Method or device for detecting CNV and predicting maternal pollution

Abstract

The present invention relates to methods of determining the chromosomal methylation level of a sample when genetically detecting an embryo, and using a trained machine learning model to process a methylation level dataset of markers, to predict the proportion of maternal contamination in the sample, and optionally to correct the sample CNV. The invention further relates to a method, a device, equipment or a system for predicting the maternal pollution proportion in a sample and optionally correcting CNV of the sample, and is used for avoiding false positive or false negative of genetic detection results of embryos.

Description

Technical Field

[0001] The present invention relates to the field of genetic testing of embryos. Specifically, the present invention relates to a method or device for predicting whether there is maternal contamination and the proportion of maternal contamination in embryo-derived samples by using the difference in methylation levels between embryos and parents while detecting copy number variation (CNV) in embryo-derived samples, thereby avoiding false positives or false negatives in embryo CNV test results. Background Art

[0002] In the field of medical testing, "maternal contamination" refers to the mixing of tissues, cells or free DNA from the parents when testing samples from offspring embryos, which may interfere with the accuracy of genetic diagnosis of offspring embryos and the interpretation of results. This phenomenon usually occurs in scenarios such as prenatal diagnosis, abortion product testing, and preimplantation genetic testing (PGT). For example, when performing chromosome analysis on abortion product tissue to determine whether the embryo has genetic abnormalities, if parental cells are mixed in, the reliability of the test results will be affected. Similarly, if pathological or molecular biological analysis of the abortion product is required, maternal contamination will also significantly affect the accuracy of the results.

[0003] In order to exclude the influence of maternal contamination on the accuracy of embryo test results, detecting maternal contamination is a very important quality control indicator. In clinical practice, commonly used detection methods include short tandem repeat (STR) method and single nucleotide polymorphism (SNP) method. These methods use the difference in genetic information between parents and offspring to determine whether the offspring samples are affected by maternal contamination. The "Guidelines for the Application of Chromosome Microarray Analysis Technology in Prenatal Diagnosis (2023)" published in the "Chinese Journal of Obstetrics and Gynecology" emphasizes that a sound quality control system is the core of accurate diagnosis of chromosome microarray analysis (CMA) technology, and emphasizes the importance of determining the fetal nature of the specimen. The guideline points out that STR analysis of fetal samples is required before testing to exclude maternal contamination.

[0004] In 2019, the "Expert Consensus on the Application of Low-depth Whole-genome Sequencing Technology in Prenatal Diagnosis" published in the "Chinese Journal of Medical Genetics" (Liu Hongqian, Liu Juntao, Wu Lingqian, [J]. Chinese Journal of Medical Genetics, 2019, 36(4).) also proposed that in prenatal diagnosis, it is recommended to combine copy number variation sequencing (CNV-seq) technology with STR testing. STR testing can help determine the maternal contamination of the sample and ensure that the test results truly reflect the condition of the fetus.

[0005] However, there are also some shortcomings in using STR for maternal contamination detection. First, it is necessary to add additional operational steps on the basis of the original detection scheme, such as the need to implement both CMA and STR or the need to implement both CNV-seq and STR, which not only increases the cost of detection, but also increases the complexity of operation. Secondly, STR testing has certain requirements for the amount of DNA samples. In certain specific scenarios, such as PGT testing, there may be only a few embryonic cells or less free DNA from embryos, which will reduce the accuracy of STR testing and lead to misjudgment of maternal contamination.

[0006] Therefore, there is a need in the art to develop methods and devices that can accurately detect CNVs in embryo-derived samples even with trace amounts of cell-free DNA (cfDNA) present in spent embryo culture medium (SCM) and blastocoel fluid (BF), and predict whether there is maternal contamination in embryo-derived samples and the proportion of maternal contamination. Summary of the invention

[0007] The present invention provides a method and device for predicting whether there is maternal contamination and the proportion of maternal contamination in the embryo-derived sample while detecting CNV in the embryo-derived sample when performing genetic testing on the embryo, even if there is only a trace amount of embryonic DNA in the embryo-derived sample.

[0008] In a first aspect, the present invention provides a method for determining the chromosome methylation level of a test sample, the method comprising:

[0009] a. Treat the test sample with a methylation-sensitive restriction endonuclease.

[0010] b. performing whole genome amplification on the DNA of the test sample treated with methylation-sensitive restriction endonuclease to construct a sequencing library;

[0011] c. Implement high-throughput sequencing to obtain raw reads data;

[0012] d. Align the raw reads data from step c to the human genome reference sequence version such as NCBI Build37 / hg19, hg38 using alignment software, such as BWA software;

[0013] e. performing de-repeating sequence analysis on the alignment result of step d to remove duplicate reads, for example, using Sambamba software to perform de-repeating sequence analysis on the alignment result of step d, and after the analysis, the obtained read sequence set is used as a set of unique matching sequences;

[0014] f. Extracting reads that overlap with the CCGG sequence obtained from the human reference genome and meet the conditions from the read sequence set obtained in step e, and obtaining a read sequence set that overlaps with the ccGG sequence of the human reference genome, for example, the conditions to be met are that the mapq of the reads is between 20-60, for example, between 30-60, between 35-60, between 40-60 or between 50-60, for example, the mapq is greater than 30 and less than 60; and the NM value is between 0-2, for example, the NM value is less than or equal to 1;

[0015] g. Divide each human chromosome into N regional segments, where each regional segment is a window, also called a "bin". For example, divide each human chromosome according to a window size of 5Mb-50Mb (for example, a window size of 5Mb, 10Mb, 15Mb, 20Mb, 25Mb, 30Mb, 35Mb, 40Mb, 45Mb or 50Mb, etc.), count the number of reads overlapping the window in the reads sequence set obtained in step f that overlaps with the human reference genome ccgg sequence, marked as bin_reads_count, and count the number of reads containing CCGG sequences from these overlapping reads, marked as bin_ccgg_reads_count, and then calculate the CCGG ratio of each window, and the calculation formula is:

[0016] CCGG ratio=bin_ccgg_reads_count / bin_reads_count (Formula 1)

[0017] Thus, the methylation level of each window on the chromosome of the test sample is determined, and a data set of the methylation level on the chromosome of the test sample is obtained.

[0018] In some embodiments, the methylation-sensitive restriction endonuclease used in the method for determining the chromosome methylation level of a test sample of the present invention does not cut the DNA methylation region. For example, the methylation-sensitive restriction endonuclease is selected from HpalI, SmaI, MspI, XmaI and BstUI, etc.

[0019] In some embodiments, the rawreads data in the method of determining the chromosome methylation level of a test sample of the present invention is obtained using a second-generation sequencing platform selected from, for example, Illumina, Life Technologies, and BGI.

[0020] In some embodiments, the test sample used in the method of determining the chromosome methylation level of the test sample of the present invention is an offspring or embryonic sample, for example, selected from: embryonic tissue (e.g., abortion product tissue), maternal amniotic fluid, maternal peripheral blood, umbilical cord blood, embryo biopsy cells, trace embryonic cells, and free DNA in embryo culture medium.

[0021] In a second aspect, the present invention provides a method for establishing a model for predicting maternal source pollution, the method being implemented on a computer system comprising one or more processors and a system memory, the method comprising:

[0022] a. providing a training set consisting of a plurality of (e.g., 2-400, 5-200, 10-150, 20-120, 30-100) test samples on a computer system, and determining the maternal contamination ratio of the plurality of test samples in the training set according to the relative allele frequency (e.g., B allele frequency);

[0023] b. implementing the method described in the first aspect of the present invention on the plurality of test samples in the training set to obtain the methylation level of each window on the chromosome of the test sample;

[0024] c. Performing a correlation analysis (e.g., Pearson correlation analysis, Spearman rank correlation analysis, Kendall rank correlation analysis) on the methylation level of each window on the chromosome of the test sample in step b and the maternal contamination ratio determined in step a, and obtaining the window of the difference in methylation level between the embryo and the parent as a marker;

[0025] d. training a model for predicting maternal contamination using the training set of step a for the markers of step c, for example, and evaluating the performance of the model for predicting maternal contamination;

[0026] e. Use an independent validation set consisting of multiple additional test samples (e.g., 2-400, 5-200, 10-150, 20-120, 30-100) to validate the model for predicting maternal contamination in step d to obtain performance characteristics of the model for predicting maternal contamination, for example, a false negative rate (FNR) of 0, an accuracy of 0.85, and a recall of 1.

