High resolution and non-invasive fetal sequencing

Abstract

Provided herein are computer-implemented methods for assigning maternal or fetal origin to one or more genetic variations in cell-free DNA (cfDNA) of a sample from a pregnant mammal, preferably a pregnant human, it uses a probabilistic model for assigning maternal or fetal origin to genetic variations in DNA from a sample obtained from a pregnant mammal, where the model assigns maternal or fetal origin based on a combination of fetal fraction and DNA fragment size.

Description

[0001] CLAIM OF PRIORITY

[0002] This application claims the benefit of U.S. Provisional Application Serial No. 63 / 402,379, filed August 30, 2022. The entire contents of the foregoing application are incorporated herein by reference.

[0003] FEDERALLY SPONSORED RESEARCH OR DEVELOPMENT

[0004] This invention was made with government support under Grant Number HD081256 awarded by the National Institutes of Health. The government has certain rights in the invention. TECHNICAL FIELD

[0005] Provided herein are methods (e.g., computer-implemented methods) for assigning maternal or fetal origin to one or more genetic variants in cell free DNA (cfDNA) from a sample from a pregnant mammal (preferably a pregnant human) using a probabilistic model for assigning maternal or fetal origin to genetic variants in DNA from a sample obtained from a pregnant mammal, wherein the model assigns maternal or fetal origin based on a combination of fetal fraction and DNA fragment size. BACKGROUND

[0006] Early genetic diagnosis of the fetus can guide clinical management, predict outcomes, and provide a basis for precision medicine 1-3 . SUMMARY

[0007] Provided herein are computer-implemented methods that can be used to assign maternal or fetal origin to one or more genetic variants in cell free DNA (cfDNA) from a sample from a pregnant mammal (preferably a pregnant human). The method comprises (a) accessing from memory a probabilistic model for assigning maternal or fetal origin to genetic variants in DNA from a sample obtained from a pregnant mammal, wherein the model assigns maternal or fetal origin based on a combination of fetal fraction and / or DNA fragment size and other sequencing characteristics; (b) inputting into the model a set of values representing one or more genetic variants detected in cfDNA from a peripheral blood sample from a pregnant mammal, wherein the values include empirically determined sequence information, e.g., the proportion of different bases in reads, and DNA fragment size information, e.g., rank sum statistics for each genetic variant; and (c) using the model to assign maternal or fetal origin to the one or more genetic variants.

[0008] In some embodiments, the genetic variants include single nucleotide variants (SNVs), insertions or deletions (indels), and / or copy number variants (CNVs).

[0009] In some embodiments, the initial set of values representing one or more genetic variants is obtained by a method comprising: aligning raw sequencing reads derived from cfDNA to a reference genome sequence; converting the raw sequencing reads to consensus reads; re-aligning the consensus reads to the reference genome sequence, thereby producing a set of aligned consensus reads; identifying consensus reads that differ from the reference genome; assigning consensus reads that differ from the reference genome as alternate alleles and consensus reads that match the reference genome as reference alleles, and determining a fragment size rank sum statistic representing a distribution of estimated fragment sizes of reads supporting reference alleles compared to a distribution of fragment sizes of reads supporting alternate alleles, thereby obtaining an initial set of values representing sequence identity and DNA fragment size rank sum statistics of one or more genetic variants.

[0010] In some embodiments, each raw sequencing read comprises a unique molecular identifier (UMI); and the method comprises converting raw sequencing reads to a single consensus read per UMI.

[0011] In some embodiments, the method further comprises selecting the set of candidate variants prior to step (b) by a method comprising: accessing a machine learning classifier from a memory, optionally a model based on a random forest, wherein the machine learning classifier is trained using a predetermined set of filtering criteria and a subset of loci present in the sample or in a reference sample to identify potential false positive (FP) loci; inputting the initial set of variants into the machine learning classifier; and filtering using the trained machine learning classifier to remove the set of variants that are rich in false positive (FP) loci, thereby selecting the set of candidate variants from the initial set.

[0012] In some embodiments, the probabilistic model uses a k-mean or Bayesian Mixture Model that simultaneously estimates fetal fraction and assigns fetal or maternal origin to each variant locus in the set. In some embodiments, the Bayesian Mixture Model is a Bayesian Gaussian Mixture Model constrained via variant allele fraction and fragment size, e.g., fragment size rank sum statistic.

[0013] In some embodiments, the fetal fraction of the sample is modeled as a latent variable (f), and the mean of the variant allele fraction distribution is set for each component based on f.

[0014] In some embodiments, the fetal fraction is estimated according to a reference fetal fraction determined based on clusters derived from VAFs across sites.

[0015] In some embodiments, the method further comprises outputting a list of one or more genetic variants identified as having a fetal origin and / or one or more genetic variants identified as having a maternal origin.

[0016] In some embodiments, the method further includes comparing the genetic variants to a database comprising a list of genetic variants and information about variants of potential medical relevance to the fetus or mother; identifying variants present in the potentially medically relevant fetus or mother; and outputting a list of potentially medically relevant one or more genetic variants identified as having a fetal origin and / or one or more genetic variants identified as having a maternal origin.

[0017] In some embodiments, the method further comprises the method may further comprise recommending further testing based on the presence of a potentially medically relevant variant.

[0018] In some embodiments, further testing includes amniocentesis or chorionic villus sampling (CVS); further monitoring of the fetus via ultrasound; or genetic testing of the mother.

[0019] In some embodiments, the method further comprises using high-throughput sequencing on cfDNA extracted from a single sample of peripheral blood from the mother, optionally wherein exome capture is performed prior to sequencing. The present method does not require (and typically does not) use a paternal blood sample or sequence (e.g., for benchmarking or any other purpose), and optionally does not use a separate maternal-only sample (e.g., for benchmarking or any other purpose); the method can include, but is not required to, determine the maternal genotype from white blood cells as described herein, and in some embodiments, the method is performed using only cfDNA from a single sample of maternal plasma. Thus, the present method can be performed using a single sample without the need for independent samples from maternal and paternal genomes, or normalization to a reference panel.

[0020] In some embodiments, adaptors with universal PCR primer sequences and unique molecular identifiers (UMIs) are ligated to the cfDNA and PCR amplified prior to sequencing.

[0021] In some embodiments, the method further comprises enriching fetal DNA in the sample, optionally by contacting the cfDNA with a plurality of oligonucleotides that bind to portions of the fetal genome, optionally including fetal protein-coding genes or other regions of the fetal genome that can be relevant for clinical interpretation or variant calling.

[0022] 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 to which this application belongs. Methods and materials are described herein for use of the present application; other suitable methods and materials known in the art can also be used. The materials, methods, and examples are illustrative only and not intended to be limiting. All publications, patent applications, patents, sequences, database entries, and other references mentioned herein are incorporated by reference in their entirety. In case of conflict, the present specification, including definitions, will control.

[0023] Other features and advantages of the present application will be apparent from the following detailed description, and from the drawings, and from the claims. BRIEF DESCRIPTION OF DRAWINGS

[0024] Figure 1A -C. Workflow of non-invasive fetal exome screening with NIFS. Figure 1A Procedure showing extraction of cell-free DNA (cfDNA) from maternal plasma followed by exome capture. We generated 51 libraries across gestational age and calculated the fetal fraction (fetal cfDNA / total cfDNA) in each sample. Figure 1B The novel variant calling method developed is highlighted to account for the maternal and fetal genotype combinations present in the cfDNA to explain the fetal fraction and the corresponding unique allele fraction at each locus. Each cluster represents a unique maternal / fetal genotype combination, and the clusters are colored according to the genotypes generated from direct exome sequencing (ES) of the maternal and fetal DNA. These data highlight the merging of clusters at heterozygous loci in mothers with lower coverage and fetal fraction, and unambiguous resolution of de novo and paternal origin variants (fetal 0 / 1; maternal 0 / 0) irrespective of the fetal fraction. Figure 1CRepresentative variants of clinical interest in Table 4, including a likely pathogenic splice variant in COL2A1 (NC_000012.12: g.47982610C>T) in a fetus with micrognathia deformity consistent with Stickler syndrome, a 4 MB pathogenic deletion on chromosome 7 (NC_000007.14: g.155368937-159327017del) in a fetus with multiple congenital deformities, and a PAH risk variant (NC_000012.12: g.102866632C>T) associated with phenylketonuria (PKU) detected in a maternal carrier screen, are shown with NIFS application to 14 cases referred for invasive testing. All variants were orthogonally validated (Table 2).

