, and copy number estimator 124 have the same or similar functions. Database 244 can be similar to... Figure 1 Database 132. Database 244 could be, for example, an SQLite 3 database embedded in computing device 200.

[0143] Therefore, in some implementations, the genetic sequencer component 242 may reside entirely on a local device (e.g., on computer 104, mobile device 106, etc.), may be partially located within cloud service provider 102, and / or may be organized in a hybrid local and cloud configuration. In some implementations, the genetic sequencer component 242 may be an application that can execute on computer 104, mobile device 106, cloud service provider 102, genetic sequencer 103, computing device 200, etc., or some combination thereof.

[0144] Figure 2 Database 244 can be similar to or equivalent to Figure 1 Database 132. That is to say, reference sequence 246, partitioned sequence 248, model parameters 250, user account 254, and read segment sequence 252 can be similar to or equivalent to... Figure 1 The reference sequence 142, the partitioned sequence 144, the model parameters 146, the user account 150, and the read sequence 151.

[0145] Figure 10A flowchart of a method 1000 for estimating the copy number of a genetic sequence (e.g., a transgene integrated into a CHO cell line for recombinant protein production) according to one embodiment of this disclosure is shown. Method 1000 begins by isolating and purifying a genomic sample (operations 1002-1004). A PCR-free shotgun library is then created from the sample (operation 1006) and sequenced to generate reads (operation 1008). The reads are aligned to a reference genome consisting of a host reference sequence (e.g., CHO) and any vector sequence (operation 1010). The host genome is divided into fragments of different sizes, e.g., from 50 bp to 12800 bp (operation 1012). Coverage of each fragment size is analyzed to identify consecutive coverage peaks with equal spacing, corresponding to ploidy levels (operation 1014). A corresponding copy number is assigned to each peak based on the coverage distance (operation 1016). Finally, a model is generated to correlate coverage and copy number (operation 1018). This model can then estimate the copy number of any target sequence (e.g., a transgene) based on observed coverage. This method offers several advantages over existing approaches. The absence of PCR-free library preparation reduces bias. In some specific implementations, the combination of genome partitioning and ploidy analysis enables accurate copy number prediction without prior knowledge of the target location. The model provides sequence-independent estimations. Overall, this method provides rapid, accurate, and cost-effective copy number and molecular characterization for cell lines used in biopharmaceutical manufacturing.

[0146] Variations can be achieved using different fragment sizes, peak numbers, regression models, confidence interval calculations, quality control analyses, and visualizations. This method helps to accurately determine the copy number of integrating vectors and transgenes, detect rearrangements, tandem events, variants, and integration sites. It represents a significant improvement over traditional Southern blotting and dPCR methods used for cell line characterization.

[0147] refer to Figure 10 An additional embodiment of method 1000 is as follows. Operation 1002 involves isolating a genomic nucleic acid sample from a biological sample to be characterized. This provides raw genomic material for subsequent analysis. The sample may contain homogeneous cell types or a heterogeneous mixture of multiple cell types, such as clonal cell lines expressing recombinant proteins. To isolate the nucleic acids, the sample first undergoes cell lysis to break the cell membrane and release its contents. Cell lysis can be achieved by various methods, including chemical lysis, enzymatic lysis, mechanical lysis by bead milling, or thermal lysis.

[0148] Once cells are lysed, the crude lysate contains a complex mixture of biomolecules, including genomic DNA, RNA, proteins, lipids, and metabolites. To specifically isolate genomic DNA, a series of extraction and purification steps are employed. Possible extraction methods include phenol-chloroform extraction, silica column-based extraction, magnetic bead-based extraction, and extraction kits using proprietary reagents.

[0149] Preferably, the extraction protocol maximizes DNA yield while minimizing shearing. This protocol may involve steps to separate proteins and other contaminants, using enzymes such as proteinase K or RNase A to digest proteins and RNA. Repeated cycles of ethanol precipitation concentrate the DNA while washing away impurities.

[0150] The quality and quantity of isolated genomic DNA can be examined. Quality assessment by gel electrophoresis or fluorescence methods can be used to determine DNA integrity, size distribution, and potential RNA / protein contamination. DNA quantification can be performed using spectrophotometry, such as UV absorption at 260 nm, or fluorescence assays using dyes that selectively bind to DNA.

[0151] The amount of starting genomic DNA depends on the downstream library preparation and sequencing methods. For PCR-free shotgun library preparation, hundreds of nanograms to micrograms of DNA can be used.

[0152] If the initial isolation results in low DNA yield or poor quality, optimization techniques can be used to increase the yield. Possible optimizations include adjusting cell number, lysis buffer, purification reagents, precipitation parameters, and extraction kit.

[0153] Once a sufficient quantity and quality of genomic DNA is obtained, it undergoes further preparation and quality control before shotgun library construction. This may include additional purification, size selection via gel extraction or SPRI beads to remove small DNA fragments, and examination of DNA size distribution. Appropriately isolated genomic DNA is used as input for Operation 1004.

[0154] Step 1004 involves purifying the genomic nucleic acid sample isolated in the previous step 1002. This purification step removes contaminants and enzyme inhibitors that may interfere with subsequent library preparation and sequencing.

[0155] Purification is achieved using various methods, such as silica-based columns, magnetic beads, filter plates, or precipitation techniques. Commercial kits utilizing these techniques are available, such as Beckman Coulter's Agencourt AMPure XP kit, New England Biolabs' Monarch Genomic DNA Purification kit, and Zymo Research's DNA Clean and Concentrator kit.

[0156] Purified DNA samples can undergo additional quality control steps, such as quantification, purity checks using spectrophotometry or fluorescence methods, and integrity analysis using agarose gel electrophoresis. Quantification determines DNA concentration, while purity checks assess for contaminants such as proteins, RNA, or salts. Gel electrophoresis verifies DNA integrity by ensuring the presence of high-molecular-weight fragments.

[0157] During purification, measures are taken to avoid DNA shearing and fragmentation. Gentle handling, wide-mouth pipette tips, and avoiding excessive vortexing help maintain high molecular weight DNA. Optimize purification parameters such as buffer conditions and bead-to-sample ratio to maximize DNA recovery.

[0158] The purified DNA was eluted in a low-salt buffer such as 10 mM Tris-HCl or TE buffer. The elution volume was adjusted to achieve the DNA concentration required for subsequent library preparation steps.

[0159] For the PCR-free library preparation described herein, in some implementations, PCR inhibitors such as phenol, ethanol, salts, or metal ions should be avoided. Additional washing steps can be used to thoroughly remove any co-purified inhibitors.

[0160] Purified DNA samples can be quantified using fluorescent dsDNA binding dyes such as Qubit or PicoGreen. Quality checks are performed using spectrophotometric ratios such as A260 / A280 and A260 / A230 to verify that protein and solvent contaminants are within acceptable limits. DNA integrity is checked by running aliquots on an agarose gel to confirm the presence of high molecular weight DNA.

[0161] Operation 1006 involves creating a shotgun library from a purified genomic nucleic acid sample. This operation prepares the nucleic acid sample for sequencing in subsequent steps.

[0162] To create a shotgun library, genomic DNA samples are first fragmented into smaller fragments by mechanical shearing or enzymatic digestion. The size of the DNA fragments is then optimized based on the sequencing technology used. For Illumina sequencing, fragment sizes of 200-500 bp can be used.

[0163] The fragmented DNA is then end-repaired to generate blunt ends. The 5'-phosphorylated blunt ends can be used for adapter ligation in subsequent operations. Optionally, the end-repaired DNA fragments can be size-selected using agarose gel extraction to obtain fragments within a predetermined size range.