[0027] In the third aspect, the present invention provides the use of the window of difference in methylation levels between embryos and parents on chromosomes obtained by the present invention, for generating a prediction of the proportion of maternal contamination by applying the model for predicting maternal contamination established in the second aspect of the present invention to a dataset of methylation levels in the window of difference in methylation levels between embryos and parents on chromosomes.

[0028] In a fourth aspect, the present invention provides a method for predicting maternal contamination of a sample, comprising:

[0029] a. implementing the method of the first aspect of the present invention on the sample, determining the methylation level of the window on the chromosome where the methylation level of the embryo is different from that of the parent, and obtaining a data set of the methylation level of the window on the chromosome where the methylation level of the embryo is different from that of the parent;

[0030] b. generating a prediction of the proportion of maternal contamination by applying the model for predicting maternal contamination established in the second aspect of the present invention to the data set of step a,

[0031] Thus, the maternal contamination of the sample is predicted.

[0032] In a fifth aspect, the present invention provides a method for correcting maternal contamination of CNV of a sample, comprising:

[0033] a. implementing the method of the fourth aspect of the present invention on the sample to obtain a predicted maternal contamination ratio of the sample,

[0034] b. Use the sample maternal contamination ratio in step a to correct the CNV of the sample.

[0035] In a sixth aspect, the present invention provides an apparatus, device or system for predicting the proportion of maternal contamination of a sample, comprising:

[0036] A data receiving unit, configured to receive a dataset of methylation levels of markers of a sample, wherein the dataset is, for example, a dataset of methylation levels of 2-400, 5-200, 10-150, 20-120, or 30-100 markers;

[0037] At least one processor and at least one memory, wherein the at least one memory stores thereon a trained machine learning model of the present invention, and when executed by the at least one processor, the trained machine learning model processes the methylation level dataset of the marker to generate the maternal contamination ratio of the sample.

[0038] In a seventh aspect, the present invention provides an apparatus, device or system for predicting the proportion of maternal contamination in a sample and correcting CNV in a sample, comprising:

[0039] A data receiving unit, configured to receive CNV data of a sample and a dataset of methylation levels of markers of the sample, wherein the dataset is, for example, a dataset of methylation levels of 2-400, 5-200, 10-150, 20-120, or 30-100 markers;

[0040] At least one processor and at least one memory, wherein the at least one memory stores thereon a trained machine learning model of the present invention and a computer program for correcting sample CNV. When executed by the at least one processor, the trained machine learning model processes the methylation level dataset of the marker to generate the maternal contamination ratio of the sample; and processes the CNV data of the sample by the computer program for correcting sample CNV to generate a corrected sample CNV, and outputs the CNV correction result of the sample.

[0041] The device, apparatus or system of the present invention can be used to predict the proportion of maternal contamination in a test sample, and optionally correct the CNV of the sample. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Combined with the following Figure 1 When reading together, the preferred embodiments of the present invention described in detail below will be better understood. For the purpose of illustrating the present invention, the drawings show the currently preferred embodiments. However, it should be understood that the present invention is not limited to the precise arrangement and means of the embodiments shown in the drawings.

[0043] Figure 1 A flowchart showing the operation when using fetal tissue, umbilical cord blood, or maternal amniotic fluid containing a large amount of fetal DNA as a sample.

[0044] Figure 2 The flowchart shows an operation when one or more human embryo-derived cells or embryo culture medium containing a trace amount of embryonic DNA are used as samples.

[0045] Figure 3 The principle of analyzing the chromosome methylation level of a sample using the methylation-sensitive restriction endonuclease Hpa II is shown.

[0046] Figure 4 This shows a qualitative ROC curve for a random forest model.

[0047] Figure 5 The CNV diagram of sample 1 in Example 4 before and after correction for maternal contamination is shown. DETAILED DESCRIPTION

[0048] Before describing the present invention in detail, it should be understood that the present invention is not limited to the specific methods and experimental conditions in this specification, because the methods and conditions can be changed. In addition, the terminology used herein is only for the purpose of describing specific embodiments and is not intended to be limiting.

[0049] I. Definitions

[0050] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art. For the purposes of the present invention, the following terms are defined below.

[0051] The term "about" when used in conjunction with a numerical value is meant to encompass the numerical value within a range having a lower limit that is 10% less than the specified numerical value and an upper limit that is 10% greater than the specified numerical value.

[0052] The term “and / or”, when used to connect two or more optional items, should be understood to mean any one of the optional items or any two or more of the optional items.

[0053] As used herein, the term "comprising" or "including" means including the described elements, integers or steps, but does not exclude any other elements, integers or steps. In this article, when the term "comprising" or "including" is used, unless otherwise specified, the situation consisting of the described elements, integers or steps is also covered.

[0054] "Preimplantation genetic testing" (PGT) is a genetic test performed on embryos based on the first or second generation of fertilization. It aims to select healthy embryos for transplantation to avoid the birth of children with genetic defects and improve the success rate of in vitro fertilization. PGT can be further divided into chromosome aneuploidy testing (PGT-A), single gene genetic disease testing (PGT-M) and chromosome structural abnormality testing (PGT-SR), which have slight differences in their indications for testing.

[0055] Preimplantation Genetic Testing for Aneuploidies (PGT-A) confirms whether the embryo is normal by detecting the number of chromosomes in the embryo. If the number of chromosomes in the embryo is incorrect, it is an aneuploid embryo; if the number of chromosomes in the embryo is correct, it is a euploid embryo.

[0056] Aneuploidy refers to the phenomenon that the number of chromosomes in the body deviates from the normal value. A normal human body has 23 pairs, a total of 46 chromosomes, including 22 pairs of autosomes and 1 pair of sex chromosomes. Aneuploidy is mainly divided into two categories: monosomy and trisomy. Monosomy patients are missing a chromosome. Therefore, for a certain chromosome, there is only one in the human body instead of the normal two. Trisomy patients have an extra copy of a chromosome in their bodies. Chromosomal aneuploidy can occur in any pair or several pairs of chromosomes and cause the carrier to die directly during the embryonic or fetal period. However, a few of them have the possibility of live birth, such as Down syndrome (trisomy 21), Edwards syndrome (trisomy 18), Patau syndrome (trisomy 13), Klinefelter syndrome (XXY syndrome), Turner syndrome (monosome X syndrome), XXX syndrome, etc.

[0057] When one or several fragments in a chromosome are aneuploid, this abnormal number is called copy number variation (CNV), which usually refers to deletion or duplication. It generally refers to the insertion or deletion variation of a genomic sequence with a length of 100bp to 10Mb. For example, the alleles of a gene / base genome region on a pair of chromosomes appear in pairs, and each allele is a copy, so the copy number of alleles in a normal person is 2. Methods for detecting CNV include, for example: methods based on polymerase chain reaction (such as timed quantitative polymerase chain reaction, multiplexed probe amplification technology, multiplexed amplifiable probe hybridization technology, etc.), methods based on hybridization technology (such as in situ immunofluorescence, Giemsa banding, Southern blotting technology, comparative genomic hybridization technology, whole genome scanning methods, etc.), methods based on chip technology (such as chip comparative genomic hybridization technology, single nucleotide polymorphism chip, etc.), and methods based on high-throughput sequencing technology (such as CNV-seq).

[0058] Preimplantation Genetic Testing for Monogenic (PGT-M) is used to prevent the occurrence of monogenic genetic diseases. PGT-M can detect whether the embryo has known pathogenic gene mutations or disease-related mutations, thereby selecting normal embryos for transplantation, thereby avoiding the occurrence of monogenic genetic diseases in the next generation.

[0059] Preimplantation Genetic Testing for Chromosomal Structural Rearrangements (PGT-SR) is mainly used for carriers of chromosomal structural rearrangements, such as patients with balanced translocation or Robertsonian translocation. Chromosomal rearrangements can be inherited or spontaneous.