[0025] Figure 2 . Exemplary workflow. An exemplary workflow in which data processing is divided into three stages. In the alignment and preprocessing stage, raw sequencing reads from cfDNA-derived exome-based sequencing (ES) are aligned to a reference genome, grouped by unique molecular identifier (UMI), and converted to a single consensus read for each UMI. The consensus reads are then realigned to the reference and recalibrated base quality scores, producing a set of aligned consensus reads that are prepared for downstream variant calling and analysis. Next, in the variant filtering and genotyping stage of the workflow, candidate variant sites are identified using Mutect2; variants are filtered using a set of hard filters and a random forest-based model trained on a subset of sites present in the sample; and a Bayesian mixture model is used to simultaneously estimate fetal fraction and assign fetal and maternal genotypes to each site. Finally, in variant interpretation, all passing variants are annotated and evaluated to produce a list of clinically relevant variants for interpretation.

[0026] Figure 3A -C. Evaluation of "unfiltered variant detection", "filtered variant detection", "overall genotyping performance", "paternal or de novo variant detection", and "genotyping accuracy of heterozygous variants in the mother" on fetal fraction. Theoretical sensitivity and detection of non-maternal variants are strong on fetal fraction, while genotyping accuracy, especially for variants that are heterozygous in the mother, is worse at lower fetal fractions. Sensitivity: TP / (TP+FN); PPV: TP / (TP+FP); Genotype accuracy: percentage of maternal heterozygous variants assigned correct fetal genotype.

[0027] Figure 4We were able to distinguish male and female cases by assessment of sequencing coverage of chrY. Examination of the number of intervals of chrY with mapped sequencing reads allowed us to detect a confirmed vanishing twin with a male fetus in a male (Table 13), which had a much greater proportion of read coverage on chrY than other samples from pregnancies with female fetuses. Investigation of the predicted chrY duplication status was less accurate, but we did find one extreme case in which the mother had received a stem cell transplant from a male donor and therefore had 6-fold higher coverage on chrY than expected. In addition, lower normalized chrY depth distinguished twin pregnancies with discordant gender. We also highlighted single samples with lower than expected fetal fraction that exhibited reduced coverage on chrY, as well as pregnancies with two XX fetuses that clustered with other samples that had been pregnant with a single XX fetus.

[0028] Figure 5 . Variant allele fraction plot. Histogram of observed variant allele fraction (proportion of reads supporting the alternative allele) as plotted for all autosomal sites in a sample with 38% fetal fraction at 268x coverage. Peaks showing the distribution based on their mean and variance of fetal fraction and coverage, with assignment of them to maternal or fetal genomic genotypes.

[0029] Figure 6 . Overview of non-invasive fetal sequencing (NIFS). We collect plasma (a mixture of fetal and maternal DNA; step 1), then extract DNA using (step 2). UMI are ligated and exome sequences are captured (step 3). Bioinformatics analysis is then performed to generate accurate exome-wide fetal and maternal genotypes (step 4). Interpretation is performed using best clinical practice (step 5).

[0030] Figure 7A - A) shows the process for extracting cell-free DNA (cfDNA) from maternal plasma followed by exome capture. We are able to extract both plasma consisting of fetal and maternal DNA as well as DNA from white blood cells that is only maternal DNA. Unique maternal DNA from white blood cells can be used for independent variant validation and maternal carrier screening. B) ligate UMI, capture exome, build sequencing library and sequence on Illumina machine.

[0031] Figure 8Fragment size as a genotyping feature. Each plot shows the distribution of fragment sizes that can unambiguously assign a read to either fetal or maternal origin. If a read overlaps a variant site that allows one allele to be unambiguously associated with either the fetus or mother based on the genotype derived from maternal and cord blood or amniocentesis WES, the read is assigned origin. For example, if the maternal genome is homozygous reference and the fetal genome is heterozygous, all reads supporting the alternate allele can be assigned fetal origin. In each sample, fetal origin fragments are enriched for small fragments and depleted of larger fragments relative to maternal origin fragments.

[0032] Figure 9 Filtering and genotyping performance. Cell-free fetal DNA is enriched for short fragments compared to maternal and we designed a rank sum test to show these biases. Lower rank sum statistics indicate an increased number of shorter fragments, which indicates that the variant is more likely to be of fetal origin. This information correlates well with VAF predictions and we use both metrics in our genotyping method.

[0033] Figure 10 Variant calling workflow. Overview of the variant calling process involving initial variant detection with mutect, which can optionally be filtered by maternal genotype if generated from white blood cells as described in Figure 7. Variants are initially filtered to remove false positives with a machine learning technique and genotyped with a Bayesian Gaussian mixture model as described below.

[0034] Figure 11 Model diagram for graphical model used for genotype assignment. This model is a Bayesian Gaussian mixture model defined on the variant allele fraction at each site and the fragment size statistic (computed by the InsertSizeRankSumTest), where the mean of the variant allele fraction component is constrained by a latent variable that estimates the fetal fraction. The model information is shown in the table below.

[0035] Figure 12 Example data processing workflow.

[0036] Figure 13 is a schematic diagram of an example computer system. DETAILED DESCRIPTION

[0037] Non-invasive prenatal screening (NIPS) has had a transformative effect on the discovery of aneuploidy. However, recent systematic benchmarking studies show that this current state-of-the-art low-resolution approach only captures a small fraction of pathogenic and likely pathogenic variants in fetuses with ultrasound-detected structural anomalies (~26% diagnostic yield from aneuploidy screening alone), whereas a much larger diagnostic yield can be achieved through sequencing and analysis of nucleotide changes and copy number variants (CNVs) that alter protein-coding sequences in the human genome (42-48%) (Lowther et al. 2020, Biorxiv; to be published in AJHG). Recent studies have shown that targeted approaches to a small set of genes or CNVs known to be associated with prenatal diagnosis can target in cfDNA 4-15 However, comprehensive genetic screening of fetal coding sequences or the whole genome during pregnancy still requires invasive procedures such as amniocentesis that pose inherent risks to both the mother and fetus.

[0038] Demonstrated here is an integrated molecular and computational process to facilitate scalable and non-invasive fetal sequencing (NIFS) to discover and annotate single nucleotide sequence changes and CNVs from circulating cell-free DNA (cfDNA) extracted from maternal plasma. This high-resolution NIFS approach can survey all >22,000 protein-coding genes in the fetal exome (NIFS-E) that encompass almost all interpretable pathogenic changes in fetal diagnostic testing (see Lowther et al., benchmarking from invasive studies of fetal structural anomaly (FSA) cases) or applied to whole genome sequencing (NIFS-G). Here, we focus on NIFS-E as a novel approach to provide non-invasive survey of the complete fetal exome as well as routine maternal carrier screening during pregnancy without the need for paternal samples Figure 1A -C). The success of this approach has implications for the replacement of current standard-of-care microarray and exome sequencing from invasive procedures for prenatal genetic diagnosis, as well as exploration of newborn sequencing, infant screening, and maternal carrier detection enterprises.

[0039] We sequenced samples from 51 pregnancies, including 37 samples from third trimester pregnancies for method development and 14 samples collected during early (n=5) or mid-pregnancy (n=9) consistent with current applications of NIPS, and observed fetal fractions in the range of 6% to 51%. Applying NIFS, we captured and sequenced 22,995 genes from the fetal genome (18,049 protein-coding genes). We detected and genotyped single nucleotide variants (SNVs) and insertions or deletions using a custom pipeline that applies a Bayesian Gaussian mixture model to account for fetal fraction while incorporating methodological features such as fragment length and sequencing characteristics of cfDNA into our model. We further utilized these features and read depth ratios to discover CNVs from cfDNA samples. The NIFS method generated a median of 203-fold coverage of the exome at a much lower sequencing cost than clinical whole-genome sequencing. We further benchmarked NIFS against 11 samples with germline exome sequencing from cord blood or amniocentesis and captured 99.7% of all 298,576 SNV sites detectable from standard exome sequencing while maintaining a median sensitivity of 93.0% and accuracy of 93.4% after accurate genotyping of all variants. The fetal sex was accurately inferred for all samples. Importantly, the fetal fraction from our method had minimal impact on de novo or paternally inherited variants, indicating the ability to screen for de novo variants in the fetus at very early stages of pregnancy.

[0040] As a validation experiment, we evaluated the clinical utility of NIFS in 14 pregnancies that underwent routine genetic testing and detected 100% of the variants of interest identified from clinical testing from current standard of care, including a pathogenic de novo CNV in a fetus with multiple congenital malformations, a likely pathogenic splicing variant in COL2A1 in a fetus with a small jaw deformity, and a homozygous pathogenic insertion or deletion in CFTR that likely causes cystic fibrosis. This NIFS method captures both maternal and fetal cfDNA, which also provides high sensitivity discovery of maternal SNVs (98.3% sensitivity relative to standard exome sequencing) and carrier screening that yields at least one reportable variant in 57.1% of the mothers evaluated, consistent with previous estimates 16,17 .