[0164] The adapter oligonucleotides are then ligated to both ends of the size-selected DNA fragment. The adapter provides primer sites for the amplification and sequencing of the library fragment. Optional adapter sequences with different molecular barcodes can be used to allow for sample multiplexing.

[0165] After adapter ligation, the ligation products can undergo several purification and size selection steps to remove unligated adapters, enzyme reagents, and small DNA fragments. This improves the quality of the sequencing library.

[0166] In some protocols, fragments ligated with adapters can be amplified by PCR to enrich fragments with adapters at both ends and increase library yield. However, PCR amplification can introduce bias and errors, so some PCR-free methods omit this step.

[0167] Overall, this shotgun library preparation workflow randomly fragments the genome and adds platform-specific adapters to generate complex libraries representing the genome or target regions. This unbiased library is compatible with downstream high-throughput sequencing, enabling the parallel sequencing of millions of DNA fragments.

[0168] Specific reagents and kits used for fragmentation, end repair, size selection, adapter ligation, and purification can vary depending on the sample type and sequencing platform. The library preparation workflow offers both automated and manual options. Overall, Operation 1006 can create sequencing-ready shotgun libraries from genomic DNA samples.

[0169] Operation 1008 involves sequencing the shotgun library created in operation 1006 using a genetic sequencer 103 to generate sequencing reads 105 representing the genomic contents of the sample. This operation can utilize high-throughput next-generation sequencing (NGS) technology to sequence millions or billions of DNA fragments from the shotgun library in parallel.

[0170] In one implementation, operation 1008 uses reversible dye terminators for sequencing by synthesis (SBS). Shotgun library fragments are cloned and amplified on the surface of flowing cells, forming individual clusters. Fluorescently labeled nucleotides are then added one at a time, with imaging performed after each addition to identify the added bases. This cycle is repeated to determine the base sequence in each fragment, generating reads of 100 bp or 150 bp in length.

[0171] Operation 1008 utilizes Illumina sequencing platforms such as NovaSeq, NextSeq, and MiSeq to perform high-throughput sequencing. These instruments offer ultra-high throughput, outputting data ranging from megabytes to terabytes, ensuring comprehensive sampling and coverage of shotgun libraries. Sequencing can be performed using kits and reagents optimized for PCR-free library preparation workflows.

[0172] In another implementation, operation 1008 can utilize DNABSEQ sequencing, a tag-based method offered by Illumina. Here, the shotgun library undergoes an enzymatic fragmentation and tag ligation process called tagging. This allows for simultaneous fragmentation and tagging of DNA fragments prior to cloning amplification and sequencing.

[0173] Operation 1008 results in the generation of numerous sequencing reads 105, which represent the genomic composition of sample 101. Reads 105 can be paired-end reads, ranging in length from 100 to 250 bp, providing sequence information from both ends of each library fragment. Sequencing depth is adjusted based on the size and complexity of the reference genome to ensure optimal coverage of the region of interest.

[0174] The raw read data 105 from Operation 1008 can be evaluated for quality metrics such as base mass fraction, GC content, and kmer diversity to ensure compatibility with downstream analyses. In some implementations, it can be assumed that the sequencer output produces acceptable data because each sample will have a large number of reads. Additionally or alternatively, in some implementations, FastQC is not performed prior to the bioinformatics pipeline. In other implementations, mapping metrics in the final results can be used to check for and / or reject unacceptable samples. Overall, Operation 1008 provides data that forms the basis for subsequent bioinformatics workflows for determining copy numbers, identifying variants, or characterizing integration vectors.

[0175] Operation 1010 involves using a processor to align the sequencing reads generated in operation 1008 to a reference sequence formed from the reference genome and vector sequence.

[0176] Optimized alignment algorithms, such as BWA, BWA-MEM, Sentieon BWA, and Bowtie2, are used for alignment to efficiently map short sequencing reads to the reference genome. In some implementations, operation 1006 may take into account specific considerations for the PCR-free library preparation workflow used to generate sequencing data.

[0177] Specifically, in one implementation, the alignment process involves the following detailed steps: 1. Retrieving and preprocessing raw sequencing reads from Operation 1008. This includes filtering low-quality reads, trimming adapter sequences, and removing duplicate reads. 2. The preprocessed reads are then aligned to a composite reference sequence using algorithms such as BWA, BWA-MEM, Sentieon BWA, and Bowtie2 with optimized parameters. The reference sequence contains a host reference genome (e.g., a CHO genome) tandem with the complete vector sequence. 3. Multithreading and GPU acceleration technologies significantly improve the speed of the read alignment process. This allows for the alignment of hundreds of millions of reads within a reasonable timeframe. 4. For reads generated from longer DNA fragments, proper end-to-end alignment is enforced, requiring consistent alignment at both ends. This reduces the occurrence of spurious alignments. 5. The alignment quality score is recalibrated to correct for biases and errors that may be introduced during sequencing. This helps improve the accuracy of downstream analyses. 6. PCR duplicate reads can be folded to avoid overestimating coverage. In some implementations, this can be used to account for the duplication artifacts inherent in PCR amplification. 7. Reads can be realigned around insertions and deletions using GATK's RealignerTargetCreator and IndelRealigner tools to clean up erroneous alignments. 8. The resulting BAM file contains the alignment positions of all reads relative to the reference genome, along with metadata such as mapping quality and flags. This serves as input for downstream copy number analysis. 9. Quality control checks are performed to ensure satisfactory alignment rates, correct pairing rates, expected insert sizes based on library preparation, reasonable coverage distribution, and no contamination. 10. To improve structural variant detection, long reads from technologies such as Oxford Nanopore can be integrated into the alignment pipeline. In summary, Operation 1010 performs alignments of sequencing reads with combined reference genome and vector sequences. The output provides a foundation for accurate characterization of copy numbers, variants, and integration sites.

[0178] Operation 1012 involves using a processor to divide the reference genome into multiple sizes. This operation is performed by the copy number estimator component after the read alignment step.

[0179] This partitioning allows for the analysis of read coverage of the reference genome on fragments of different sizes. This enables the determination of the relationship between read coverage and copy number, which can be used to develop accurate copy number estimation models.

[0180] Specifically, the reference genome is divided into non-overlapping partitions of varying sizes, ranging from 50 base pairs (bp) to 51,200 bp. In one implementation, partition sizes include 50 bp, 100 bp, 200 bp, 400 bp, 800 bp, 1,600 bp, 3,200 bp, 6,400 bp, 12,800 bp, 25,600 bp, and 51,200 bp.

[0181] The copy number estimator obtains coverage information for each partition size by extracting relevant data from alignment files containing mapped sequencing reads. Coverage represents the number of sequencing reads aligned to each partition.

[0182] By analyzing coverage at different partition sizes, copy number estimators can identify trends and patterns related to the correlation between coverage and copy number. Smaller partition sizes help assess coverage variability and noise, while larger partitions are more useful for identifying copy number differences.

[0183] In one implementation, the copy number estimator uses partition sizes of 800bp, 1,600bp, 3,200bp, 6,400bp, 12,800bp, 25,600bp, and 51,200bp to estimate the copy number based on the coverage peak. Smaller partitions of 50bp to 400bp are used to estimate coverage variability and standard deviation to determine confidence intervals.

[0184] This partitioning allows for adjusting fragment sizes to optimize peak resolution in coverage-based ploidy distribution analyses. It also enables focused analysis of coverage for specific genes or regions of interest.

[0185] In summary, the partitioning operation, by fragmenting the reference genome into regions of different sizes and evaluating coverage at various partition sizes, facilitates the determination of the relationship between coverage and copy number. This supports the development of accurate models for estimating target sequence copy numbers.

[0186] Operation 1014 involves using a processor to determine the average read coverage for each size of the divided reference genome until a predetermined number of consecutive average read coverage peaks are formed, and the consecutive average read coverage peaks have equal coverage distances between them.