[0060] In the present invention, the sample to be tested is a sample of offspring or embryo, including but not limited to: embryonic tissue (e.g., abortion product tissue), maternal amniotic fluid, maternal peripheral blood, umbilical cord blood, embryo biopsy cells, trace embryonic cells, free DNA in embryo culture fluid, etc. The term "offspring or embryo trace DNA sample" or "trace DNA sample" can be used interchangeably herein to refer to a sample containing DNA from offspring or embryos, wherein the DNA from offspring or embryos is present in trace amounts in the sample, for example, less than 1ng, 500pg, 100pg or 10pg or even less (e.g., about 1-6pg, e.g., 5pg). In this article, "DNA from offspring or embryos" or "offspring or embryo-derived DNA" can be used interchangeably to refer to DNA from, for example, offspring or embryo cells, offspring or embryo body fluids, or offspring or embryo cell culture fluids, whose genotype is substantially equivalent to the offspring or embryo genome genotype. Therefore, in some embodiments, the trace DNA sample can be an embryo produced by in vitro fertilization (IVF) cultured in a culture medium, and there is a sample of trace embryonic genomic DNA released by the cultured embryo in the culture medium. In some embodiments, the sample is an embryo or fetal biopsy cell sample, for example, an IVF blastocyst trophoblast cell. In some embodiments, the sample is blood and / or plasma from a pregnant woman, which contains a trace amount of cell-free DNA (cfDNA) from the fetus.

[0061] The term "offspring" as used herein includes, but is not limited to, mammalian, such as human, offspring, and refers to born or unborn offspring.

[0062] The term "blastocyst" is an embryo 5 or 6 days after fertilization that has an inner cell mass, an outer cell layer called the trophectoderm, and a fluid-filled blastocoel cavity that contains the inner cell mass from which the entire embryo was derived. The trophectoderm is the precursor to the placenta.

[0063] Maternal contamination refers to the maternal contamination caused by the introduction of some parental (i.e., paternal and / or maternal) components into the embryo sample during prenatal diagnosis. "Maternal contamination" is one of the main factors affecting the accuracy of embryo chromosome detection results in embryo chromosome detection, and tends to increase the detection failure rate, false positive and false negative probability of the sample. In some embodiments, the maternal contamination comes from the cumulus cells directly surrounding the egg.

[0064] The term "epigenetic modification" refers to the modification of genomic function, which affects gene expression by changing gene activity through a series of biological modifications rather than the nucleotide sequence of DNA. The modification of genomic function mainly includes modifications to DNA, RNA, and histones, which change the local electrochemical properties and conformation of chromatin, thereby regulating the transcriptional activity of genes. DNA methylation is an important research direction in the field of epigenetics. The most common DNA methylation modification in eukaryotes is 5-methylcytosine (5mC). There are about 30 million CpG sites in the human reference genome, of which about 70-80% are methylated.

[0065] Chromosome microarray analysis (CMA) technology, also known as chromosome chip. According to the different chip designs and detection principles, CMA technology can be divided into array-based comparative genomic hybridization (aCGH) technology and single nucleotide polymorphism (SNP) microarray technology, which can simultaneously detect almost all microscopic and submicroscopic CNVs in the genome. aCGH is a chip technology developed on the basis of comparative genomic hybridization technology. Its principle is to use different colors of fluorescence (such as green Cy3 fluorescence and red Cy5 fluorescence) to label the test sample and the whole genome DNA of the healthy person reference sample in equal proportions, and then competitively hybridize with the microarray sequence fixed on the carrier. The two fluorescence intensities of the test sample and the reference sample are compared by computer software analysis. The deletion area in the test sample will produce a stronger Cy5 signal (red), while the duplication area will produce a stronger Cy3 signal (green), thereby determining the copy number changes at the whole genome level of the test sample. Unlike aCGH technology, SNP microarray technology only needs to label the genomic DNA of the sample to be tested and hybridize with the SNP microarray. The A and B alleles of a given SNP are labeled with different fluorescent dyes, the absolute fluorescence intensity of the sample to be tested is measured, and the copy number change is determined by comparing it with the fluorescence intensity of the normal reference sample database (computer simulation).

[0066] The term "short tandem repeat (STR)" is composed of 2-7 base pairs as the core unit, which are repeated in tandem and are commonly found in introns of non-coding regions. Due to the different repeating units and the number of repetitions, the distribution of STR in different races and populations is very different, which constitutes the genetic polymorphism of STR. The typing combination of an individual at multiple STR loci constitutes the unique genetic profile of this individual. Selecting several STR loci can be used as effective individual identification markers with individual specificity. Quantitative fluorescence PCR (QF-PCR) is based on PCR amplification and capillary electrophoresis separation technology, which can qualitatively and quantitatively analyze the polymorphism of STR and is one of the commonly used technical means to detect STR loci.

[0067] The term genome "copy number variation sequencing (CNV-seq)" is based on high-throughput sequencing technology. It uses the principle of sequencing by synthesis to perform low-depth whole-genome level detection on sample DNA, compares the sequencing results with the human reference genome, and performs bioinformatics analysis to discover aneuploidy and CNVs of the tested samples. The accuracy of CNV detection is higher than CMA.

[0068] High-throughput sequencing is also known as Next-Generation Sequencing (NGS) and Second-Generation Sequencing. Compared with the first-generation Sanger sequencing method, NGS technology has many obvious advantages, such as the sequencing library used by NGS can be constructed and amplified in vitro, while the construction and amplification of the library of the first-generation sequencing technology still depends on the transformation and reproduction of E. coli; NGS technology has extremely high throughput, and can obtain up to hundreds of G of data in one run, while the amount of data obtained by the first-generation sequencing technology in one run is less than 1M; the sequencing cost of NGS technology is only one hundred thousandth of the traditional sequencing cost, and this ratio is constantly decreasing.

[0069] The term "sequencing depth" refers to the ratio of the total number of bases obtained by sequencing to the size of the genome to be tested. Assuming a genome size of 2M and a sequencing depth of 10X, the total amount of data obtained is 20M. Sequencing depth can be expressed as the ratio of the total number of bases (bp) obtained by sequencing to the size of the genome (Genome).

[0070] The term "coverage" refers to the ratio of the sequence portion of a genome, transcriptome or chromosome segment for which sequence information is known to the entire group or segment.

[0071] Allele frequency or gene frequency is the frequency of the allele of the gene (or variant of the gene) relative to other alleles of the gene, which can be expressed as a decimal or percentage.Allele frequency is usually associated with a specific genomic locus, because the gene is usually located at one or more loci.However, the allele frequency as used herein can also be associated with the fragment (size-based bin) based on size of DNA fragmentation.In this sense, the DNA fragmentation (such as cfDNA) comprising the allele is assigned to different fragments based on size.Relative to the frequency of other alleles, the frequency of the allele in the fragment based on size is the allele frequency.

[0072] B allele frequency (BAF) is used to reflect the heterozygosity of chromosome segments. When the copy number is normal, the B allele has three frequencies, which can reflect three banding types (genotype AA = 0, genotype AB = 0.5, genotype BB = 1); when the copy number increases, the banding type may change accordingly, for example, when the copy number is 3, four banding types appear (AAA = 0, AAB = 0.33, ABB = 0.67, BBB = 1). Therefore, BAF values ​​0, 0.5 and 1 represent normal copy numbers (n = 2); when the copy number increases, BAF values ​​are around 0, 0.33, 0.67 and 1.0 (genotypes AAA, AAB, ABB and BBB). Based on this regular change of BAF, the BAF value can be used to estimate abnormal copy number variation and the proportion of exogenous contamination DNA in genomic DNA (gDNA) samples.

[0073] The term "read" refers to a DNA sequence output by a sequencer, a sequence obtained from a portion of a nucleic acid sample. Each sequence in the sequencing data is a read. The read can be symbolically represented by a base pair sequence (represented by A, T, C, or G) of the sample portion. The read can be stored in a memory device and appropriately processed to determine whether it matches a reference sequence or meets other criteria. In some cases, a read is a DNA sequence of sufficient length (e.g., at least about 25 bp) that can be used to identify, for example, a larger sequence or region that can be aligned and specifically assigned to a chromosome or genomic region or gene.

[0074] The term "alignment" refers to the process of comparing a read to a reference sequence and thereby determining whether the reference sequence contains the read sequence. If the reference sequence contains the read, the read sequence can be mapped to the reference sequence, or in some embodiments, to a specific position in the reference sequence. In some cases, the alignment is used to indicate whether the read is a member of a particular reference sequence (i.e., whether the read is present or absent in the reference sequence). In some cases, the alignment is used to indicate the position in the reference sequence to which the read is mapped.

[0075] The term "mapping" refers to the use of alignment to specifically assign sequence reads to a larger sequence (eg, a reference genome) or a subsequence of the larger sequence.