[0041] The potential utility of NIFS in prenatal screening is broad. The method provides nucleotide resolution screening to replace current low resolution NIPS methods. It can also provide rapid reflex testing for pregnancies with ultrasound abnormalities prior to the need for invasive procedures. When current exome sequencing is required, neonatal variant findings can be utilized and reinterpreted 18,19 , which can significantly reduce the time to diagnosis of many conditions. We also identify variants that can modulate risk of related late onset conditions (i.e. BRCA2 variants related to breast cancer risk), and such variants can be interpreted based on the parent risk based on the American College of Medical Genetics standard for reportable secondary findings 20,21 . In fact, the NIFS method provides genetic data comparable to current invasive prenatal exome sequencing, and requires the same detailed guidelines and procedures for interpretation of results and appropriate feedback 22 . These analyses demonstrate that the complete fetal exome can be accessed from the same maternal plasma sample that has been routinely collected for lower resolution fetal screening using new molecular and analytical methods such as NIFS.

[0042] Non-invasive fetal sequencing (NIFS) method

[0043] Variant allele fraction can be used to inform predictions about small variant genotypes, as the genetic variants present in cfDNA are a mixture of fetal and maternal fragments. The reads supporting a variant depend on the maternal and fetal genotypes and the fetal fraction; these patterns help to predict the genotype of both the mother and the fetus. Figure 5 An exemplary plot showing VAF plotted against annotation frequency to show the components assigned using the Bayesian Gaussian mixture model described herein. Fetal fraction reduction causes the cluster means to shift. Sequencing depth can also affect the outcome, as lower coverage results in higher intra-cluster variance; low coverage and low fetal fraction can pose a challenge to the ability to distinguish fetal genotypes based on VAF in sites where the mother is heterozygous alone.

[0044] In the present method, fetal variants can be uniquely detected with high sensitivity and specificity using the NIFS analysis pipeline as described herein, which also takes into account fragment size.

[0045] Figure 6 A schematic overview of an exemplary NIFS workflow is provided; the exemplary workflow is in Figure 2As shown, the method is performed on a sample collected from a pregnant woman (step 1, although the method can be performed on a previously collected sample and the method does not require a sample collection step). In step 2, cfDNA (and optionally maternal DNA, e.g., obtained from white blood cells) is extracted from the sample. Next, exome capture is optionally performed, and in step 3, the cfDNA (and optionally maternal DNA) is sequenced. In step 4, bioinformatics analysis of the sample is performed, and in step 5, variant interpretation is performed.

[0046] The sample (i.e., a peripheral blood sample) can be collected using methods known in the art. In some embodiments, 5-40 ml, e.g., 20 ml, is collected in a pregnant subject via a blood draw. The present method can be used in mammals, e.g., humans or non-human veterinary subjects. The present method does not require (and typically does not) use a paternal blood sample or sequence (e.g., for benchmarking or any other purpose), and optionally does not use a separate maternal-only sample (e.g., for benchmarking or any other purpose); the method may include, but is not required to, determine the maternal genotype from white blood cells as described herein, and in some embodiments, the method is performed using only cfDNA from a single sample of the mother's plasma. Thus, the present method can be performed using a single sample without the need for independent samples from the maternal and paternal genomes, or normalization to a reference group.

[0047] After the sample is plasma separated, cfDNA is extracted from the plasma (representing a mixture of fetal and maternal cfDNA); DNA can also be optionally extracted from white blood cells (only maternal DNA can be used for verification). DNA extraction can be performed using methods known in the art, such as Figure 7AExemplary methods are described below in the section titled Library Creation Methods; briefly, plasma is mixed with magnetic beads that bind to cfDNA, then a magnetic field is applied to concentrate the magnetic beads, which are then washed, isolated, and eluted. Other methods for isolating cfDNA are known in the art and can be used, e.g., centrifugal column-based methods, manual magnetic bead-based methods, and automated magnetic bead-based methods (see, e.g., Polatoglou et al., Diagnostics (Basel). 2022 Oct; 12(10): 2550; Bronkhorst et al., Tumour Biol. 2020 Apr; 42(4): 1010428320916314; Bronkhorst et al., Tumour Biol. 2019 Aug; 41(8): 1010428319866369; Michelson et al., Mitochondrion. 2023 Jul; 71: 26-39; Streleckiene et al., Biopreserv Biobank. 2019 Dec; 17(6): 553-561; Solassol et al., Clin Chem Lab Med. 2018 Aug 28; 56(9):e243-e246). Many kits are available for isolation, including QIAamp Circulating Nucleic Acids Kit (QiaM, 55114 Qiagen GmbH, Hilden, Germany), NucleoSpin Plasma XS (Macherey-Nagel 740900.50, High Sensitivity Protocol - MNaS, Macherey-Nagel GmbH, Düren, Germany), QIAmp MinElute cfDNA Mini Kit (QiaS, 55204, Qiagen GmbH, Hilden, Germany), cfPure Cell-Free DNA Extraction Kit (BChM, K5011610-BC, BioChain Inc., Newark, CA, USA), MagMAX Cell-Free DNA Isolation Kit (TFiM, A29319, Thermo Fisher Scientific, Waltham, MA, USA) and automated methods including MagNAPure 24 Total NA Isolation Kit (RocA, 07658036001, Roche Diagnostics GmbH, Penzberg, Germany), NextPrep-Mag cfDNA Isolation Kit (PerkinElmer), and cfNA ss2000 protocol on MagNA Pure 24 System (Roche Diagnostics). TM cfDNA automated isolation kits (PerkinElmer), and cfNA ss2000 protocol on MagNA Pure 24 System (Roche Diagnostics).

[0048] Preferably, adapters with universal primer sequences and unique molecular identifiers (UMIs) are ligated to the DNA to maximize sequence coverage, and the library is amplified using PCR. The method can then include an optional step of enriching the sample for fetal DNA, for example, by contacting the cfDNA with a plurality of oligonucleotides that bind to portions of fetal protein-coding genes, such as the TWIST targeted panel (Alliance Clinical Research Exome), optionally targeting all 22,995 genes (or 18,049 protein-coding genes, or a subset thereof) from the fetal genome, or a subset thereof, or other regions of the fetal genome that can be relevant to clinical interpretation or variant identification, or the method can include sequencing all nucleotides in the genome without exome capture (this method is referred to herein as NIFS, Genomic; NIFS-G). The UMI-labeled DNA (from the total cfDNA population, such as genomic DNA or exome-enriched DNA) is then sequenced using a high-throughput / next-generation sequencing method, preferably to an average sequencing depth of about 100X, 150X, or 200X. Filtered sequencing depth (i.e., after using the UMIs to filter out related reads) of at least 200, 250, or 300X in early and mid-pregnancy and at least 100X (but more preferably 200, 250, or 300X) in late pregnancy is preferred. See Figure 7B .

[0049] Sequencing can be performed using methods known in the art, including automated Sanger sequencing (e.g., using an ABI3730xl Genetic Analyzer), pyrosequencing on a solid support (e.g., using 454 sequencing, Roche), reverse-terminating sequencing-by-synthesis (e.g., using an Illumina®Genetic Analyzer), ligation sequencing or sequencing-by-synthesis with virtual terminators ​Molecular sequencing (see Voskoboynik et al. eLife 2013 2: e00569 and U.S. Patent Application No. 13 / 608,778, filed September 10, 2012); DNA nanoball sequencing; single molecule real-time (SMRT) sequencing; nanopore DNA sequencing; sequencing by hybridization; mass spectrometry sequencing; and microfluidic Sanger sequencing. Exemplary next generation sequencing methods known to those of skill in the art include massively parallel signature sequencing (MPSS), polony sequencing, pyrosequencing (454), Illumina (Solexa) sequencing by synthesis, SOLiD sequencing by ligation, Ion Torrent sequencing, DNA nanoball sequencing, chain termination sequencing (Sanger sequencing), heliscope single molecule sequencing, single molecule real-time (SMRT) sequencing (Pacific Biosciences); flow-based sequencing (e.g., Ultima sequencing) and nanopore sequencing such as described at the World Wide Web at nanoporetech.com.

[0050] Variant detection and identification from sequencing data is then performed using novel bioinformatics analysis methods to discover short variants (e.g., single nucleotide variants (SNVs) and insertions or deletions) and CNVs. Generally, data processing methods can be divided into three stages: alignment and preprocessing; variant filtering and genotyping; and variant interpretation. See, e.g., Van Loo et al. Nat. Methods 2010 7: 247-249; and Van Loo et al. Genome Biol. 2010 11 : R36. Figure 9 .