[0187] More specifically, the processor divides the reference genome into multiple fragment sizes ranging from 50 bp to 51,200 bp, including 50 bp, 100 bp, 200 bp, 400 bp, 800 bp, 1600 bp, 3200 bp, 6400 bp, 12800 bp, 25600 bp, and 51200 bp. The processor obtains the aligned read coverage for each fragment size from the alignment output stored in memory or on disk.

[0188] The processor calculates the median coverage value for each fragment size to determine the average read coverage. The median, rather than the average, can be used to minimize the impact of outliers.

[0189] The processor analyzes the average read coverage across the fragment size spectrum. It identifies consecutive peaks within the coverage where the distance between peaks is approximately equal. For example, three consecutive peaks can be detected at coverage levels of 100×, 200×, and 300×, indicating that the equal coverage distance between the peaks is 100×.

[0190] The number of consecutive equidistant peaks is predetermined, for example, set to three peaks. The processor continues to divide the reference genome into progressively larger sizes, determining the average read coverage for each size, until the predetermined number of equidistant coverage peaks are reached.

[0191] Once a predetermined number of peaks with equal coverage distances are identified, the associated fragment size is used for subsequent copy number determination. For example, if three equidistant peaks are found at a fragment size of 12800 bp, then 12800 bp is used to estimate the copy number.

[0192] In some implementations, a 12800 bp size represents the optimal balance between capturing integer copy numbers with clear peaks and minimizing coverage noise or variability. Smaller fragment sizes exhibit more variability, while larger fragments do not resolve different copy numbers well. In some implementations, some cell lines may be unique, thus using different fragment sizes, such as 51200 bp. Another factor is sequencing depth—low sequencing depth samples can use larger fragment sizes. However, in some implementations, the assay can produce relatively consistent sequencing depths (e.g., 300-500 million reads per sample); therefore, cell line specificity regarding sequencing depth may or may not be considered.

[0193] In some implementations, the processor can optimize or adjust fragment sizes to improve peak resolution in the coverage spectrum. For example, the processor can evaluate 10bp interval fragment sizes around 12800bp, such as 12790bp, 12800bp, and 12810bp, and determine which size produces the clearest equidistant peak to improve copy number estimation.

[0194] Then, the resulting optimized fragment size, the number of observed peaks, and the distance to equal peak coverage are used by the processor to estimate the copy number and generate a model relating coverage to copy number, as described in operations 1016 and 1018.

[0195] This detailed, multi-step process, encompassing adaptive fragment size adjustment, average coverage calculation, isometric peak identification, and fragment size optimization, enables robust and accurate estimation of genome copy number from aligned sequence reads. This algorithmic approach provides a data-driven method for selecting optimized parameters tailored to specific sample sequencing results.

[0196] Operation 1016 involves using a processor to determine the corresponding copy number associated with each of a predetermined number of consecutive average read segment coverage peaks.

[0197] In one implementation, this determination is performed by analyzing the equal coverage distance between consecutive average read coverage peaks and comparing it to the coverage value of each peak. For example, if there are three consecutive peaks with equal coverage distances between them, the first peak is considered to represent 1× copy number. This is done by comparing the coverage value of the first peak with the equal coverage distances—if they are approximately equal, the first peak is assigned a copy number of 1×.

[0198] The second peak is then assigned a copy number of 2×, since its coverage value is approximately twice the equal coverage distance. Similarly, the third peak is assigned a copy number of 3×. In this way, the corresponding copy number of each peak can be inferred based on its coverage value relative to the equal coverage distance between peaks.

[0199] In another implementation, copy number determination utilizes a linear regression model that correlates coverage values ​​with copy number. This model is trained using the average coverage value of a segmented reference genome fragment with a known copy number. This allows for the generation of a trend line capturing the relationship between coverage and copy number. Using this trend line equation, the copy number corresponding to each peak coverage value can be estimated.

[0200] In addition, quality control analysis can be performed on read coverage data to remove potential sequencing biases and errors. Thresholds can be set to filter out peaks caused by coverage anomalies. The remaining high-confidence peaks are then used for copy number estimation.

[0201] Copy number analysis can also incorporate factors such as the variability of coverage across different fragment sizes and the reliability of coverage measurements. Statistical techniques can determine confidence intervals for each copy number prediction based on fragment size and observed coverage distribution.

[0202] To further optimize copy number estimation, the ploidy distribution of the genome can be analyzed by plotting the coverage and frequency of different fragment sizes. The position and width of peaks in the ploidy distribution can provide insights into the true copy status. For example, the dominant peak typically represents the major copy number.

[0203] In summary, Operation 1016 analyzes read coverage data to infer the corresponding copy number of consecutive coverage peaks. This is achieved by comparing equal coverage distances, applying a trained regression model, quality control filtering, confidence interval calculation, and ploidy distribution analysis. The resulting copy number profile helps to accurately estimate the copy status of target gene sequences.

[0204] Operation 1018 involves using a processor to determine a model configured to estimate the copy number of the target sequence. This model is generated by leveraging the correlation between read coverage and copy number observed in the host genome.

[0205] Specifically, the processor acquires coverage information for segmented fragments of different sizes on the host genome, ranging from 50 bp to 51,200 bp, increasing exponentially. The processor calculates the median coverage for each fragment size and plots the coverage distribution. A series of evenly spaced continuous peaks are identified in the coverage map, indicating the major ploidy status of the host genome.

[0206] The distance between peaks corresponds to the expected 1× coverage. The processor assigns a putative copy number to each peak by comparing peak position to 1× coverage. This establishes a correlation between host genome observation coverage and copy number.

[0207] The processor then performs regression analysis, such as least squares regression, to fit a trend line representing the mathematical relationship between coverage (independent variable) and copy number (dependent variable). The resulting regression formula serves as a model for predicting copy number based on observed coverage.

[0208] To estimate the copy number of a target sequence (e.g., a gene of interest (GOI) within a vector), the processor determines its coverage from aligned reads. This measured coverage is then input into a regression model equation to calculate the predicted copy number of the target sequence.

[0209] The model is trained and optimized using the coverage of host genome fragments and known copy numbers. In some specific implementations, it does not require the genomic location of the target sequence. This allows for unbiased copy number prediction of any target sequence based solely on its observed coverage.

[0210] Furthermore, the processor can use the variability in coverage across different fragment sizes of the host genome to estimate the confidence interval for copy number prediction. Fragment coverage bias can be correlated with copy number to calculate the prediction error range at a predetermined confidence level.

[0211] The coverage-copy number model can be further optimized by determining the ploidy distribution of the host genome. The processor plots the frequency of observed coverage to reveal different ploidy peaks. The initial coverage-copy number ratio is then adjusted based on the location of the main peak. This improves the accuracy of the model for subsequent copy number estimation.

[0212] In summary, operation 1018 involves developing a mathematical model representing the relationship between observed sequence coverage and copy number. This model is applied to estimate the copy number of any target sequence based solely on the measured coverage obtained from aligned sequencing reads. In some embodiments of this disclosure, the model does not require prior knowledge of the target sequence location.

[0213] Example refer to Figure 1 , 11 And 12, use Figure 1 Examples of System 100 are as follows: In some implementations, DNA fragment libraries are prepared using a shotgun-based PCR-free DNA library preparation workflow. In some implementations, PCR-free DNA libraries are prepared using kits selected from the group consisting of: Illumina... ® DNA-free PCR Prep and labeling kit; TruSeq ® DNA-free PCR kit; PerkinElmer NEXTFLEX ® Rapid XT kit; and Qiagen QIAseq ® FX DNA Library Kit.