[0076] The term "reference genome" or "reference sequence" refers to any specific known genomic sequence, either partial or complete, of any organism or virus that can be used to reference an identified sequence from a subject. For example, reference genomes for human subjects and many other organisms can be found at ncbi.nlm.nih.gov. For example, the commonly used human reference genomes are hg19 / GRCh37 hg38 / GRCh38 and other versions.

[0077] The term "parameter" herein refers to a numerical value that characterizes an attribute of a system such as a physical feature, the value of which or other characteristics may affect relevant conditions such as a sample or DNA molecule. In some cases, the term "parameter" is used with reference to a variable that affects a mathematical relationship or model output, which may be an independent variable (i.e., an input to the model) or an intermediate variable based on one or more independent variables. Depending on the scope of the model, the output of one model may become the input to another model, thereby becoming a parameter of another model.

[0078] The term "MAPQ" is an abbreviation for mapping quality, which is a quality score for the alignment, used to describe the accuracy and reliability of mapping the sequenced reads to the reference genome. The higher the MAPQ value, the more reliable the read alignment. A MAPQ value of 60 (which is the maximum MAPQ value in some software) means that the software considers the alignment to be very reliable.

[0079] The Wilcoxon rank-sum test, also known as the Mann-Whitney U test, is a nonparametric test used to compare whether the medians of two independent samples are significantly different.

[0080] A CNV breakpoint is a location on a chromosome where a copy number variation (eg, gain or loss) occurs.

[0081] The term "differentially methylated region" refers to a genomic region that has a differential methylation status between the fetus and the parent. In the present specification, Table 4 shows the differentially methylated regions identified using the HpaII enzyme.

[0082] II. Determine the methylation level of each window on the chromosome of the test sample

[0083] DNA methylation is an important epigenetic modification that involves the addition of a methyl group (-CH 3 ). The most common form of methylation occurs at CpG dinucleotides, i.e., a sequence of cytosine followed by guanine. In the human genome, approximately 70-80% of CpG sites are associated with methylation. DNA methylation has significant characteristics in stages such as germ cell formation and embryonic development, and plays an important biological function. Through in-depth research, the inventors have developed a method for quickly and inexpensively obtaining the methylation level of a chromosome region (also called a "chromosome window") of a sample, and using the methylation level to determine whether there is maternal contamination.

[0084] The method of identifying whether a fetal DNA sample has maternal contamination is as follows: Figure 1 and Figure 2 As shown. The method of the present invention can detect DNA samples of pg level to microgram level (for example, 1pg to 1μg, for example, 5pg to 100ng). The DNA sample is treated with methylation-sensitive DNA restriction endonucleases (including but not limited to HpaII, SmaI, MspI, XmaI and BstUI, etc.), and then the sequencing library is prepared and high-throughput sequencing is performed. The obtained sequencing data can be analyzed for methylation levels based on the proportion of recognition sites of methylation-sensitive DNA restriction endonucleases in the sequencing data, thereby obtaining information on maternal contamination.

[0085] In the method of the present invention, the methylation-sensitive DNA restriction endonuclease has the characteristic of not cutting the DNA methylation region. Commonly used methylation-sensitive restriction endonucleases are, for example, Hpa II, Msp I (recognition sequence CCGG, but when the cytosine in the sequence is methylated, CCGG cannot be cut); Sma I, XmaI (recognition sequence CCCGGG), etc.

[0086] In some embodiments, when embryonic tissue, umbilical cord blood, or maternal amniotic fluid containing a large amount of embryonic DNA is used as a test sample, the method of the present invention for determining the methylation level of each window on the chromosome of the test sample includes:

[0087] - Extract DNA from the test sample;

[0088] - Treat DNA with methylation-sensitive restriction endonucleases;

[0089] -Perform whole genome amplification on DNA treated with methylation-sensitive restriction endonucleases and construct sequencing libraries;

[0090] -Perform high-throughput sequencing and data analysis, including obtaining the methylation level of each window on the chromosome of the test sample based on the proportion of recognition sites of methylation-sensitive DNA restriction endonucleases in the sequencing data.

[0091] In some embodiments, when one or several trace embryonic cells containing trace embryonic DNA or free DNA in embryo culture medium is used as a test sample, the method of the present invention for determining the methylation level of each window on the chromosome of the test sample comprises:

[0092] - Treating the test sample with a cell lytic enzyme (e.g., proteinase K);

[0093] - Then, treat with methylation-sensitive restriction endonucleases such as Hpa II, Msp I, Sma I or XmaI;

[0094] - Whole genome amplification and construction of sequencing library;

[0095] -Perform high-throughput sequencing and data analysis, including obtaining the methylation level of each window on the chromosome of the test sample based on the proportion of recognition sites of methylation-sensitive DNA restriction endonucleases in the sequencing data.

[0096] In some embodiments, the methylation level analysis is a larger-scale differential methylation block (DMB) analysis.

[0097] In some embodiments, high-throughput sequencing is performed and calculation of methylation levels is performed as follows.

[0098] Use a high-throughput sequencer (including but not limited to various sequencing platforms of brands such as MGI, GeneMind, ThermoFisher, Illumina, etc.) to sequence the library. The sequencing type can be selected: SE55, SE100, PE100 or PE150. The amount of sequencing data should not be less than 200K reads. The sequencing data quality requirements are Q30>70% and Q20>85%.

[0099] i) sequencing the DNA sample to generate raw reads data, and storing the data in a fastq file, for example, using a second-generation sequencing platform to generate raw reads data, wherein the second-generation sequencing platform is selected from, for example, second-generation sequencing platforms of Illumina, Life Technologies, BGI, Zhenmai, and Cellu;

[0100] ii) aligning the raw reads data in the fastq file of step i) to the human genome reference sequence to obtain an alignment result (bam file), wherein the human genome reference sequence is, for example, the human genome reference sequence version: NCBIBuild37 / hg19, hg38; for example, aligning the raw reads data of step i) to the human genome reference sequence by using the alignment software BWA (Burrows-Wheeler Alignment tool);

[0101] iii) performing markup on the alignment result (bam file) of step ii) to remove duplicate reads and obtain the alignment result (bam file) after markup, such as using sambamba software to remove duplicates. After the analysis, the obtained sequence set is used as a set of unique matching sequences, also known as "effective data volume";

[0102] iv) Obtain CCGG sequences from the human reference genome. Specifically, for each chromosome, starting from the first base, search for subsequent base sequences CCGG sequences in steps of 1, and store the genomic position of the CCGG sequence as a bed file. The bed file includes three columns of information, namely chromosome number (chrom), starting position (start), and ending position (end), where each row represents the position of a CCGG sequence.

[0103] v) Extracting the reads that overlap with those in step iv) and meet the conditions from the markdup bam file in step iii) and storing them in a new BAM file (ccgg bam), wherein the conditions met include, for example, mapq of the read is greater than 30 (for screening reads with high credibility) and NM value of the read is less than or equal to 1 (for screening reads with low sequencing error rate), wherein the NM value is the edit distance between the base of the read and the base of the reference genome, and the smaller the NM value, the smaller the difference with the reference genome and the lower the sequencing error rate.

[0104] vi) Divide each human chromosome into N regional segments, where each regional segment is a window (also called "bin"). For example, divide each human chromosome into 5Mb-50Mb window sizes (e.g., 10Mb, 15Mb, 20Mb, etc., where 1Mb = 1,000,000 bases), count the number of reads in each window where ccgg bam overlaps with the window (i.e., the start position of the read falls within the bin), marked as bin_reads_count, count the number of reads containing CCGG sequences from these overlapping reads, marked as bin_ccgg_reads_count, and then calculate the CCGG ratio of each window using the following formula:

[0105] CCGG ratio=bin_ccgg_reads_count / bin_reads_count (formula 1)

[0106] III. Prediction of the proportion of maternal source pollution

[0107] The present invention establishes a model for predicting maternal source pollution, which is a machine learning model.

[0108] The present invention provides a method for training a machine learning model to predict the proportion of a sample contaminated by its parents, the method comprising the following steps.

[0109] i) Load the dataset and split it into a training set and a separate test set.

[0110] A number of samples (e.g., 2-400, 5-200, 10-150, 20-120, 30-100) are randomly selected as training sets, and a number of samples (e.g., 2-400, 5-200, 10-150, 20-120, 30-100) are randomly selected as independent validation sets.