[0051] In the alignment and preprocessing stage, raw sequencing reads derived from cfDNA are aligned to a reference genome, grouped by UMI, and converted to single consensus reads for each UMI. The consensus reads are then realigned to the reference and optionally recalibrated for base quality scores to improve read quality, resulting in a set of aligned consensus reads that are prepared for downstream variant calling and analysis. Maternal genotyping can optionally be performed, and then the maternal genome or database can be used to filter germline variants. Genotyping data is optionally in variant call format (VCF), a text format for encoding genetic variant sites and genotypes.

[0052] Next, in the variant filtering and genotyping phase of the workflow, candidate variant sites are first identified by comparison to a reference genome (e.g., GRCh38 using Mutect2). Candidate variants can be filtered to remove potential false positive (FP) sites, e.g., using a hard filter set and a machine learning classifier such as a random forest-based model, support vector machine (SVM), or neural network trained on a subset of sites present in the sample. A probabilistic model is then applied to estimate fetal fraction, and fetal and / or maternal genotypes are assigned to all variant sites observed in the cfDNA sequencing data; e.g., a k-means or Bayesian mixture model can be used, e.g., to simultaneously estimate fetal fraction and assign fetal and maternal genotypes to each site.

[0053] In some embodiments, the probabilistic model uses a constrained 2D Bayesian Gaussian mixture model with five components, each representing a different combination of maternal and fetal genotypes for an autosomal variant, to simultaneously estimate fetal fraction and assign fetal and maternal genotypes to all variant sites observed in the cfDNA sequencing data. The combination is defined in two dimensions: variant allele fraction (VAF) and a fragment size rank sum statistic summarizing the difference in fragment sizes supporting the reference and alternate alleles, e.g., as described herein (e.g., in the section Variant calling of cfDNA with Mutect2); see, e.g., https: / / github.com / illumina / mutect2 / blob / master / src / main / java / com / illumina / mutect2 / VariantCaller.java#L 1 1 1 1 Figure 8 The centers of the cfDNA VAF clusters are determined by fetal fraction (FF). The model (shown in Figure 11 ) can be fitted using stochastic variational inference (e.g., using Pyro).

[0054] Table A shows the five components used in an exemplary Bayesian Gaussian mixture model.

[0055] Table A - Variant Allele Fraction

[0056] Fetal genotype Maternal genotype Expected VAF 0 / 1 0 / 0 ff / 2 0 / 0 0 / 1 (1 - ff) / 2 0 / 1 0 / 1 0.5 1 / 1 0 / 1 ff + (1 - ff) / 2 0 / 1 1 / 1 ff / 2 + (1 - ff) 1 / 1 1 / 1 1

[0057] ff = fetal fraction; VAF = variant allele fraction: variant allele reads / total reads

[0058] As shown in Figure 9 , incorporating fragment size and variant allele fraction (VAF) into the probabilistic model allows for accurate assignment of origin (maternal or fetal or both), e.g., based on assignment of one of the five components shown above.

[0059] Finally, in variant interpretation, all passed variants are annotated and evaluated to produce a list of clinically relevant variants for interpretation. Annotations can be made by reference to one or more databases, for example, variants can be annotated with gene and functional consequences (e.g., based on RefSeq 4 ), allele frequency (e.g., based on gnomAD v2.1.1 and gnomAD v3.0), Rare Exome Variant Ensemble Learner (REVEL) 16 score predicting the deleteriousness of each nucleotide change in the genome, ClinVar 17 annotations (updated 2023-04-30), and disease information for each gene such as inheritance type (e.g., recessive, e.g., based on Online Mendelian Inheritance in Man (OMIM, version 2022-07-08 18 ). Variants can be further filtered, for example, included if they have an allele frequency of <5 or are not reported in gnomAD v2.1.1 and gnomAD v3.0 6 , or excluded if determined to be likely benign or benign / likely benign in ClinVar, or synonymous. See, e.g., Figure 12 Exemplary data processing workflow.

[0060] This result can then be used to output a list from each sample for further review, preferably including all ClinVar annotated pathogenic / likely pathogenic variants, all frameshift / gain-of-stop variants, all predicted splice variants with a splice AI score 19 of >0.95, all non-frameshift variants of >15 amino acids, and all non-synonymous variants with a REVEL score of >0.7. This list can be shared with, e.g., a healthcare provider or with the mother.

[0061] Based on the presence of variants in the fetus that have the potential to have a deleterious impact on the health of the fetus or the mother (e.g., pathogenic or likely pathogenic variants, including those associated with adverse health conditions or outcomes), the method can further include recommending further testing, e.g., invasive testing such as amniocentesis or chorionic villus sampling (CVS), and / or further monitoring via ultrasound.

[0062] Based on the presence of variants in the fetus that have the potential to be deleterious, the method can further include recommending further testing, e.g., genetic testing to confirm the variants.

[0063] Standard computing devices and systems can be used and implemented to perform the methods described herein. Computing devices include various forms of digital computers, such as laptops, desktops, mobile devices, workstations, personal digital assistants, servers, blade servers, mainframes, and other appropriate computers. In some embodiments, the computing device is a mobile device, such as a personal digital assistant, a mobile telephone, a smartphone, a tablet computer, or other like computing device. The components of the computing device described herein, their connections and relationships, and their functions, are meant to be exemplary only, and are not meant to limit implementations of the applications described and / or claimed in this document.

[0064] Computing devices typically include one or more processors, memory, storage devices, one or more high-speed interfaces, and one or more low-speed buses connecting the various components. Each of the components are interconnected using various buses, and can be mounted on a common motherboard or in other manners as appropriate. The processor can process instructions for execution within the computing device, including instructions stored in the memory or on the storage device to display graphical information for a GUI on an external input / output device, such as a display coupled to the high-speed interface. In other implementations, multiple processors and / or multiple buses can be used, as appropriate, along with multiple memories and types of memory. Also, various other components can be connected to the computing device, such as a keyboard, a printer, and other input and output devices not shown.

[0065] Figure 13 An example computer system 500 is shown including a processor 510, a memory 520, a storage device 530, and an input / output device 540. Each of the components 510, 520, 530, and 540 can be connected to the other components over one or more buses 550. The processor 510 can process instructions for execution within the system 500. In some implementations, the processor 510 is a single-threaded processor, a multi-threaded processor, or another type of processor. The processor 510 can process instructions stored in the memory 520 or on the storage device 530. The memory 520 and the storage device 530 can store information within the system 500.

[0066] Input / output devices 540 provide input / output operations for system 500. In some embodiments, input / output devices 540 can include one or more of network interface devices (e.g., an Ethernet card), a serial communication device (e.g., an RS-232 port), or a wireless interface device (e.g., an 802.11 card, a 3G wireless modem, a 4G wireless modem, or a 5G wireless modem, or the like). In some embodiments, input / output devices can include driver devices configured to receive input data and send output data to other input / output devices, e.g., keyboard, printer and display devices 560. In some embodiments, mobile computing devices, mobile communication devices, and other devices can be used. In some embodiments, the present methods are performed using equipment including sequencing machines, e.g., Illumina sequencers.

[0067] Embodiments

[0068] The present invention is further described in the following examples, which do not limit the scope of the invention described in the claims.

[0069] Methods

[0070] The following methods were used in the examples below.

[0071] Sample collection

[0072] Samples were collected from the Brigham and Women’s Hospital (BWH) LIFECODES longitudinal biorepository 23 and the MAPing Pregnancy Research Biorepository at the Center for Fetal Medicine at BWH (Tables 1-2). This resource was designed to facilitate research on prenatal screening and diagnosis and to deepen the understanding of the genetic basis of fetal structural abnormalities. We collected samples from any gestational age, with initial technical development focused on late pregnancy, while the 14 samples carrying fetal abnormalities were primarily accrued from early and mid-pregnancy (Tables 3 and 12). Women enrolled at the time of a prenatal visit at any gestational age and a peripheral blood sample was collected in two Streck collection tubes (Streck, La Vista, NE) providing up to 20 mL of maternal blood.

[0073] A total of 51 samples were collected from 49 singleton pregnancies, 1 dizygotic twin pregnancy, and 1 monozygotic twin pregnancy. Sample collection spanned all three trimesters, and 11 samples had a matched confirmation sample for obtaining benchmarking data using fetal DNA obtained from cord blood or amniocentesis.