[0214] In some embodiments, the input test material used to generate next-generation sequencing (NGS) libraries comprises genomic DNA (gDNA) isolated from a Chinese hamster ovary (CHO) clonal cell line that produces immunoglobulins. Alternatively, in some embodiments, gDNA is isolated from a Homo sapiens cell line engineered to produce therapeutic proteins. In some embodiments, the engineering of CHO clonal cell lines involves the following processes: plasmid transfection, plate pooling (5000 cells / well, Gln-medium), 96-well plate static assay, fed-batch productivity assay, limiting dilution clones (0.5 cells / well), 96-well plate static assay, fed-batch productivity assay, or stability studies and fed-batch evaluation. In some embodiments, 300 to 2000 nanograms (300 ng–2000 ng) of gDNA sample are used in library construction. In some embodiments, library preparation is based on a 300 ng gDNA sample.

[0215] In some implementations, the library preparation method includes: 1. Tagging genomic DNA (manual operation: 5 minutes, total: 10 minutes); 2. Post-tagging purification (manual operation: 10 minutes, total: 20 minutes); 3. Ligation indexing (manual operation: 5 minutes, total: 45 minutes); 4. Library purification (manual operation: 15 minutes, total: 45 minutes); 5. Quantification and merging of the library; 6. Dilution to the starting concentration (reagent: RSB); and 7. Sequencing setup.

[0216] In some embodiments, modifications to step 4 (the second cleanup step) allow for adjustment of the library insert size. In some embodiments, the final cleanup (step 4) and size selection affect the resulting library. In some embodiments, double-sided solid-phase reversible fixation (SPRI) cleanup and size selection discard low and high molecular weight fragments, reducing yield but tightening the size distribution. In some embodiments, adjusting the volume of the sample purification beads (SPB) during double-sided SPRI cleanup allows for fine-tuning of the final library insert size. In some embodiments, the fragment size distribution within the library is sensitive to the volume of the sample preparation beads (SPB), especially during the first cleanup. Increasing the SPB volume in a cleanup step reduces the median fragment size. While an insert size of approximately 350 bp is optimal for many applications, longer insert sizes have been observed to allow for more robust integration site detection.

[0217] In some implementations, the DNA library is sequenced using NGS. In some implementations, NGS uses sequencing-by-synthesis (SBS) technology. In some implementations, Illumina is used. ® NextSeq™ 2000 and DNABSEQ™ G400 instruments (MGI Tech) sequence DNA libraries. In some implementations, sequencing yields approximately 400 million to 2 billion reads from the DNA library.

[0218] In some implementations, the reads are aligned to the genome of a control organism for genetic stability analysis. In some implementations, the control organism is selected from the CHO or Homo sapiens reference genome and aligned to the inserted vector sequence.

[0219] In some implementations, the NGS reads generated during sequencing may not perfectly match the reference vector sequence. In some implementations, small variants, such as single nucleotide mutations, small insertions, or deletions, exist in the sequencing reads. In some implementations, if PCR amplification is used, these variants are mutations or sequencing errors introduced during library preparation. In some implementations, the PCR-free library preparation workflow used herein allows the application of filtering algorithms to reduce potential sequencing errors.

[0220] In some implementations, the NGS library is loaded into NextSeq. ® On the 2000 instrument, sequencing and primary data processing are performed to generate Fastq reads. The resulting Fastq files are then transferred to the SB platform for further analysis using a bioinformatics pipeline. In some implementations, the pipeline output includes at least one of the following: copy number of the gene of interest, vector integration coordinates, and vector sequence variants.

[0221] Example 1. Preparation of input materials The input test material used to generate next-generation sequencing (NGS) libraries contains genomic DNA (gDNA) isolated from Chinese hamster ovary (CHO) clonal cell lines that produce immunoglobulins. The engineering of the CHO clonal cell lines involves the following process: plasmid transfection → plate pooling (5000 cells / well, Gln-medium) → 96-well plate static assay → fed-batch productivity assay → limiting dilution clones (0.5 cells / well) → 96-well plate static assay → fed-batch productivity assay → stability studies and fed-batch evaluation.

[0222] All test samples used in the development of this assay were fully characterized, and different numbers of vector copies integrated were determined by digital droplet PCR (ddPCR). The plasmid sequences used to develop the various clonal cell lines are provided in Table 1 below.

[0223] Table 1

[0224] Using QIAGEN ® The EZ1 DNA Tissue Kit was used to extract DNA from engineered CHO cell lines according to the procedure outlined in OPRD 1046. No exogenous controls were added to the test samples prior to extraction. The purified DNA was eluted with the buffer provided with the kit, and the quantity was assessed using a Qubit fluorometer (ThermoFisher) according to the manufacturer's instructions. In this example, Illumina was used. ® DNA PCR-free library preparation and tagging kits were used for library construction. Each test sample and each library required at least 300 to 2000 nanograms (300 ng to 2000 ng) of DNA. As shown in Table 2, sufficient material was extracted from each test sample for DNA library construction.

[0225] Table 2

[0226] Example 2. Evaluation Library Preparation Workflow The consistency and reproducibility of PCR-free library preparation workflows from different commercial vendors were compared, including DNA insert length distribution, genome-wide coverage uniformity, percentage of mapped reads, and percentage of Q30 bases used for NGS DNA library construction.

[0227] Figure 11A comparison of insert size distributions generated by different library preparation protocols is provided to compare insert length distributions from several PCR-free library preparation protocols from Illumina, Qiagen, and Perkin Elmer. Each library preparation workflow was repeated multiple times to demonstrate consistency in the resulting insert sizes. Libraries were prepared from 300 ng gDNA isolated from different engineered CHO-K1 clones, sequenced on NextSeq™ 2000, and analyzed using the Dynamic Read Analysis for GENomics (DRAGEN) Germline analysis workflow v. 3.8.4.

[0228] Illumina was observed ® DNA PCR-free library preparation and magnetic bead-based tagging yield the most consistent insert sizes. This method is described in detail in Standard Operating Procedure (SOP) draft BPRD 2489. In summary, the method includes: 1. Tagging genomic DNA (manual: 5 min, total: 10 min); 2. Post-tag cleaning (manual: 10 min, total: 20 min); 3. Ligation indexing (manual: 5 min, total: 45 min); 4. Library cleaning (manual: 15 min, total: 45 min); 5. Quantification and merging of the library; 6. Dilution to starting concentration (reagent: RSB); 7. Sequencing setup.

[0229] Modifications to step 4 of the standard workflow enable adjustment of library insert sizes. Final purification (step 4) and size selection affect the resulting library. Double-sided SPRI purification and size selection discard low and high molecular weight fragments, reducing yield but tightening the size distribution. Adjusting the volume of the sample purification bead (SPB) during double-sided SPRI purification allows for fine-tuning of the final library insert size. The fragment size distribution within the library is sensitive to the volume of the sample preparation bead (SPB), especially during the first purification step. Increasing the SPB volume in a purification step decreases the median fragment size. While an insert size of approximately 350 bp is optimal for many applications, longer insert sizes have been observed to allow for more robust integration site detection.

[0230] Library preparation was based on 300 ng gDNA samples. To characterize fragment size distribution, single-stranded fragments were double-stranded using four cycles of PCR. The Illumina used in the first or second purification step was varied. ® The volume of sample preparation beads (SPBs) was used to modify the median fragment size and yield in the library. ND, undetermined. Results are shown in Table 3.

[0231] Table 3

[0232] Other PCR-free library protocols analyzed include Perkin Elmer NEXTFLEX. ® Rapid DNA-seq Kit2.0, Illumina ® TruSeq™ Library Preparation Kit and Qiagen's QIAseq ® FX scheme. Qiagen's QIAseq was observed. ® The FX scheme consistently produces shorter insert fragments than other analyzable library types.