[0111] Based on a method for identifying the parental tendency or parental contamination of a nucleic acid sample proposed by the inventor, the parental tendency or parental contamination (the parental contamination and maternal contamination are used interchangeably in this article) ratio of the training set sample and the independent verification set sample is measured, and the method is: for a genomic DNA (gDNA) sample, by the relative allele frequency, the BAF (Ballele frequency) determined based on the nucleic acid chip identifies whether the nucleic acid sample has parental contamination (see patent application publication number: CN114214425A). The nucleic acid chip can be a single nucleotide polymorphism site microarray nucleic acid chip, a MassARRAY flight mass spectrometer chip or a combination thereof. In one embodiment, the single nucleotide polymorphism site microarray nucleic acid chip is a SNP genotyping chip.

[0112] ii) The training set in i) is subjected to correlation analysis (eg, Pearson correlation analysis) and Wilcoxon rank sum test to screen for regions where there are significant differences in methylation levels between the embryo and the parents.

[0113] Specifically, the methylation level of each window on the chromosome of each sample in the training set is determined by the method under the title "II. Determining the methylation level of each window on the chromosome of the test sample" of this article; the methylation level of each window on the chromosome of each sample in the training set is subjected to correlation analysis with the maternal contamination ratio determined by the relative frequency of alleles under i). In some embodiments, a Pearson correlation analysis is performed to obtain the Pearson correlation coefficient, and further the multiple hypothesis test of the Bonferroni, Benjamini-Hochberg, and Holm methods is used to calculate the difference significance P value of each window. The P values ​​of the multiple hypothesis tests of each window are then sorted to obtain candidate windows with significant differences in the methylation levels of the embryo and the parent. In the present invention. The windows with significant differences in the methylation levels of the embryo and the parent are also referred to as "markers" or "markers".

[0114] In some embodiments, the methylation level of each window on the chromosome of each sample in the training set is determined by the method under the title "II. Determining the methylation level of each window on the chromosome of the test sample"; each sample in the training set is divided into a case group and a control group according to different thresholds (such as 30%, 40%, 50%, etc.) set according to the maternal contamination ratio determined by the relative frequency of alleles under i). In some embodiments, when the threshold set for the maternal contamination ratio determined by the relative frequency of alleles under i) of "III. Predicting the maternal contamination ratio" is 30%, samples greater than or equal to 30% of the samples in the training set are case groups, and samples less than 30% are control groups. In some embodiments, when the threshold set for the maternal contamination ratio determined by the relative frequency of alleles under i) is 40%, samples greater than or equal to 40% of the samples in the training set are case groups, and samples less than 40% are control groups. In some embodiments, when the threshold value set for the maternal contamination ratio determined by the relative frequency of alleles under i) is 50%, samples greater than or equal to 50% of the samples in each sample of the training set are case groups, and samples less than 50% are control groups. After the maternal contamination ratio is grouped, the CCGG ratio is subjected to a Wilcoxon rank sum test; for example, in method B of the embodiment, with 30% as the cutoff, the case group has 46 samples, and the control group has 29 samples, that is, there are 46 CCGG ratios in window 1 of the case group, and there are 29 CCGG ratios in window 1 of the control group, and then the two groups of data are subjected to a rank sum test analysis to obtain the p value of the difference between the two groups, and further use the Bonferroni, Benjamini-Hochberg, Holm method Multiple hypothesis test to calculate the difference significance P value of each window. Then the P value of the multiple hypothesis test of each window is sorted to obtain candidate windows with significant differences in the methylation levels of the embryo and the parent. In the present invention. The window where there is a significant difference in methylation levels between the embryo and the parents is also called a "marker".

[0115] iii) Use the training set to train the model using the candidate markers selected in ii).

[0116] In one embodiment, when training the model, the method used is Random Forest, which is a machine learning algorithm widely used for classification and regression problems. The model is an integrated model composed of multiple decision trees, and the generalization ability and prediction accuracy of the model are improved by introducing randomness.

[0117] The evaluation method used to evaluate the performance of the model is the leave-one-out cross-validation method. The "leave-one-out cross-validation" (LOOCV) is a cross-validation method used to evaluate the performance of the model. In the leave-one-out method, one sample in the data set is used as the test set each time, and the remaining samples are used as the training set, and the test is repeated until all samples are tested. The performance is evaluated based on the prediction results returned by the training model (i.e., the probability value of predicting the sample as maternal contamination and no maternal contamination).

[0118] Evaluation metrics include:

[0119] Specificity = TN / (TN+FP)

[0120] Sensitivity = TP / (TP+FN)

[0121] Positive predictive value PPV = TP / (TP+FP)

[0122] Negative predictive value NPV = TN / (TN+FN)

[0123] False positive rate FPR = FP / (FP+TN)

[0124] False negative rate FNR = FN / (FN+TP)

[0125] Accuracy = (TP+TN) / sum(TP+TN+FP+FN)

[0126] Precision = TP / (TP+FP)

[0127] Recall = TP / (TP+FN)

[0128] (Note: TN: true negative; FP: false positive; TP: true positive; FN: false negative)

[0129] By using all the candidate makers obtained in ii) as features for auxiliary model training, the aim is to obtain a model with optimal performance, which should have all the candidate markers. A model is constructed based on all the candidate markers.

[0130] iv) Using the independent validation set in i) to validate the model obtained in iii) and evaluating the performance indicators of the independent validation set using the model performance evaluation method described in iii).

[0131] Specifically, all markers described in iii) are extracted from the independent validation set samples, and the model's maternal contamination prediction is verified using the data from the independent validation set. The prediction method is the predict function in the R language. In the R language, the predict function is used to generate prediction values ​​based on the fitted model. The predict function can be used for various types of model objects, including linear regression, generalized linear models, decision trees, random forests, etc. The basic usage of the predict function is to pass a trained model object and new data to it to obtain prediction results. Finally, the performance of the model is evaluated based on the prediction results.

[0132] In some embodiments, the model for predicting maternal contamination is verified to obtain performance characteristics of the model for predicting maternal contamination, wherein the false negative rate (FNR) is 0, the accuracy (Accuracy) is 0.85, and the recall rate (Recall) is 1.

[0133] Using the predicted maternal contamination ratio of the sample, correction of the sample CNV can be achieved.

[0134] In some embodiments, the present invention provides a method for predicting the proportion of maternal contamination in a sample, the method having the following steps:

[0135] Obtaining a dataset of methylation levels of markers of a sample; providing the dataset of methylation levels of markers as input to a trained machine learning model; and processing the dataset of methylation levels of the markers using the trained machine learning model to generate a proportion of maternal contamination in the sample.

[0136] The dataset of methylation levels of markers provided as input to the trained machine learning model is, for example, a dataset of methylation levels of 2-400, 5-200, 10-150, 20-120, 30-100 markers.

[0137] IV. Products of the method of the present invention

[0138] The present invention provides products for implementing any of the above aspects of the present invention, including but not limited to devices, systems and equipment.

[0139] In some embodiments, the present invention provides a device, apparatus or system for predicting the proportion of maternal contamination in a sample, comprising:

[0140] A data receiving unit, configured to receive a dataset of methylation levels of markers of a sample, wherein the dataset is, for example, a dataset of methylation levels of 2-400, 5-200, 10-150, 20-120, or 30-100 markers;

[0141] At least one processor and at least one memory, wherein the at least one memory stores thereon a trained machine learning model of the present invention, and when executed by the at least one processor, the trained machine learning model of the present invention processes the methylation level dataset of the marker to generate the maternal contamination ratio of the sample.

[0142] In some embodiments, the methylation level dataset of the marker is obtained from a server-based data store, and the methylation level dataset of the marker is uploaded to the data store by a client device.

[0143] In some embodiments, processing of the dataset of methylation levels of markers using the trained machine learning model is performed on a server device using the dataset of methylation levels of markers obtained from a client device.

[0144] In some embodiments, the maternal source contamination ratio is made accessible to the client device.

[0145] In some embodiments, the present invention provides a device, apparatus or system for predicting the proportion of maternal contamination in a sample and correcting CNV in a sample, comprising:

[0146] A data receiving unit, configured to receive CNV data of a sample and a dataset of methylation levels of markers of the sample, wherein the dataset is, for example, a dataset of methylation levels of 2-400, 5-200, 10-150, 20-120, or 30-100 markers;

[0147] At least one processor and at least one memory, wherein the at least one memory stores thereon a trained machine learning model of the present invention and a computer program for correcting sample CNV. When executed by the at least one processor, the trained machine learning model of the present invention processes the methylation level dataset of the marker to generate the maternal contamination ratio of the sample; and processes the CNV data of the sample by the computer program for correcting sample CNV to generate a corrected sample CNV.