[0074] Library creation methods

[0075] After sample collection, we isolated cell-free DNA (cfDNA) from maternal serum using the "Double Spin Protocol 2" recommended by Streck. The precipitated maternal leukocytes were used to extract maternal genomic DNA from all samples. In order to perform extensive benchmarking of maternal variant discovery, we collected maternal germline DNA (gDNA) from 28 mothers and performed standard exome sequencing (ES) on the Broad Institute Genomics Platform (Table 3). We extracted DNA from the cfDNA portion of the isolated serum using the NextPrep MagcfDNA Isolation Kit (Cat. # NOVA-3825-03). We then determined the cfDNA fragment size and concentration via Tapestation (cfDNA tape, Agilent Technologies) and QuBit (Broad Range DNA, Agilent Technologies), respectively. To convert cfDNA into sequenceable fragments, we used a DNA sequencing kit from New England Biolabs (NEB) according to the manufacturer's protocol. of Ultra TMII DNA library preparation kit with the following modifications: 1) NEB adaptor and USER enzyme steps replaced with direct ligation of xGen Stubby Unique Molecular Identifier (UMI) adaptors ordered from integrated DNA technologies (IDT); and 2) NEB primers replaced with xGen dual index primers (IDT). Following adaptor ligation, 12 cycles of PCR were performed. Following this initial PCR amplification, libraries were multiplexed into up to 16 batches (up to 8 ug of total material) and subjected to exome capture using an Alpha Broad Exome bait from TWIST Bioscience targeting 194,202 exonic regions under the IDT xGen hybridization protocol. Briefly, multiplexed libraries were combined with human Cot DNA and xGen Blocker oligos and resuspended in hybridization buffer and bait following dehydration. Following a 4 hour incubation, bait hybridized libraries were combined with buffer resuspended streptavidin beads and subjected to multiple washes to remove any non-hybridized library followed by 15 rounds of post-capture PCR on beads. PCR amplified libraries were purified using SPRI beads and analyzed using a tapestation (D1000, Agilent Technologies) and QuBit (Broad Range DNA, Agilent Technologies) prior to multiplexing and sequencing on an Illumina NovaSeq. We were able to obtain additional material for benchmarking analysis in a subset of participants in the study, including fetal cord blood collected at delivery (n=7), paternal DNA (n=7), and an additional 4 cases of DNA extracted from cultured cells derived from amniocentesis (Table 3).

[0076] Sequence generation

[0077] cfDNA libraries were sequenced at the Broad Institute Genomics Platform on Illumina Novaseq S4 flow cells in pooled multiplexed sequencing runs. Our multiplexing strategy was designed to generate as many unique sequencing reads as possible while keeping the original sequence duplication rate (not considering UMI) below 75%. We noted that at a depth of 200x, assuming a fetal fraction of 25% (the median fetal fraction observed in our samples), each target was expected to have a mean coverage of approximately 50 reads of fetal origin. For 8 samples (MGB038, MGB039, MGB40, MGB016, MGB043, MGB046, MGB047, and MGB048), sequencing was performed across two S4 flow cells and raw sequencing reads were combined and processed together.

[0078] Overview of data processing workflow

[0079] Figure 2 An overview of the exemplary data processing workflow is given in FIG. 1. The workflow is divided into three main sections, which are briefly described in this section and described in more detail below. The first stage pre-processes the raw sequencing reads, constructs consensus reads for each UMI discovered in the sequencing dataset, and aligns these consensus reads to the reference genome. See “Alignment and pre-processing of cfDNA sequence data” for a more detailed description of the methods and tools. In the next processing stage, candidate variant call sites are generated from the aligned consensus reads using the somatic variant caller Mutect2 (see section “Variant detection in cfDNA”); then a machine learning-based method is used to train a classifier for the dataset of each sample to filter variant sites that can be artifacts (see section “cfDNA variant filtering”); then a Bayesian mixture model is used to simultaneously estimate fetal fraction and assign fetal and maternal genotypes to each variant site (“cfDNA genotyping”). Finally, a set of protocols is developed for annotating, interpreting, and managing variant sites to produce a list of potentially clinically relevant variants for each sample (“Variant determination”).

[0080] Alignment and pre-processing of cfDNA sequence data

[0081] The following pipeline is used to generate high-quality GRCh38-aligned cram files for variant calling. First, UMI are extracted from each read using the open-source fgbio ExtractUmisFromBam from Fulcrum Genomics (github.com / fulcrumgenomics / fgbio). Several subsequent steps are performed using the open-source Picard tools from the Broad Institute at MIT and Harvard (broadinstitute.github.io / picard / ), including sorting the data by query name using Picard SortSam. Illumina adapters are recognized and marked as Picard’s MarkIlluminaAdapters. Reads are then converted to FASTQ with Picard’s SamToFastq, aligned to the GRCh38 reference genome with the open-source BWA-MEM aligner 24 25 ​The PrintReads tool in the framework (gatk.broadinstitute.org) removes a small number of degenerate mapped reads with mapped read length less than 19 bp. The reads are then grouped by UMI with the fgbio tool GroupReadsByUmi. We create consensus duplex reads with the fgbio CallDuplexConsensusReads with the parameters --error-rate-pre-umi=45 --error-rate-post-umi=30 --min-input-base-quality=10 --min-reads=0. The consensus reads are filtered with the fgbio FilterConsensusReads with the parameters --min-reads 0 0 0 --max-read-error-rate 0.35 --max-base-error-rate 0.3 --min-base-quality 40, then clipped with the fgbio ClipBam with the parameters clipping-mode=Hard --clip-overlapping-reads=true. The mate information is fixed and the mate CIGAR tag is populated with Picard's FixMateInformation. The consensus reads are sorted by coordinates with Picard's SortSam. Finally, the base quality scores are recalibrated with the GATK BaseRecalibrator and ApplyBQSR tools. The metrics are then collected, and variant calling, filtering, and genotyping are applied to the on-target intervals in the Twist Broad Custom Exome Kit (Twist Alliance Clinical Research Exome). The publicly available version of the on-target data is available at: twistbioscience.com / resources / data-files / twist-alliance-clinical-research-exome-349-mb-bed-files.

[0083] The on-target gene count is calculated by intersecting the list of target intervals with the GRCh38 refSeq database downloaded from the UCSC Genome Browser (NCBI Annotation Release 110) and counting the number of genes that are covered by the target intervals. 26 27 28

[0084] Coverage analysis ​​

[0085] We applied the Picard tool CollectHsSequencingMetrics to collect coverage statistics in the whole exome targets based on aligned consensus reads. In Table 3, we report the mean coverage in all target intervals and the fraction of target bases with at least 50x coverage for each sample by cfDNA sequencing reads (of mixed maternal and fetal origin). We also multiplied each sample’s mean coverage per target indicator by the sample’s estimated fetal fraction to produce the percentage of exome target intervals with at least 8x and 10x mean estimated fetal read coverage, and report the two values for each sample in Table 4. Finally, in samples with matching paternal gDNA ES, we report the median and interquartile range of the number of reads supporting only the paternal allele in Table 7 “Genotyping performance and coverage based on paternal gDNA ES.” These values can be used to infer the distribution of coverage at these sites by sequencing reads derived from the fetal genome, where half are expected to support the paternal allele.

[0086] Variant site detection in cfDNA

[0087] We used the open source software tool Mutect2 from the Broad Institute of MIT and Harvard 27 We identified candidate variant sites with the argument --max-mnp-distance 0 to split MNP variants into separate records, and with the following arguments to generate annotations used in filtering: -G StandardAnnotation -G StandardHCAnnotation -A MappingQualityZero -A TandemRepeat -A CountNs. We generated additional annotations for genotyping by modifying GATK to add the InsertSizeRankSum annotation to each variant based on the fragment size of reads supporting the reference and alternate alleles at each site (see section cfDNA genotyping). To produce this annotation, the estimated fragment size distribution of reads supporting the reference allele was compared to the fragment size distribution of reads supporting the alternate allele using the Mann-Whitney U test (implemented by GATK’s RankSumTest). The value of the annotation is the Z-score of the U statistic. The fragment size of each read supporting the reference at each site determined by the calling engine based on Mutect2 / GATK integration was estimated based on the mapping insert fragment size reported by BWA in the BAM file for each read pair, and adjusted to account for insertions and deletions reported in the CIGAR and matching CIGAR of the reads.

[0088] cfDNA variant filtering

[0089] To remove potential false positive (FP) sites due to sequencing, library preparation, or alignment errors, a variant site filtering was developed, which includes hard filtering rules and a random forest-based classifier that assigns a score to each variant site reflecting the likelihood that the site is a true positive (TP) variant. The filtering rules are:

[0090] 1. Any site in the cfDNA sample is filtered if Mutect2 identifies more than one alternative allele and at least one of the alternative alleles is an insertion or deletion.