[0233] Figure 12 The uniformity of Illumina DNA coverage without PCR is shown. In some specific implementations, uniform coverage helps in the accurate detection of genetic variants in sequencing data because it can identify variants that are far from the average depth. To evaluate coverage on the genome at different GC content levels, data from Illumina... ® Normalized coverage data from a DNA-free PCR-tagged library preparation kit were plotted against GC content in the CHO-K1 cell genome. This range was specifically analyzed, as 20% to 70% of the CHO-K1 genome consists of GC sequences. Figure 12 The results showed that the PCR-free kit provided a uniform level of coverage across a wide range of GC content in whole-genome sequencing data.

[0234] Example 3. Repeatability Comparing manually prepared libraries with those prepared using the automated Perkin Elmer Sciclone. ® The main parameters of libraries prepared with CHO Clone27 using different amounts of gDNA on the G3 NGSx workstation are shown. Read quality and mapping quality are consistent; although the inserted fragments in the automated libraries are shorter, they are still within acceptable limits.

[0235] The Illumina described in this paper was tested using eight independently prepared NGS libraries as described above. ® The DNA library obtained without PCRPrep and the tagging protocol had good data quality and reproducibility. The input was 300 ngg DNA from CHO-K1 Clone27. NextSeq ® The library was sequenced using Illumina 2000. ® Data analysis was performed using the DRAGEN Germline analysis workflow v. 3.8.4. The results demonstrated the reproducibility of library parameters in seven out of eight replicates.

[0236] Integrating Nanopore sequencing for copy number determination refer to Figure 13 14, 15 and 16, use Figure 1 An example of System 100 is shown below. In addition to the methods and systems previously described for determining copy numbers of genomic regions, the embodiments disclosed herein also include the use of Nanopore sequencing technology as an alternative to shotgun sequencing for generating libraries and performing sequencing. This method allows for the analysis of longer DNA fragments, providing the capability to confirm gene integrity, detect structural variations, and estimate copy numbers.

[0237] Sample preparation and genomic DNA isolation One implementation of a method involves isolating and purifying genomic DNA (gDNA) from biological samples. Specifically, gDNA is extracted from a Chinese hamster ovary (CHO)-derived cloned cell line engineered to express recombinant proteins via an integration expression construct. The extracted high-quality, high-molecular-weight gDNA can be used for subsequent fragmentation and library preparation steps.

[0238] Genomic DNA extraction can be performed using protocols optimized for purity, integrity, and / or other parameters. In this implementation, gDNA is extracted from vector-integrated CHO cells using the Qiagen EZ1 DNA Tissue Kit. Appropriate amounts of gDNA, ranging from 500 nanograms (ng) to 1 microgram (µg), can be isolated for downstream processing. The extraction process ensures that the gDNA is properly contaminated and suitable for subsequent applications. The purity and concentration of the extracted gDNA can be assessed using fluorescence-based detection methods, such as the Qubit fluorometer, which provides quantification of double-stranded DNA.

[0239] Fragmentation of genomic DNA Once the gDNA is isolated, it can be fragmented in a controlled manner to obtain DNA fragments ranging from approximately 300 base pairs (bp) to 2 kilobases (kb), with an average fragment length of approximately 1,500 bp. This fragmentation can be accomplished using mechanical shearing methods. For example, a Covaris S220 focused ultrasound system can be used, following the guidelines in "Quick Guide: DNA Shearing with S220 Focused-ultrasonicator, Covaris".

[0240] Multiple shearing cycles can be performed to ensure uniform fragmentation within the desired size range. After shearing, the fragmented gDNA can be purified using double-sided Ampure beads to remove excessively short or long fragments and / or concentrated DNA. All bead-purified material can then be combined for subsequent steps. The quality and fragment size distribution of the sheared DNA can be assessed using instruments such as the Agilent Bioanalyzer or TapeStation.

[0241] Now for reference Figure 13 The bioanalyzer trace provides a detailed representation of the fragment size distribution of cleaved genomic DNA specially prepared for Nanopore library construction in a specific implementation.

[0242] The trace shows a main peak with an average fragment length of approximately 1,664 base pairs (bp), a parameter used for downstream Nanopore sequencing applications. This average fragment length indicates controlled and efficient fragmentation achieved through mechanical shearing (e.g., using a Covaris S220 focused ultrasound system). Precise targeting within this fragment size range (approximately 300 bp to 2 kilobases (kb)) can be adjusted to balance sufficient read length coverage with the practical limitations of sequencing technology, thereby tailoring the sequencing output.

[0243] Bioanalyzer traces further reveal the fragment size distribution, providing insights into the uniformity and accuracy of the shearing process through peak full width at half maximum (FWHM) capture. The presence of any additional peaks or shoulders may indicate variability in the fragmentation process, potentially indicating the presence of uncut genomic DNA or overcut fragments, both of which can adversely affect sequencing performance in some implementations. However, Figure 13 The traces shown exhibit a narrow distribution, supporting the conclusion that the fragmentation process was properly controlled.

[0244] In addition, the trace includes quantitative metrics such as the area under the curve (representing the total amount of DNA analyzed) and the percentage of total DNA falling within the desired fragment size range. This information can be used to ensure that most of the DNA is suitable for subsequent library preparation and sequencing.

[0245] Effective fragmentation (as demonstrated by an average fragment length of 1,664 bp) can be further validated by a subsequent double-sided Ampure bead purification step, which helps remove DNA fragments that are too short or too long. This purification process may contribute to the purity and concentration of the DNA, as verified by fluorescence-based quantitative methods (such as the Qubit fluorometer), ensuring that the DNA is suitable for high-fidelity Nanopore sequencing, in this specific exemplary implementation.

[0246] Converted into sequencing-compatible libraries Starting with 1 µg of fragmented and purified DNA material, the fragmented gDNA is then converted into a sequencing-compatible DNA library suitable for Nanopore sequencing. This process may involve end repair and A-tail addition to the DNA fragments to generate blunt-ended, 3'-adenylated DNA molecules. End repair enzymes remove overhangs and fill in recessed 3' ends, while A-tail addition adds a single adenine nucleotide to the 3' end of the DNA fragment.

[0247] These prepared DNA fragments can then be ligated with sequencing adapters provided in Nanopore library preparation kits (such as the Ligation Sequencing Kit V14 (SQK-LSK114)). The adapter ligation step attaches motility proteins and unique molecular identifiers to the DNA fragments. Motility proteins facilitate efficient translocation of DNA molecules through nanopores during sequencing, while unique molecular identifiers enable accurate read identification and demultiplexing.

[0248] After adapter ligation, the library was further purified using Ampure beads to remove excess adapters, enzymes, and other reaction components. All bead-purified material was combined. The quality and concentration of the purified library were then assessed using a Qubit fluorometer and an Agilent bioanalyzer to ensure it met the requirements for optimal Nanopore sequencing performance.

[0249] Library Quantitative and Quality Assessment Sequencing libraries can be quantified using sensitive fluorescence-based methods (such as the Qubit fluorometer), which measure DNA concentration with appropriate specificity and accuracy. Furthermore, library quality and fragment size distribution can be assessed using platforms such as Agilent Bioanalysts or TapeStation. These analyses can be used to ensure that the libraries have appropriate fragment sizes and minimal degradation suitable for high-quality Nanopore sequencing.

[0250] Nanopore sequencing program A quantitative library of approximately 35 femtomolar (fmol) can be loaded into a Nanopore flow cell, such as the PromethION flow cell (R10.4.1). Sequencing runs can then be initiated using a Nanopore sequencer (such as the PromethION 2 instrument). The resulting number of sequencing reads typically ranges from 40 million to over 140 million.