[0148] In some embodiments, the CNV data of the sample and the methylation level dataset of the marker are obtained from a server-based data storage, and the CNV data of the sample and the methylation level dataset of the marker are uploaded to the data storage by the client device.

[0149] In some embodiments, processing of the marker methylation level dataset and correction of sample CNV using the trained machine learning model is performed on a server device using the CNV data of the sample and the marker methylation level dataset obtained from a client device.

[0150] In some embodiments, the maternal contamination ratio and the corrected sample CNV are made accessible to a client device.

[0151] The following examples are described to assist understanding of the present invention. The examples are not intended to and should not be interpreted in any way as limiting the scope of protection of the present invention.

[0152] Example

[0153] The invention generally described herein will be more readily understood by reference to the following examples, which are provided by way of illustration and are not intended to limit the scope of the invention. These examples are not intended to represent that the experiments below are all or the only experiments performed.

[0154] Materials and methods

[0155] Aborted fetal tissue or maternal amniotic fluid was obtained for gDNA extraction; and IVF blastocyst trophoblast cell biopsies and embryo culture fluid were obtained for single-cell genome amplification. All studies were approved by the institutional ethics committee and written informed consent was obtained before they were conducted.

[0156] Blastocyst biopsy

[0157] The blastocysts were discarded on the 5th / 6th day after in vitro fertilization for biopsy. A laser membrane rupture instrument was used to make a 10-15 μm small hole on the zona pellucida opposite to the inner cell mass of the blastocyst to assist hatching, and the culture was continued for 4-6 hours. After the trophoblast cells were hatched, laser cutting combined with aspiration biopsy was used to obtain some cells. For unhatched blastocysts, a biopsy needle was used to negatively aspirate some trophoblast cells from the opening of the zona pellucida for biopsy. After the biopsy, the blastocysts continued to be cultured.

[0158] Treatment of embryo culture medium

[0159] Embryo culture is performed according to clinical standards. IVF embryos are cultured in culture medium until day 3 (D3), and fresh culture medium is replaced. When the embryo develops to the blastocyst stage of stage 4, usually on the 5th or 6th day of in vitro culture (D5 or D6), the embryo is transferred to a new culture dish and the culture medium in the original culture dish is aspirated with a glass capillary or a pipette, usually 15 to 25 μL.

[0160] The collected embryo culture medium was placed in a 200 μL PCR tube, and proteinase K (NEW ENGLAND Biolabs, P8107S) was added. According to the manufacturer's instructions, the cell lysis reaction was performed on a PCR instrument under the following conditions (the hot lid was set at 105° C.):

[0161]

[0162] The lysate in a PCR tube, also referred to as a processed embryo culture fluid sample hereinafter, is taken out and divided into two parts, one for copy number variation (CNV) detection and the other for maternal contamination identification.

[0163] Example 1 Detection of copy number variation (CNV)

[0164] One of the processed embryo culture fluid samples described in Materials and Methods was tested for CNV by the following procedure.

[0165] 1) Determine the free DNA in the embryo culture medium, the original concentration of nucleic acid is about 1.1 pg / μl;

[0166] 2) For the free trace DNA in the embryo culture medium of step 1), use the ChromInst amplification kit (Cat. No. XK-005-96) of Yikang Gene Company to perform whole genome amplification through multiple annealing and looping based amplification cycles (MALBAC) according to the instructions of the kit;

[0167] 3) performing next-generation sequencing on the DNA product after whole-genome amplification in step 2) to generate raw reads data;

[0168] 4) Use BWA software to align the raw reads data from step 3) to the human genome reference sequence NCBIBuild37 / hg19;

[0169] 5) using Sambamba software to perform de-repeating sequence analysis on the alignment results of step 4), and then using the obtained sequence set as a set of unique matching sequences;

[0170] 6) dividing each human chromosome into 1000 kb windows, and dividing each sequence in the set of unique matching sequences obtained by analysis in step 5) into a corresponding chromosome window according to position information;

[0171] 7) For the sequences divided into the corresponding chromosome windows in step 6), GC correction is performed using the LOESS algorithm according to the GC content of the human genome reference sequence NCBI Build37 / hg19;

[0172] 8) Comparing the GC-corrected reads data of step 7) with the average reads value of the negative sample control set, wherein the negative sample control set is a set of 200 embryo samples that have been born and have normal chromosomes, and obtaining the ratio of each 1000 kb window of the sample to be tested to the negative sample used as the reference sample;

[0173] 9) The Circular Binary Segmentation (CBS) algorithm is used to connect windows with similar log2ratio values ​​to obtain a larger window area, thereby obtaining the region of copy number variation.

[0174] For gDNA extracted from aborted fetal tissue or maternal amniotic fluid, Sigma-Aldrich After whole genome amplification using the complete whole genome amplification (WGA) kit, step 3) to step 9) of Example 1 were performed.

[0175] Example 2 Determination of Chromosome Window Methylation Levels

[0176] Another aliquot of the processed embryo culture fluid sample described in the Materials and Methods was treated with a methylation-sensitive endonuclease and sequenced as follows ( Figure 2 and Figure 3 ) to obtain the methylation level of each chromosome window for the purpose of identifying maternal contamination.

[0177] 1) Determine the free DNA in the embryo culture medium, the original concentration of nucleic acid is about 1.1 pg / μl;

[0178] 2) Place the embryo culture medium from step 1) in a 200 μL PCR tube, add 1 μL rCutSmart buffer (New England BioLabs, B6004S) and 0.5 μL HpaII enzyme (New England BioLabs, R0171S), and mix well;

[0179] 3) Perform the reaction on a PCR instrument under the following conditions (heat cover set to 105°C):

[0180]

[0181] 4) ChromInst whole genome amplification library construction: Use the ChromInst whole genome amplification library construction one-step kit (XK-008, Xukang Medical Technology (Suzhou) Co., Ltd., Susu Equipment No. 20170425) to construct a sequencing library according to the instructions of the kit, and then purify the amplified sequencing library with magnetic beads, dissolve it in 20 μL of nuclease-free water, and determine the concentration by Qubit;

[0182] 5) Sequencing the library in step 4) using a high-throughput sequencer of the Illumina sequencing platform, the sequencing type is PE150, the sequencing data volume is not less than 200K reads, the sequencing data quality requirements are set to Q30>70%, Q20>85%, and raw reads data are generated and stored in fastq files;

[0183] 6) Align the raw reads data in the fastq file in step 5) to the human genome reference sequence NCBI Build37 / hg19 using the alignment software BWA to obtain the alignment result (bam file);

[0184] 7) Using sambamba software to perform markup on the alignment result (bam file) of step 6) to remove duplicate reads and obtain the alignment result (bam file) after markdup;

[0185] 8) Obtain the CCGG sequence from the human genome reference sequence NCBI Build37 / hg19, and store the genomic location of the CCGG sequence as a bed file;

[0186] 9) Extract the reads that overlap with the reads in step 8) and meet the conditions from the markdup bam file in step 7) and store them in a new BAM file (ccgg bam), where the conditions met are that the mapq of the read is greater than 30 and the NM value of the read is less than or equal to 1;

[0187] 10) Divide each human chromosome into N regional segments, where each regional segment is a 10Mb window (also called "bin"). In each window, the number of reads where ccgg bam overlaps with the window (i.e., the starting position of the read falls in the bin) is counted, marked as bin_reads_count. From these overlapping reads, the number of reads containing CCGG sequences is counted, marked as bin_ccgg_reads_count. Then, the CCGG ratio of each window is calculated using the following formula:

[0188] CCGG ratio=bin_ccgg_reads_count / bin_reads_count (Formula 1)

[0189] The gDNA extracted from aborted fetal tissue or maternal amniotic fluid was treated with 0.5 μL HpalI enzyme (New England BioLabs, R0171 S) and then purified by Sigma-Aldrich The whole genome amplification (WGA) kit was used for whole genome amplification, and then steps 5) to 10) of Example 2 were performed. Figure 1 A flowchart showing the operation when using fetal tissue, umbilical cord blood, or maternal amniotic fluid containing a large amount of fetal DNA as a sample.

[0190] Example 3 Establishment of machine learning model

[0191] 1) Load the dataset and split it into training and testing sets.