[0091] 2. A machine learning classifier (described in detail below) is applied to score variants, and any variant with a score below a cutoff (which is determined by assessing the sensitivity on a gold standard set of common variants) is filtered.

[0092] 3. Insertions or deletions that can be recurrent sequencing errors but still pass our random forest filtering are hard filtered based on a list of recurrent artificial insertion or deletion calls that we observed in our dataset. To construct this list, we identify in a subset of cfDNA samples from this study that do not have a matching cord blood or amniocentesis sample each insertion or deletion site with an allele count of at least 5 and that is not in a sample that has already been involved in a genetic test. From this resulting list, we remove any site that is present in the gnomAD v3.1.2 28 database at any allele frequency. The remaining sites are used to make a catalog consisting of 969 insertions or deletions that are recurrent artifacts in our data. We apply a filter to remove any insertion or deletion site that has a position and alternative allele that matches one of the sites in this catalog.

[0093] 4. Any site that is confidently called by Mutect2 and is in phase with a SNV site that does not pass one or more of the filters listed above is also filtered. Mutect2 calls certain sets of sites to be in phase with each other based on the number of reads that span more than one site in the set and support the same allele combination. The information is recorded in the variant’s phase set ID (PID) annotation. This filter captures sets of sites that represent mapping errors in clusters when reads originate from other paralog-specific variants in the genome that contain multiple paralogs.

[0094] In addition to the filters listed above, two filters are applied after genotyping (see section cfDNA genotyping) based on identifying variant sites with unexpectedly low counts of reads supporting alternate alleles.

[0095] The machine learning classifier described in step 2 above is trained using positive-unlabeled learning 29 The principle of the approach is built where only positive training labels are known for certain in the training dataset. We train a new instance of this classifier for each sample, using only data from that sample. Variants sites that are common in the population can be true, we assign an initial positive training label to gnomAD v3 28 sites that have a maximum sub-population frequency (as given by the AF_popmax annotation in gnomAD data) of at least 0.1. All other sites are initially assigned a negative training label. We then train a random forest with 800 estimators implemented through the scikit-learn package. After training the classifier, we then score each variant according to the classifier’s predicted probability that the site is truly positive. We then identify a cutoff for this score for the PASS filter state using the GATK’s FilterVariantTranches tool, which finds the optimal cutoff that results in the requested estimated sensitivity based on a common set of SNPs and insertions or deletions provided as a resource with the best practices pipeline. In our pipeline, we request a sensitivity of --snp-tranche 99.5 and --indel-tranche 95.0. The following features (selected to be independent of allele fraction and either generated by Mutect2 or added based on site coordinates from the genomic context) are selected for evaluation in the random forest:

[0096] • Indel: binary feature indicating whether the variant is a SNP or an insertion or deletion

[0097] • SOR: Strand-odds-ratio, a strand bias test statistic

[0098] • MQ: Mean Quality of reads

[0099] • MQRankSum: Rank Sum Test for mapping quality bias

[0100] • ReadPosRankSum: Rank Sum Test for read position bias

[0101] • BaseQRankSum: Test for base quality score bias between reference and alternate alleles

[0102] • MPOS: Median distance of the site from the end of the reads

[0103] • ECNT: Number of events in the assembled haplotype containing the variant

[0104] • NCount: Number of reads in the pileup with N base calls at the variant site (created in the formation of duplex consensus reads)

[0105] • DP: Depth of the variant site

[0106] • SEGDUP: Binary feature indicating whether the site lies within a segmental duplication

[0107] • LCR: Binary feature indicating whether the site lies within a low complexity region as defined by the LCR-hs38 resource provided by Li et al. 30

[0108] • SIMPLEREP: Binary feature indicating whether the site lies within an annotated simple repeat • STR: Binary feature indicating whether GATK / Mutect classified the site as falling within a short tandem repeat sequence.

[0109] It should be noted that the fact that a variant is observed in a region of the genome that is repetitive (as annotated by SEGDUP, LCR, SIMPLEREP, and STR) is used as a feature for training in the classifier, rather than as a hard filter, for the purpose of allowing the classifier to make confident calls in those regions of the genome.

[0110] Estimation of cfDNA genotypes and fetal fraction

[0111] ​We developed a machine learning-based model that simultaneously estimates fetal fraction and assigns fetal and maternal genotypes to all variant sites observed in cfDNA sequencing data. Our model consists of a constrained Bayesian Gaussian mixture model with five components, each representing a different combination of maternal and fetal genotypes for autosomal variants. The mixture is defined in two dimensions: variant allele fraction, described in the chapter on variant calling in cfDNA, and the fragment size rank sum statistic, which summarizes the difference in fragment size supporting the reference and alternative alleles. We model the fetal fraction of a sample as a latent variable (f) and base the mean of the variant allele fraction distribution of each component on it as follows: if we let 0 / 0 represent a homozygous reference genotype, 0 / 1 a heterozygous genotype, and 1 / 1 a homozygous alternative genotype, then the components and their means are defined as: (“cluster 0”: fetal 0 / 1, maternal 0 / 0) f / 2; (“cluster 1: fetal 0 / 0, maternal 0 / 1”) (1-f) / 2; (“cluster 2: fetal 0 / 1, maternal 0 / 1”) 0.5; (“cluster 3: fetal 1 / 1, maternal 0 / 1”) f + (1-f) / 2; (“cluster 4: fetal 0 / 1, maternal 1 / 1”) 1-(f / 2). Each data dimension is modeled independently, i.e., the covariance matrix of each component is diagonal. Before inferring the parameters of the model, we remove a subset of sites that appear to be outliers, including sites that do not pass filtering status (as set by the filtering procedure described above in the chapter on variant calling in cfDNA), sites with cfDNA VAF less than 0.025 or greater than 0.975, and sites with fragment size statistic missing less than -4 or greater than 4. To further clean the data, we remove all sites that do not pass outlier detection for cfDNA VAF and fragment size statistic. Outlier detection is implemented by fitting an IsolationForest outlier classifier from the sklearn.ensemble package to the data with a contamination parameter of 0.05. We run the model in Pyro 31We define a genotype calling mixture model and fit it to the data for each sample using stochastic variational inference. We use the AutoDelta guide functionality of Pyro to find the maximum a posteriori values for each parameter. To initialize the model, we first produce an initial estimate of the fetal fraction. We do this by identifying the location of the cluster of sites ("cluster 4") in the VAF distribution that represents maternal homozygous variants and heterozygous sites in the fetus. We initialize the fetal fraction by computing a Gaussian kernel density estimate of all sites with a VAF less than 0.975 and using the scipy.signal.argrelextrema function to identify the peak in the density with the maximum value (corresponding to cluster 4). To estimate an initial value for the mean of the fragment size statistic distribution, we find the 500 sites with the estimated fetal fraction whose cfDNA VAF is closest to the expected VAF of cluster 4 and use their median fragment size statistic value. Once we estimate the mean of the fragment size statistic distribution for maternal homozygous variants / fetal heterozygous sites, we initialize the means of the other fragment size component distributions by multiplying this value by the vector [-1.0, 0.5, 0.0, -0.5, 1.0] to match the expected relative contribution of maternal and fetal reads observed for sites in each cluster.

[0112] After fitting the model parameters using stochastic variational inference, we re-add all sites filtered from the above model to the dataset and solve for the optimal cluster assignment parameters for each autosomal site by exhaustively enumerating all potential variables using Pyro's discrete latent variable enumeration strategy with a guide function that fixes the learned model parameters but allows the assignment probabilities to vary. We then estimate the likelihood of each possible fetal genotype by summing the cluster component assignment probabilities: the likelihood of a 0 / 0 (reference / reference) fetal genotype at this site is the probability of assigning this site to cluster 1; the likelihood of a 0 / 1 (reference / alternate) fetal genotype is the sum of the assignment probabilities for clusters 0, 2, and 4; and the likelihood of a 1 / 1 (alternate / alternate) fetal genotype is the assignment probability for cluster 3. Sites in the cfDNA sample that appear homozygous alternate (i.e., where the VAF is greater than 0.975) are automatically assigned the homozygous alternate genotype. Similarly, the maternal genotype likelihoods are set as follows: the likelihood of a maternal 0 / 0 genotype is set to the assignment probability for cluster 0; the likelihood of a maternal 0 / 1 genotype is set to the sum of the assignment probabilities for clusters 1, 2, and 3; and the likelihood of a maternal 1 / 1 genotype is set to the assignment probability for cluster 4.