[0251] During sequencing, individual DNA molecules are driven through nanopores by an applied voltage. Changes in ionic current are measured as the DNA translocates through the pores. Each nucleotide base causes a characteristic interference in the ionic current, and these signals are decoded in real time to determine the nucleotide sequence of each molecule.

[0252] Data processing and comparison After sequencing, the raw Nanopore reads can undergo base identification and preprocessing to convert the raw signal data (with ion current interference) into nucleotide sequences. Base identification can be performed using software provided by the Nanopore platform (such as Guppy), which utilizes neural network algorithms to accurately decode the signal data.

[0253] The preprocessed reads can then be aligned to a composite reference sequence using alignment tools optimized for long read data (such as minimap2 or an enhanced Sentiion minimap2 alignment software with optimized parameters). The reference sequence contains both the host genome (e.g., the CHO genome) and the vector sequence, allowing for a complete mapping of the reads.

[0254] The alignment process accommodates the inherently high error rate of Nanopore sequencing by allowing for gaps and mismatches during alignment. Quality control checks are performed to ensure satisfactory alignment rates, reasonable coverage distribution, and contamination-free operation. The alignment output is a BAM file containing the position of reads relative to the reference genome, along with relevant metadata such as mapping quality scores, alignment markers, and supplementary information. This alignment dataset can be used for downstream analyses, including copy number estimation, variant detection, and integration site identification.

[0255] Copy number estimation and comparison The copy number of a target sequence (such as a GOI within a vector) can be estimated using a method similar to that described previously but adapted to long read data. The average read coverage of the target sequence can be calculated by determining the average depth of the reads aligned to the GOI region.

[0256] The estimated copy number can be determined by applying a model that correlates coverage with copy number. This model can be derived from calibration data or from statistical analysis that correlates coverage levels with known copy numbers in control samples.

[0257] exist Figure 15 The paper presents a comparison of copy number predictions based on Nanopore and Illumina sequencing reads. GOI copy numbers from Nanopore (represented by dots) fall within the 95% confidence interval of copy numbers obtained using Illumina sequencing (represented by light-colored bars). This comparison demonstrates the reliability and accuracy of Nanopore sequencing for copy number estimation, indicating that Nanopore sequencing is a viable alternative to traditional short-read sequencing methods for this application.

[0258] Confirmation of gene integrity and detection of structural variations Figure 16 A detailed visualization of gene-of-interest (GOI) coverage achieved through Nanopore sequencing is provided, as shown using the Integrative Genomics Viewer (IGV). This figure demonstrates Nanopore sequencing's ability to generate long reads spanning the entire GOI, ensuring accurate mapping and confirmation of gene integrity.

[0259] Figure 16The top portion features a bar that clearly marks the GOI region within the vector sequence. This visual marker spans the entire length of the GOI, providing a reference for subsequent alignment data. Genomic coordinates along the axes indicate the location of the GOI within the broader genomic context.

[0260] Below the GOI marker, a coverage trajectory is displayed, showing the depth and distribution of sequencing reads across genome regions. Coverage is represented by a series of arrows and bars, each arrow indicating a read aligned to a reference sequence. Continuous coverage on the GOI region demonstrates the uniformity and integrity of sequencing data obtained from Nanopore technology.

[0261] Alignments of individual Nanopore reads are shown below the coverage trajectory. Each read is depicted as a horizontal bar, with alignment details indicating how the read traverses the entire GOI. The highlighted annotation "6 reads cover the entire GOI" emphasizes that six independent reads completely cover the GOI, thus confirming the intact and unbroken integration of the gene within the genome.

[0262] Should Figure 16 This demonstrates the advantages of Nanopore sequencing in terms of read length and coverage continuity. The ability to generate long reads covering the entire genome region can be used to validate gene integrity, especially in applications where full-length sequences are used for functional validation, such as the production of therapeutic proteins.

[0263] One of the significant advantages of Nanopore sequencing is its ability to generate long reads that can span entire genes or large genomic regions. This capability helps confirm the integrity of GOIs, ensuring they are full-length and unchanged, which could be important for the production of therapeutic proteins.

[0264] Furthermore, long reads can improve the detection of structural variations such as insertions, deletions, inversions, or rearrangements. These variations may be difficult to detect with short read sequencing technologies due to read length limitations and difficulties in mapping repetitive or complex regions.

[0265] Integrating Nanopore sequencing into this method offers several potential benefits: Long read capability: This enables sequencing of longer DNA fragments, improving resolution of repetitive or complex genomic regions and providing more continuous coverage of target sequences. In some specific implementations, this may enhance the ability to detect and characterize large structural variations and assemble complete gene sequences.

[0266] Structural variation detection: Helps identify large-scale genomic alterations that may be undetectable by short-read sequencing technologies. Long reads can span breakpoints or junctions, allowing for proper and accurate mapping of structural variations.

[0267] Gene integrity verification: This helps to verify the complete sequence of GOI, ensuring it is full-length and unchanged. This can be important for the production of therapeutic proteins, as changes in gene sequence can affect protein function and safety.

[0268] Flexible and rapid analysis: Nanopore sequencing offers the ability to generate data in real time and adjust operating parameters, potentially providing a flexible platform for a wide range of genomic analyses. In some implementations, its rapid turnaround time and scalability make it suitable for high-throughput applications.

[0269] The use of Nanopore sequencing also allows for direct comparison with traditional short-read sequencing methods. As shown in Figures 14 and 15, the coverage consistency and copy number estimation between Nanopore and Illumina sequencing are highly correlated, with a correlation coefficient r = 0.868, validating the effectiveness of the method.

[0270] Data validation and quality control Now for specific reference Figures 14A-14C This paper presents a detailed comparative analysis of sequencing coverage between Nanopore and Illumina sequencing methods. The figure is divided into three different panels, each providing insights into the consistency and correlation of coverage among these sequencing technologies.

[0271] The top panel illustrates the base-level coverage of the vector sequence, with different lines distinguishing sequencing methods. The first line represents coverage obtained from Illumina sequencing, while the second and third lines depict coverage from two independent Nanopore sequencing replicates. A box marks the location of the gene of interest (GOI) within the vector sequence, highlighting the region of interest for copy number analysis. The coverage pattern demonstrates the consistency and reliability of Nanopore sequencing in capturing transgenic regions, and the correlation coefficient r = 0.868 indicated at the bottom shows a strong correlation between sequencing methods.

[0272] The middle panel provides a scatter plot analysis of the correlation between the two Nanopore repeats. This analysis confirms the reproducibility of the Nanopore sequencing method, with a correlation coefficient r = 0.868, reflecting high consistency between independent sequencing runs. The scatter plot shows minimal coverage variation, ensuring reliable data for downstream analysis.

[0273] The panel below displays a scatter plot comparing base level coverage between Nanopore and Illumina sequencing. While the plot indicates a correlation coefficient r = 0.8, this suggests a robust but slightly lower correlation compared to Nanopore repeats. In some specific implementations, this panel highlights the complementary nature of Nanopore and Illumina sequencing, where Nanopore's long-read capabilities can be used to provide additional insights into genomic regions that may be challenging for short-read techniques.

[0274] Now for specific reference Figure 15 It provides a comparison of copy number predictions for genes of interest (GOIs) derived from Nanopore and Illumina sequencing technologies. Figure 15 The specific implementation demonstrated the accuracy and reliability of Nanopore sequencing for copy number estimation.

[0275] The central feature of this graph is a bar chart representing the current Aptegra copy number predictions from Illumina sequencing. Light-colored bars represent the mean, while dashed lines above and below the bars represent the 95% confidence intervals (CIs). This visual representation provides a reference frame for comparing Nanopore predictions.