[0192] Based on a method for identifying parental tendency or parental contamination of nucleic acid samples proposed earlier by the inventors, for genomic DNA (gDNA) samples, the relative allele frequency is used to identify whether the nucleic acid sample has parental contamination based on the BAF (Ballele frequency) determined by the chip (see patent application publication number: CN114214425A), and this parental contamination ratio is used as the gold standard for identifying parental contamination.

[0193] Using the method disclosed in patent application publication number CN1 14214425A, the parental bias or parental contamination of several nucleic acid samples was identified, from which 75 samples were randomly selected as a training set and 34 samples were randomly selected as an independent validation set.

[0194] Table 1 below shows the gold standard measurement values ​​of the maternal contamination ratio from 75 samples (blastocyst culture fluid samples from clinical tests) used as a training set, and Table 2 below shows the gold standard measurement values ​​of the maternal contamination ratio from 34 samples used as an independent validation set.

[0195] Table 1 Names of 75 samples in the training set and gold standard values ​​of maternal source pollution ratios

[0196]

[0197]

[0198]

[0199]

[0200] Table 2 Names of 34 samples in the independent validation set and gold standard values ​​of maternal contamination ratios

[0201]

[0202]

[0203]

[0204] 2) The steps of the above-mentioned embodiment 2 of the present invention were implemented on the 75 samples of the training set. The training set in step 1) of embodiment 3 was subjected to Pearson correlation analysis and Wilcoxon rank sum test to screen the windows with significant differences in methylation levels.

[0205] Method A: The CCGG Ratio value of the first window (chr1: 0-10000000) was calculated according to formula 1 in Example 2 of the present invention, and then the CCGG Ratio values ​​of the 75 samples in the training set (expressed as "ccgg_ratio_of_samples") and the maternal contamination ratio of the corresponding gold standard in Table 1 (expressed as "contam_ratio_list") were subjected to Pearson correlation analysis, wherein the scipy.stats method in Python was used: correlation, p_value=stats.pearsonr(contam_ratio_list, ccgg_ratio_of_samples), and the obtained p_value was further used to calculate the P value of the multiple hypothesis test of the Bonferroni method, that is, p_value*n (when the genome is divided into window sizes according to 10M bin size and the step size is set to 2Mb, the number of windows n is 1464 windows; p_value*n means multiplying the p value by n, thereby obtaining the P value of the multiple hypothesis test). The P values ​​of the multiple hypothesis tests for each window were then ranked.

[0206] Method B: Alternatively, the maternal contamination ratio determined according to the gold standard is set at 30% as the threshold, and the training set with a maternal contamination ratio greater than or equal to 30% in Table 1 is divided into a case group (46 cases) and a control group (29 cases) with a maternal contamination ratio less than 30%. According to the calculation method of formula 1 in the above embodiment 2 of the present invention, the CCGG Ratio value of the first window (chr1: 0-10000000) is calculated, and then the CCGG Ratio values ​​of the corresponding 46 case group samples (the 46 CCGG ratio values ​​are written into the variable "ccgg_ratio of_case_samples") and the CCGGRatio values ​​of the corresponding 29 control group samples (the 29 CCGG ratio values ​​are written into the variable "ccgg_ratio_of_control_samples") are subjected to Wilcoxon rank sum test, wherein the scipy.stats method in Python is used: stat, p = mannwhitneyu (ccgg_ratio_of_case_samples, ccgg_ratio_of_control_samples, alternative = 'two-sided'), and the obtained p value is used to further calculate the P value of the multiple hypothesis test of the Bonferroni method, that is, p_value*n (when the genome is divided into 10M bins size is used to divide the window size, and when the step size is set to 2Mb, the number of windows n is 1464 windows; p_value*n means multiplying the p value by n, thereby obtaining the P value of the multiple hypothesis test), and then sorting the P value of the multiple hypothesis test of each window. Calculate the windows with significant differences in methylation levels between the maternal pollution group and the control group determined by the gold standard method, and obtain the windows with significant differences in methylation levels.

[0207] By implementing the above two methods and evaluating the model performance of the two methods, the following Table 3 is obtained.

[0208] Table 3 Comparison of Method A and Method B

[0209]

[0210] Compare the results of method A and method B. As shown in Table 3, by Pearson correlation analysis, the Bonferroni P value was corrected, and the optimal P value cutoff value was 0.0001. When the P value was less than 0.0001, 218 regions were obtained on the genome as candidate markers for windows with significant differences in embryonic and parental methylation levels (Table 4). That is, with a P value of 0.0001 as a cutoff value, when the P value was less than 0.0001, the window with significant differences in embryonic and parental methylation levels was determined as a candidate marker.

[0211] Table 4 Chromosome locations of windows with significant differences in methylation levels between embryos and parents

[0212]

[0213]

[0214]

[0215]

[0216]

[0217] 3) The candidate markers selected in step 2) of Example 3 are used to train the model using the training set. When training the model, the method used is Random Forest.

[0218] The evaluation method used to evaluate the model performance is the leave-one-out method.

[0219] The 218 candidate markers obtained under the combination of the parameters in step 2) of Example 3 (i.e., the first column of data of method A in Table 3 is used as the combination of parameters) are used as features for auxiliary model training, aiming to obtain a model with optimal performance, which should have these 218 markers. Thus, a model is constructed based on these 218 candidate markers.

[0220] 4) Use the independent validation set samples in Table 2 to validate the model constructed in step 3), and use the model performance evaluation method in step 3) to evaluate the performance indicators of the independent validation set.

[0221] First, the methylation level of each chromosome window of each sample was obtained for the 34 samples in the independent validation set of Table 2 using the method of Example 2 (except that the samples used were different), and the methylation level of the marker was extracted from it to obtain the marker-related methylation level data set of the independent validation set. Secondly, the marker-related methylation level data set of the independent validation set was used to verify the maternal contamination prediction of the model. The prediction method is the predict function in the R language, which is used to make predictions based on existing models and can be applied to many model objects, including linear regression, generalized linear models, decision trees, random forests, etc. Finally, the performance of the model was evaluated based on the prediction results, and the results are shown in FIG. Figure 4 As shown in Table 5, the qualitative performance indicators of the random forest model are given when the cutoff value (cutoff) of the maternal pollution group vs the non-maternal pollution group is 0.3 and the P value cutoff (cutoff) is 0.0001.

[0222] Table 5 Qualitative performance indicators of random forest model

[0223]

[0224]

[0225] 5) Quantitative analysis of maternal pollution: The best 218 marker training set samples in step 2) are subjected to regression analysis, and the predicted value of maternal pollution is obtained through the following format of the predict function: predict(object, newdata, ...). Thus, the maternal pollution of the sample is predicted through the trained model to obtain the predicted value of the maternal pollution of the sample.

[0226] The method of Example 3 was further applied to other blastocyst culture fluid samples from clinical tests in parallel. The comparison of the maternal contamination ratio obtained by methylation detection and the maternal contamination ratio obtained by the gold standard is shown in Table 6.

[0227] Table 6 Prediction value of maternal contamination by methylation and proportion of maternal contamination obtained by gold standard

[0228]

[0229]

[0230] As can be seen from Table 6, the methylation test can accurately predict the proportion of maternal contamination.

[0231] Example 4 CNV maternal contamination calibration

[0232] Maternal contamination correction can detect chromosomes related to fetal karyotype abnormalities and avoid false negatives. CNV correction is applied to another sample tested in parallel. Copy number correction of fetal CNV is performed according to the maternal contamination ratio. CNV of the fetal DNA sample is corrected using the following formula to obtain the corrected CNV:

[0233] Male X chromosome correction formula:

[0234]

[0235] N: corrected CNV copy number (calculated according to haploid);

[0236] S: CNV copy number before correction (calculated according to haploid);

[0237] M: Maternal source pollution ratio (>0 and <1)

[0238] Male Y chromosome correction formula:

[0239]

[0240] N: corrected CNV copy number (calculated according to haploid);

[0241] S: CNV copy number before correction (calculated according to haploid);

[0242] M: Maternal source pollution ratio (>0 and <1)

[0243] Correction formula for autosomes and female X chromosome:

[0244]

[0245] N: corrected CNV copy number (calculated according to haploid);

[0246] S: CNV copy number before correction (calculated according to haploid);

[0247] M: Maternal source pollution ratio (>0 and <1)

[0248] Note: For the autosome and female sex chromosome copy number, it is necessary to convert it to diploid, and the output copy number value should be 2*N; that is, female sex chromosome copy number / autosome copy number = 2*N.