[0113] We applied this model to all autosomal variants in each sample, as well as to variants on the X chromosome in samples where the fetal sex chromosome ploidy was predicted to be XX. For samples with a predicted XY fetal sex chromosome ploidy, we used this model to genotype only variants within the pseudoautosomal region (PAR) of the X chromosome. For X chromosome variants in XY samples outside of the PAR, we defined three Gaussian components for each pair of possible maternal and fetal genotypes (excluding homozygous variants in both the mother and fetus). We defined these components based on the parameters learned in the training autosomal model as follows: for the cluster representing maternal heterozygous variants where the fetus carries the variant, the VAF mean was set to 1 / (2-f); for the cluster representing maternal heterozygous variants where the fetus does not carry the variant, the VAF mean was set to (1-f) / (2-f); and the third cluster, representing homozygous reference and variants in the fetus (i.e., de novo mutations), had a VAF mean of f / (2-f). The fragment size mean for these clusters was set to the mean learned in the autosomal model for clusters 1, 3, and 0, respectively, with a variance equal to the fragment size variance from the autosomal cluster 0 multiplied by 5 (to account for the additional variation observed at these sites). We assigned genotypes to these variants by computing the likelihood that each variant was generated from each of these Gaussian components and assigning the variant to the cluster accordingly.

[0114] After genotyping, we applied two additional filters to the resulting variant calls. First, we filtered out calls where the variant allele fraction was too low to be generated by the cluster representing variants where the fetus is heterozygous and the mother is homozygous reference (cluster 0). To do this, we performed a low-tailed binomial distribution test on the number of reads supporting the alternative allele out of the total depth observed at the site, with a binomial probability of f / 2, the expected VAF for this cluster, and filtered out any sites where the p-value of this test was less than le-5. Second, we filtered out any insertion or deletion calls where the alternative allele was supported by three or fewer reads, as we found the error rate to be high in these variants.

[0115] Real data processing: Standard ES of gDNA variant calls from maternal, paternal, fetal cord blood, and amniocentesis samples

[0116] gDNA libraries were prepared from maternal, paternal, fetal cord blood, and amniocentesis samples following the standard ES protocol of the Broad Institute Genomics Platform (Cambridge, MA). After Illumina sequencing, reads were aligned, and variant calls were made following the GATK Best Practices Guide 25 Variant calling. Briefly, after marking and trimming of the adapter sequences, BWA-MEM with default parameters was used to align the reads to the human genome (hg19) 24Pre-processed reads were aligned to the human reference. Repeated reads were marked using Picard MarkDuplicates and excluded from downstream analysis. Base recalibration was performed using GATK BaseRecalibrator and ApplyBQSR (using known variant sites from the GATK reference bundle). Germline single nucleotide variants (SNVs) and indels were called for each sample using GATK HaplotypeCaller in GVCF mode, followed by joint genotyping of all maternal and fetal DNA-derived samples and variant filtering with GATK VQSR. To ensure a high-quality set of genotypes for benchmarking, we further applied a stringent set of variant filters previously used for a large-scale family sequencing project 32 Briefly, if a variant site overlapped a low-complexity region of the genome, it was removed; variant genotypes satisfying any of the following criteria were filtered: depth less than 10; allele balance <0.25 or >0.75; probability of allele balance (based on a binomial distribution with mean 0.5) lower than le-9; or fewer than 90 reads were reference informative for the genotype. For amniocentesis sample of study participant MGB043, sequencing was performed at Boston Children's Hospital (Boston, MA) using a protocol from GeneDx (Stamford, CT). Sequencing data from this sample was re-aligned to hg38 and then re-processed according to the informatics steps listed above; for this sample alone, we restricted benchmarking evaluation to the intersection of the Broad custom exome kit and the exome target regions of the GeneDx kit.

[0117] Benchmarking and evaluation

[0118] Variants were compared to "true" genotype data from ES derived from matched cord blood, amniocentesis, maternal DNA collected from white blood cells, or paternal samples (see section gDNA ES variant calling in maternal, paternal, fetal cord, and amniocentesis samples). For cfDNA variant comparisons to cord blood or amniocentesis, we performed five sets of evaluations, reported in Tables 5, 6, 8, and 10:

[0119] • Site-level comparison of variants that were not removed by our filtering methods (see section "cfDNA variant filtering") that do not take into account the fetal genotype at the site (Table 10, "Variant calling after filtering variants"). This evaluation provides an estimate of the sensitivity limit of cfDNA sequencing at the depth used in this study after attempting to remove sequencing artifacts and other errors from the sequencing data. As with the filtered variant calling evaluation below, we exclude from this evaluation maternal variants that were not transmitted to the fetus, so that the PPV metric shows the ability of this method to distinguish errors from true biological variants.

[0120] • Site-level comparison of all variant sites detected in the cfDNA sequencing data (so of maternal or fetal origin, or both), without taking into account the final filtering status assigned to the site by our bioinformatics pipeline or the fetal genotype (also in Table 10 "Unfiltered variant calling"). This evaluation provides an estimate of the theoretical limit of sensitivity of cfDNA sequencing at the depth used in this study, without attempting to remove sequencing artifacts and other errors. We exclude from this evaluation any sites that are present in the mother but not transmitted to the fetus (according to maternal and cord blood or amniocentesis gDNA ES data), so that FPs in this evaluation are expected to represent true sequencing or mapping errors, rather than failures in fetal / maternal genotyping.

[0121] • Comparison of all fetal genotypes assigned by our model to all genotypes called in the cord blood or amniocentesis ES data (Table 5, "Overall genotyping performance"). This evaluation assesses the accuracy of our genotyping model that attempts to assign fetal and maternal genotypes to all sites detected in the cfDNA, which is a mixture of cfDNA fragments with maternal and fetal origin. See the Supplementary Methods section cfDNA genotyping for a description of the genotyping model. In contrast to the "unfiltered variant calling" and "variant calling after filtering variants" evaluations, maternal variants that are not transmitted are included. The results of this evaluation are representative of the ability of our informatic method to determine the fetal genotype at each site in the exome given only cfDNA sequencing samples.

[0122] • Comparison of all variants assigned by this participant a fetal heterozygous genotype and a maternal homozygous reference genotype (Table 6, "Predicted paternal or de novo variant calling") to sites that are present in the cord blood or amniocentesis gDNA ES data but not in the maternal gDNA ES data. In this evaluation, we exclude from the evaluation any variant sites detected in the maternal gDNA ES data, and only evaluate variants called in the NIFS data that have been assigned to the "fetal 0 / 1; maternal 0 / 0" cluster by the genotyping pipeline. This evaluation characterizes the accuracy of this method in detecting paternal genetic variants as well as de novo mutations.

[0123] • Assess our method's ability to accurately genotype variants that are heterozygous in the mother (Table 8, "NIFS genotype accuracy for heterozygous variants in the mother"). This evaluation focuses only on sites where maternal gDNA ES data indicates that the mother is heterozygous for the variant with a pass filter status. These sites are important for recessive disease diagnosis, but are more difficult to genotype with low fetal fraction. For this evaluation, we report a single accuracy metric, which is the percentage of true maternal heterozygous sites assigned by NIFS a pass filter status and the correct fetal genotype.

[0124] All of the above evaluations, except "NIFS genotype accuracy for heterozygous variants in the mother", were performed with the vcfeval tool from RealTime Genomics 33,34 (RTG; realtimegenomics.com / products / rtg-tools), which performs haplotype-based analysis to match variants between samples, and is a widely accepted standard for genomic variant call evaluation. All benchmarking analysis was restricted to intervals targeted by the exome capture on autosomes. The "unfiltered variant detection" and "after filtering variant detection" evaluations comparing to cord blood and amniocentesis samples were performed by matching sites, regardless of called genotype. In both evaluations, the presence of the same variant site matched on genomic position and alternate allele in both cfDNA sample and confirmation data was considered a true positive (this was implemented using the vcfeval parameter— squash-ploidy). On the other hand, "overall genotyping performance" comparisons require that each called fetal allele in the output of our pipeline matches an allele present in the called genotype in the confirmation data. The benchmarking evaluation true positives (TP), false positives (FP), true negatives (TN), and false negatives (FN). Sensitivity and PPV were computed by RTG vcfeval as:

[0125] • PPV = TP / (TP + FP)

[0126] • Sensitivity = TP / (TP + FN)

[0127] For all of the above evaluations, except for "Genotype accuracy of heterozygous variants in mothers", we excluded from the evaluation all regions where the ES data from the cord blood or amniocentesis sample had less than 10 reads of coverage - in other words, called variant sites in these regions were not considered TP, FP, or FN. This evaluation was performed using the same analysis script as described below for "Genotype accuracy of maternal and fetal genotypes" reported in Table 9. We note that the metrics present in this evaluation can be computed by summarizing the results in Table 9 corresponding to each genotype cluster for maternal heterozygous variants. See also Figure 3A - C.