[0276] The figure shows two dots labeled "Nanopore Prediction 1" and "Nanopore Prediction 2". Both predictions fall within the 95% CI of Illumina's prediction-based approach, as indicated by the red dots within the dashed line. In some specific implementations, this location highlights the effectiveness of Nanopore sequencing as a reliable alternative to copy number determination, providing accuracy comparable to traditional Illumina methods.

[0277] Throughout the process, quality control checks can be performed to ensure data reliability. This includes verifying satisfactory alignment rates, assessing the uniformity of coverage distribution, and checking for potential contamination. In some specific implementations, these steps can be used to accurately characterize copy numbers, detect variants, and precisely identify integration sites.

[0278] Various alternatives and modifications can be devised by those skilled in the art without departing from this disclosure. Therefore, this disclosure is intended to cover all such alternatives, modifications, and variations. Furthermore, although several embodiments of this disclosure have been shown in the accompanying drawings and / or discussed herein, they are not intended to limit this disclosure, but rather to broaden its scope as permissible in the art, and the specification is read accordingly. Therefore, the above description should not be construed as restrictive, but merely as examples of particular embodiments. Moreover, those skilled in the art will contemplate other modifications within the spirit and scope of the appended claims. Other elements, steps, methods, and techniques substantially different from those described in the foregoing and / or appended claims are also intended to fall within the scope of this disclosure.

[0279] The embodiments shown in the accompanying drawings are merely illustrative of certain examples of this disclosure. Furthermore, the described drawings are purely illustrative and not restrictive. In the drawings, the sizes of some elements may be exaggerated and not drawn to scale for illustrative purposes. Additionally, elements with the same numbering in the drawings may be the same or similar elements, depending on the context.

[0280] When the term "comprising" is used in this specification and claims, it does not exclude other elements or steps. When an indefinite or definite article is used to refer to a singular noun, such as "an," "a," or "the," this includes the plural form of that noun unless otherwise specified. Therefore, the term "comprising" should not be construed as limited to the items listed thereafter; it does not exclude other elements or steps, and thus the scope of the expression "device comprising items A and B" should not be limited to a device consisting solely of components A and B. This expression indicates that, for the purposes of this disclosure, the only relevant components of the device are A and B.

[0281] Furthermore, the terms "first," "second," "third," etc., used in both the specification and claims, are intended to distinguish similar elements and do not necessarily imply a descriptive order or chronological sequence. It should be understood that such terms are interchangeable where appropriate (unless expressly stated otherwise), and the embodiments of this disclosure described herein can operate in orders and / or arrangements other than those described or illustrated herein.

Claims

1. A method for determining the copy number of a genomic region, the method comprising: Isolate genomic nucleic acid samples; Purify the genomic nucleic acid sample; A shotgun library was created from the genomic nucleic acid sample; The shotgun library was sequenced to generate reads; The processor is used to align the read with a reference sequence formed by the reference genome and vector sequence; The processor is used to divide the reference genome into multiple sizes; The processor is used to determine the average read coverage of each size of the divided reference genome until a predetermined number of consecutive average read coverage peaks are formed, and the consecutive average read coverage peaks have equal coverage distances between them. The processor is used to determine the corresponding copy number associated with each of the predetermined number of consecutive average read segment coverage peaks; and The processor is used to determine the model configured to estimate the copy number of the target sequence.

2. The method according to claim 1, wherein the reference sequence is further formed by at least one auxiliary sequence.

3. The method according to claim 2, wherein the at least one auxiliary sequence comprises an Escherichia coli genome sequence.

4. The method according to claim 1, wherein the target sequence is a sequence within the vector sequence.

5. The method according to claim 1, wherein the target sequence is the vector sequence.

6. The method of claim 1, wherein the target sequence is a predetermined gene within the vector sequence.

7. The method of claim 1, wherein the model is configured to estimate the copy number without requiring genomic position parameters.

8. The method according to claim 1, wherein the reference sequence is the DNA genome of a Chinese hamster ovary cell line.

9. The method of claim 1, wherein the reference sequence is the DNA genome of a mammalian cancer cell line.

10. The method of claim 1, wherein the shotgun library is prepared using an NGS shotgun library with a PCR-free workflow.

11. The method of claim 1, wherein the sequencing of the shotgun library uses sequencing-by-synthesis.

12. The method of claim 1, wherein the shotgun library is sequenced using Illumina sequencing.

13. The method of claim 1, wherein the sequencing of the shotgun library is performed using DNABSEQ sequencing.

14. The method of claim 1, wherein the plurality of partitioned reference genomes are from 100 bp to 51,200 bp in size.

15. The method of claim 1, wherein the plurality of sizes of the partitioned reference genome includes 50bp, 100bp, 200bp, 400bp, 800bp, 1600bp, 3200bp, 6400bp, 12800bp, 25600bp, and 51200bp.

16. The method of claim 15, wherein the processor uses sizes of 800bp, 1600bp, 3200bp, 6400bp, 12800bp, 25600bp, and 51200bp to estimate the copy number.

17. The method of claim 15, wherein the processor uses 50bp, 100bp, 200bp, 400bp, 800bp, 1600bp, 3200bp, 6400bp, and 12800bp to estimate the standard deviation of the copy number.

18. The method of claim 1, wherein all divisions of the divided reference genome are non-overlapping.

19. The method of claim 1, wherein at least two divisions of the divided reference genome overlap.

20. The method of claim 1, wherein all partitions of the partitioned reference genome are sequential partitions.

21. The method of claim 1, wherein the predetermined number of consecutive average read coverage peaks is three.

22. The method of claim 1, wherein the act of using the processor to determine the corresponding copy number associated with each of the predetermined number of consecutive average read coverage peaks comprises: The equal coverage distance is compared with a first value of a first peak in the predetermined number of consecutive average read coverage peaks to determine a first corresponding copy number of the first peak corresponding to the predetermined number of consecutive average read coverage peaks.

23. The method of claim 22, wherein the first corresponding copy number is one when the first value is approximately equal to the equal coverage distance.

24. The method of claim 22, wherein the second peak in the predetermined number of consecutive average read coverage peaks corresponds to a second corresponding copy number.

25. The method of claim 22, wherein when the first value is approximately twice the equal coverage distance, the first corresponding copy number is two.

26. The method of claim 25, wherein the second peak in the predetermined number of consecutive average read coverage peaks corresponds to a second corresponding copy number three.

27. The method of claim 1, wherein the model is a regression model that correlates coverage number with copy number.

28. The method of claim 1, wherein the partitioned reference genome corresponds to at least one gene.

29. The method of claim 1, further comprising normalizing the read coverage using a reference coverage value obtained from a control sample.

30. The method of claim 1, further comprising estimating the standard deviation of the estimated copy number based on the read coverage and the size of the target sequence.

31. The method of claim 30, further comprising determining a confidence interval for the estimated copy number based on the standard deviation and the estimated copy number.

32. The method of claim 1, wherein the alignment of the read segment with the reference sequence is performed using an accelerated version of the BWA alignment program.

33. The method of claim 1, wherein determining the average read coverage for each size of the partitioned reference genome comprises calculating the median coverage value for each size.

34. The method of claim 1, wherein the model is trained using the average base level coverage of each partition and the known copy number within the reference sequence.

35. The method according to claim 1 further includes performing quality control analysis on the read coverage to identify sequencing errors and filtering out false positive copy number determinations.

36. The method of claim 1, further comprising visualizing the estimated copy number and its confidence interval on a circos plot.

37. The method of claim 1, wherein the target sequence is one or more genes of interest (GOI) within the vector sequence.

38. The method of claim 37, further comprising determining the copy number of the one or more GOIs based on the observed read coverage of the one or more GOIs and the model.

39. The method of claim 1, wherein the vector sequence is an integrated full-size vector or a partial-size vector.

40. The method of claim 1, further comprising identifying and reporting any vector-vector tandem events based on split reads mapped to different portions of the vector sequence.