[0249] For the specific method of CNV correction, please refer to a method or device for identifying maternal contamination and copy number correction previously proposed by the inventor (see patent application publication number: CN117558343A).

[0250] Figure 5Table 7 shows the comparison of fetal DNA CNV results before and after correction of CNV maternal contamination in a later parallel test sample.

[0251] Table 7. Comparison of CNV before and after correction for maternal contamination

[0252]

Claims

1. A method for determining the chromosome methylation level of a test sample, the method comprising: a. Treat the test sample with a methylation-sensitive restriction endonuclease, b. performing whole genome amplification on the DNA of the test sample treated with methylation-sensitive restriction endonuclease to construct a sequencing library; c. Implement high-throughput sequencing to obtain raw reads data; d. Align the raw reads data from step c to the human genome reference sequence version using alignment software; e. performing de-repeating sequence analysis on the alignment result of step d to remove duplicate reads, and after the analysis, the obtained read sequence set is used as a set of unique matching sequences; f. Extracting reads from the read sequence set obtained in step e that overlap with the CCGG sequence obtained from the human reference genome and satisfy the read alignment quality between 20-60 and the NM value between 0-2, and obtaining a read sequence set that overlaps with the ccgg sequence of the human reference genome; g. Divide each human chromosome into N regional segments according to a window size of 5Mb-50Mb, where each regional segment is a window, also called "bin". In each window, count the number of reads overlapping the window in the set of reads sequences overlapping the human reference genome ccgg sequence obtained in step f, marked as bin reads count, and count the number of reads containing CCGG sequences from these overlapping reads, marked as bin_ccgg_reads_count, and then calculate the CCGG ratio of each window. The calculation formula is: CCGG ratio=bin_ccgg_reads_count / bin_reads_count (formula 1) Thus, the methylation level of each window on the chromosome of the test sample is determined, and a data set of the methylation level on the chromosome of the test sample is obtained.

2. The method according to claim 1, wherein the method comprises: a. treating the test sample with a methylation-sensitive restriction endonuclease, wherein the methylation-sensitive restriction endonuclease does not cut the DNA methylation region, for example, the methylation-sensitive restriction endonuclease is selected from HpaII, SmaI, MspI, XmaI and BstUI b. performing whole genome amplification on the DNA of the test sample treated with methylation-sensitive restriction endonuclease to construct a sequencing library; c. Implement high-throughput sequencing to obtain raw reads data; d. Align the raw reads data from step c to the human genome reference sequence version such as NCBIBuild37 / hg19, hg38 using alignment software, such as BWA software; e. performing de-repeating sequence analysis on the alignment result of step d to remove duplicate reads, for example, using Sambamba software to perform de-repeating sequence analysis on the alignment result of step d, and after the analysis, the obtained read sequence set is used as a set of unique matching sequences; f. Extracting reads from the read sequence set obtained in step e that overlap with the CCGG sequence obtained from the human reference genome and meet the conditions such as mapq greater than 30 and NM value less than or equal to 1, and obtaining a read sequence set that overlaps with the ccGG sequence of the human reference genome; g. Divide each human chromosome into N regional segments according to, for example, 5Mb, 10Mb, 15Mb, 20Mb, 25Mb, 30Mb, 35Mb, 40Mb, 45Mb or 50Mb, wherein each regional segment is a window, also referred to as a "bin", and count the number of reads overlapping the window in the set of reads sequences overlapping the human reference genome ccgg sequence obtained in step f in each window, marked as bin reads count, and count the number of reads containing CCGG sequences from these overlapping reads, marked as bin_ccgg_reads_count, and then calculate the CCGG ratio of each window, and the calculation formula is: CCGG ratio=bin_ccgg_reads_count / bin_reads_count (formula 1) Thus, the methylation level of each window on the chromosome of the test sample is determined, and a data set of the methylation level on the chromosome of the test sample is obtained.

3. The method according to claim 1 or 2, wherein the raw reads data is obtained using a second generation sequencing platform selected from, for example, Illumina, Life Technologies and BGI.

4. The method according to any one of claims 1-3, wherein the sample is an offspring or embryo sample, for example selected from: embryonic tissue (e.g., abortion product tissue), maternal amniotic fluid, maternal peripheral blood, umbilical cord blood, embryo biopsy cells, trace embryonic cells, free DNA in embryo culture medium.

5. A method for establishing a model for predicting maternal source contamination, which is implemented on a computer system including one or more processors and system memory, the method comprising: a. providing a training set consisting of a plurality of (e.g., 2-400, 5-200, 10-150, 20-120, 30-100) test samples on a computer system, and determining the maternal contamination ratio of the plurality of test samples in the training set according to the relative allele frequency (e.g., B allele frequency); b. implementing the method according to any one of claims 1 to 4 on the plurality of test samples in the training set to obtain the methylation level of each window on the chromosome of the test sample; c. Performing a correlation analysis (e.g., Pearson correlation analysis, Spearman rank correlation analysis, Kendall rank correlation analysis) on the methylation level of each window on the chromosome of the test sample in step b and the maternal contamination ratio determined in step a, and obtaining the window of the difference in methylation level between the embryo and the parent as a marker; d. training a model for predicting maternal contamination using the training set of step a for the markers of step c and evaluating the performance of the model for predicting maternal contamination; e. Use an independent validation set consisting of multiple additional test samples (e.g., 2-400, 5-200, 10-150, 20-120, 30-100) to validate the model for predicting maternal contamination in step d to obtain performance characteristics of the model for predicting maternal contamination, for example, a false negative rate (FNR) of 0, an accuracy of 0.85, and a recall of 1.

6. Use of the window of difference in methylation level between the embryo and the parent on a chromosome, for generating a prediction of the proportion of maternal contamination by applying the model for predicting maternal contamination established by the method of claim 5 to a dataset of methylation levels in the window of difference in methylation level between the embryo and the parent on a chromosome, wherein the window and methylation levels of the window of difference in methylation level between the embryo and the parent on a chromosome are obtained by steps ac in the method of claim 5, for example, the window of difference in methylation level between the embryo and the parent on a chromosome is the window shown in Table 4.

7. Methods for predicting sample maternal contamination, including: a. implementing the method according to any one of claims 1 to 4 on the sample, determining the methylation level of the window on the chromosome where the methylation level of the embryo is different from that of the parent, and obtaining a data set of the methylation level of the window on the chromosome where the methylation level of the embryo is different from that of the parent; b. generating a prediction of the proportion of maternal contamination by applying the model for predicting maternal contamination established by the method of claim 5 to the data set of step a, Thus, the maternal contamination of the sample is predicted.

8. A method for correcting maternal contamination of CNV of a sample, comprising: a. Implementing the method of claim 7 on the sample to obtain a predicted maternal contamination ratio of the sample, b. Use the sample maternal contamination ratio in step a to correct the CNV of the sample.

9. A device, apparatus or system for predicting the proportion of maternal contamination in a sample, comprising: A data receiving unit, configured to receive a dataset of methylation levels of markers of a sample, wherein the dataset is, for example, a dataset of methylation levels of 2-400, 5-200, 10-150, 20-120, or 30-100 markers; At least one processor and at least one memory, wherein the at least one memory stores thereon the trained machine learning model of claim 5, and when executed by the at least one processor, the trained machine learning model processes the methylation level dataset of the marker to generate the maternal contamination ratio of the sample.

10. A device, apparatus or system for predicting the proportion of maternal contamination in a sample and correcting CNV in a sample, comprising: A data receiving unit, configured to receive CNV data of a sample and a dataset of methylation levels of markers of the sample, wherein the dataset is, for example, a dataset of methylation levels of 2-400, 5-200, 10-150, 20-120, or 30-100 markers; At least one processor and at least one memory, the at least one memory having stored thereon the trained machine learning model of claim 5 and a computer program for correcting sample CNV, wherein when executed by the at least one processor, the trained machine learning model processes the methylation level dataset of the marker to generate a maternal contamination ratio of the sample; The CNV data of the sample is processed by a computer program for correcting the sample CNV to generate a corrected sample CNV, and the CNV correction result of the sample is output.

11. Use of the device, apparatus or system according to claim 9 or 10 for predicting the proportion of maternal contamination in a test sample, and optionally correcting the CNV of the sample.

Images