41. The method of claim 1, further comprising detecting and reporting sequence variants in the integrated vector by comparing the read with the vector sequence.

42. The method of claim 1, further comprising determining the location of at least one vector integration site in the reference genome by analyzing chimeric reads in the aligned reads mapped to the vector region and the paired reads in the aligned reads.

43. The method of claim 1, wherein the method is used for molecular characterization of Chinese hamster ovary (CHO) cells for the production of recombinant proteins.

44. The method of claim 1, wherein the method is used for molecular characterization of the reference genome.

45. The method of claim 1, wherein the method is used for molecular characterization of mammalian cancer cell lines.

46. ​​The method of claim 1, further comprising calculating the coverage variability of each size of the partitioned reference genome to estimate a reliability parameter of the estimated copy number.

47. The method of claim 1, wherein determining the average read coverage for each size of the partitioned reference genome comprises calculating the average base level coverage for each partition by taking into account the size and copy number of each fragment.

48. The method of claim 1, further comprising determining coverage anomalies.

49. The method of claim 1 further includes identifying homologous sequence regions within the vector sequence.

50. The method of claim 1, further comprising comparing the estimated copy number of the target sequence with a predetermined threshold to estimate copy number stability and integrity.

51. The method of claim 1, further comprising using molecular characterization selected from Southern blotting, ddPCR, and qPCR to verify accuracy.

52. The method of claim 1, wherein the reads are generated using long-read sequencing technology, thereby improving the accuracy of variant detection and integration site annotation.

53. The method of claim 1, further comprising detecting and reporting rearrangements or structural variations within the reference sequence by analyzing the alignment pattern of the read segments.

54. The method of claim 1, further comprising generating a molecular characterization report that summarizes the estimated copy number, variant profile, and integration site of the target sequence.

55. The method of claim 1, further comprising determining the ploidy distribution of the reference genome based on the observation coverage of fragments of different sizes, to refine the estimated copy number of the target sequence.

56. The method of claim 1, further comprising using at least one of Fish analysis, ddPCR analysis, and qPCR to verify the accuracy of the estimated copy number.

57. The method of claim 1 further comprises obtaining the coverage of different fragments of different sizes in the reference genome by extracting coverage information from the aligned reads, the sizes including 50, 100, 200, 400, 800, 1600, 3200, 6400 and 12800 bp.

58. The method of claim 1, further comprising normalizing coverage based on the maximum peak among the consecutive average read coverage peaks, by dividing the coverage of each point by the coverage of the maximum peak to provide a normalized coverage value.

59. The method of claim 1, further comprising normalizing the occurrence count by dividing the occurrence count of each point by the sum of the occurrence counts of all points to provide a normalized occurrence value.

60. The method of claim 1, further comprising: Extract the average read coverage of each of the plurality of known copy sizes; as well as Calculate the standard deviation for each size to assess the variability in fragment coverage.

61. The method of claim 1, further comprising determining a tandem event of the carrier sequence.

62. The method of claim 61, further comprising determining a connected read segment having a first segment of the vector sequence and a second segment of the vector sequence, wherein the first segment and the second segment are connected together within the connected read segment of the read segment.

63. The method of claim 1, further comprising identifying chimeric reads from the aligned reads, wherein the chimeric reads are partially mapped to the carrier sequence and partially mapped to a non-carrier region.

64. The method of claim 63, further comprising determining vector-vector linkages based on the chimeric reads, thereby determining the insertion location of the vector within the reference genome.

65. The method of claim 64, further comprising de novo assembly of the chimeric reads and their respective pairs to generate a longer contiguous group representing an extension of a continuous sequence.

66. The method of claim 65, further comprising aligning the assembled contiguous groups with the reference genome, the reference genome comprising both the host genome and the vector sequence, to identify the reference genome-vector boundary and create an alignment file.

67. The method of claim 1, wherein the shotgun library has an average fragment length of about 1,500 base pairs, and the sequencing of the shotgun library is performed using Nanopore sequencing technology.

68. The method of claim 67, wherein the genomic nucleic acid sample is isolated from a cloned cell line derived from the Chinese hamster ovary (CHO).

69. The method of claim 68, wherein the genomic nucleic acid sample comprises genomic DNA (gDNA) isolated from the CHO-derived clonal cell line.

70. The method of claim 69 further comprises fragmenting the genomic DNA into fragments ranging in size from about 0.3 to 2 kb.

71. The method of claim 70, further comprising converting the fragmented genomic DNA into a sequencing-compatible DNA library.

72. The method of claim 71, further comprising quantifying the sequencing-compatible DNA library using a Qubit fluorometer.

73. A system for determining the copy number, the system comprising: A processor configured to perform the method according to any one of claims 1-72.

74. An apparatus for determining a copy number, the determination of the copy number being implemented by a set of operable processor-executable instructions configured to perform one or more actions of the method according to any one of claims 1-72.

75. A computer-readable medium storing instructions that, when executed by a processor, cause the processor to perform the following operations: The processor is used to align the reads with a reference sequence formed from the reference genome and vector sequence; The processor is used to divide the reference genome into multiple sizes; The processor is used to determine the average read coverage of each size of the divided reference genome until a predetermined number of consecutive average read coverage peaks are formed, and the consecutive average read coverage peaks have equal coverage distances between them. The processor is used to determine the corresponding copy number associated with each of the predetermined number of consecutive average read segment coverage peaks; and The processor is used to determine the model configured to estimate the copy number of the target sequence.

76. The medium of claim 75, wherein the computer-readable medium is further configured to cause the processor to perform the method of any one of claims 2-72.

77. The medium according to claim 75 or 76, wherein the computer-readable medium is a non-transitory computer-readable medium.

78. The method of claim 1, further comprising the following actions: (i) Genomic DNA (gDNA) is extracted from mammalian cell lines to obtain said genomic nucleic acid samples; (ii) Tagging the gDNA sample with a mass in the range of 300 ng to 2000 ng to form a labeled gDNA fragment; (iii) The sample is first purified using sample preparation beads; (iv) Link the index to the labeled gDNA fragment from step (ii); as well as (v) The sample is purified a second time using sample preparation beads to obtain a library for NGS.

79. The method according to claim 78, wherein the mammalian cell line is the Chinese hamster ovary (CHO) cell line or the Homo sapiens cell line.

80. The method according to any one of claims 78 and 79, wherein the DNA is of high molecular weight.

81. The method according to any one of claims 78 to 80, wherein the number of mutations, insertions or deletions is reduced compared to the number of mutations, insertions or deletions occurring in a library preparation method using a polymerase chain reaction (PCR) amplification step.

82. The method according to any one of claims 78 to 81, further comprising adjusting the volume of the sample preparation beads during the first purification step or the second purification step to change the median library fragment size.

83. The method of claim 82, wherein the median library fragment size is greater than 350 base pairs (bp).

84. The method according to any one of claims 78 to 83, wherein the first purification step comprises a sample preparation bead with a volume selected from the range of 30 μl to 50 μl.

85. The method according to any one of claims 78 to 84, wherein the second purification step comprises a sample preparation bead with a volume selected from the range of 10 μl to 20 μl.

86. The method according to any one of claims 78 and 85, wherein the volume of the sample preparation beads in the first purification step is 33 μl, and wherein the volume of the sample preparation beads in the second purification step is 15 μl.

87. The method according to any one of claims 84 and 85, wherein the volume of the sample preparation beads in the first purification step is 45 μl, and wherein the volume of the sample preparation beads in the second purification step is 12 μl.

88. A method for reducing sequencing errors during high-throughput next-generation sequencing (NGS), the method comprising: (i) Perform high-throughput next-generation sequencing (NGS) on the library prepared by any one of the steps of claims 78 to 87.

Images