[0128] In addition to the above described analyses, we also used the matched cord blood or amniocentesis gDNA ES data to break down the sensitivity and genotype accuracy of NIFS for all confirmed variants in the fetal and maternal exomes in more detail (Table 9, "Genotype accuracy of maternal and fetal genotypes"). For this evaluation, we compared the union set of all NIFS calls made from cfDNA to all variants called in maternal gDNA ES or cord blood / amniocentesis gDNA ES. For each combination of maternal and fetal genotypes present in the comparison set for maternal and fetal variants, we computed the percentage of sites present in the raw Mutect2 cfDNA VCF with matching position and alternate allele (as reported in Table 10) and the percentage of those sites that were unfiltered and assigned the correct fetal genotype by the cfDNA variant calling pipeline. These evaluations were performed with a custom analysis script that matched variant calls in maternal gDNA ES, cord blood or amniocentesis gDNA ES, and cfDNA sequencing data by genomic position and alternate allele (as opposed to the haplotype-based approach implemented in RTG vcfeval).

[0129] The second set of evaluations compared the maternal genotypes predicted by our model to the variants detected in ES sequencing of maternal gDNA extracted from pelleted maternal white blood cells. The results of this evaluation are reported in two parts, "Detection of maternal variants" and "Maternal genotyping performance" in Table 11. We allowed any called variant site to match the "Detection of maternal variants" comparison, regardless of the maternal genotype assigned by NIFS or gDNA variant calling. We required a perfect genotype match between gDNA and ES calls for the "Maternal genotyping performance" evaluation. For these maternal evaluations, we excluded any sites where the maternal gDNA ES data had less than 10x read coverage. These evaluations were performed using the RTG vcfeval tool.

[0130] Finally, for participants with matching gDNA ES data from the parental blood samples, we performed an evaluation of the proportion of sites assigned non-reference fetal genotypes in the cfDNA data, excluding sites present in the maternal gDNA ES data that were present in the paternal gDNA ES data. The results of this evaluation are reported in Table 7, “Genotyping performance and coverage based on paternal gDNA ES.” This evaluation is another way of calculating the PPV of NIFS calls predicted to be paternal inheritance or de novo mutations, in addition to the “Predicted paternal or de novo variant calls” results reported in Table 6. For this analysis, we used RTG vcfeval to calculate the PPV of all calls assigned to cluster 0 (representing clusters of fetal heterozygous and maternal homozygous reference variants) against the paternal ES variant set, and we restricted the evaluation to sites that did not match variant sites in the maternal ES data. We excluded any regions from this evaluation where the paternal ES data had less than 10x read coverage. We also report the number of reads supporting the alternate allele for each of the confirmed paternal variants detected by NIFS.

[0131] For sample MGB043, the amniocentesis sample was sequenced at Boston Children’s Hospital using a different exome capture kit provided by GeneDx, and we therefore restricted all evaluations to the set of exome target intervals covered by both the Twist custom exome list used for NIFS samples and the GeneDx exome targets (n = 194,202 intervals).

[0132] Family relationship inference

[0133] Predicted genetic relationships (cfDNA, parents, and between cord blood and amniocentesis samples) after variant calling with KING 35 Validation. To validate the suspicious family relationships in our cohort, we filtered the cfDNA variants to include only those with a gnomAD allele frequency (AF_popmax) greater than 0.05 and a quality score for fetal genotype inference greater than 10. Processing the resulting predicted genotypes with KING (parameters related - degree 2) verified that the expected relationships had an estimated proportion of identically by descent genomes (KING metric PropIBD) of at least 0.4 (with one exception, a paternal-fetal pair with a PropIBD of 0.32, which we manually confirmed).

[0134] Detection of copy number variants (CNVs)

[0135] We developed a sliding window binning approach to investigate significant deviations in copy state using coverage collected from GATK CollectReadCounts with GC correction. We normalized copy states against a subset of control NIFS libraries (without cases of fetal abnormality, Table 12) using GATK CreateReadCountPanelOfNormals and DenoiseReadCounts. We filtered out highly variable capture intervals in control cfDNA samples with median absolute deviation (MAD) greater than the 3rd quartile + 1.5*interquartile range (IQR). We then computed the median copy fraction for each sample in bins representing a sliding window of 3MB in size with 100kb offsets across the genome. We applied a final filtering step to remove 1MB bins with >10 control samples that were classified as outliers based on IQR analysis for each bin. We only observed one validated CNV event in our study cohort, so we could not perform extensive benchmarking or sensitivity analysis for CNVs. We note that our previous gDNA ES study, as reported by Fu et al 32 has demonstrated accurate CNV discovery beyond single gene resolution—up to routinely discovered events spanning >2 exons—and has noted the potential for discovery of CNVs at single exon resolution. Detection of these events in cfDNA would be difficult due to the mixture of maternal and fetal DNA, but more data will allow the development of improved methods and thorough benchmarking.

[0136] Gender determination

[0137] Given the strong coverage of chrY and chrX, we explored the ability of NIFS to determine fetal gender. Given that any reads on chrY should indicate male fetal gender, except for some artifacts, we initially focused on gender calling on chrY. In fact, the presence of any coverage (from GATK CollectReadCounts) on the chrY binning intervals was highly discriminative for gender determination Figure 4 ), although accurate prediction of chrY copy state by dividing the median coverage in all intervals on chrY by the fetal fraction was still challenging due to the relatively low and variable coverage on chrY compared to the rest of the genome.

[0138] Variant classification

[0139] We used genotype calls derived from cfDNA results to analyze potential pathogenic variations in each sample in the fetus and the mother. We applied bcftools 36Merge tools to create a multi-sample VCF of all samples with cffDNA sequencing. Use ANNOVAR 37 and bcftools, with gene and functional results (RefSeq 26 ), allele frequencies (gnomAD v2.1.1 and gnomAD v3.0), REVEL 38 scores, ClinVar 39 annotations (updated 2023-04-30), and each gene disease information such as inheritance pattern (e.g. recessive) from the Online Mendelian Inheritance in Man (OMIM, version 2022-07-08 40 ) annotate the merged VCF. If a variant has an allele frequency <5 or is not reported in gnomAD v2.1.1 and gnomAD v3.0 28 we include the variant, and exclude synonymous variants. We then create a list from each sample for further review, including all ClinVar annotated pathogenic / likely pathogenic variants, all frameshift / gain-of-stop variants, all predicted splice variants with a splice AI score 41 >0.95, all non-frameshift variants >15 amino acids, and all non-synonymous variants with a REVEL score >0.7. In addition to ClinVar P / LP variants, remove variants that do not pass filtering. In this set, filter variants with <4 alternate reads and those determined as likely benign or benign / likely benign in ClinVar.

[0140] We determined fetal genotypes using the methods described above, with the note that for a small subset of insertions or deletions phased with high-quality SNVs, we used the SNV genotypes given the higher SNV genotype accuracy. Variants from disease genes from OMIM were selected for further analysis. We manually inspected each remaining variant using the Integrative Genomics Viewer (IGV) and removed variants that appeared low quality or were present in multiple NIFS samples (indicating they were likely technical artifacts). Variants were checked for pathogenicity and assessed for clinical relevance based on ACMG criteria 42-44 as described by Riggs et al. 45Clingen and ACMG guidelines were provided to assess CNVs. We assessed 28 samples for potential carrier variants using matched maternal germline exome sequencing data (Table 3), which we further filtered based on their ClinVar pathogenicity. Variants were considered if they were listed as pathogenic or likely pathogenic in ClinVar with a clinical significance corresponding to 2 or more gold stars (i.e. practice guidelines, reviewed by expert panel, or with multiple submitters providing and no conflicts of interest). Variants with genotypes corresponding to maternal carrier status were selected. As before, the potential clinical relevance of the variants was examined (Table 14). All identified variants were confirmed by maternal germline ES.

[0141] Table 1. Characteristics of study samples

[0142]

[0143] IQR - Interquartile Range

[0144] Table 2. Representation of study participants

[0145]

[0146] Table 3. Sample information

[0147]

[0148]

[0149] Table 4. Sample coverage and sequencing metrics

[0150]

[0151]

[0152] Table 5. Overall genotyping performance

[0153]

[0154]

[0155] * Evaluation of fetal genotype assigned to each locus by NIFS compared to genotype of the validation sample.

[0156] All loci in the exome target regions with sufficient depth in the validation sample were considered.

[0157] Table 6. Predicted paternal or de novo variant detection

[0158]

[0159]

[0160] For sites present in the confirmation sample and not present in the maternal gDNA sample, evaluate all sites that NIFS predicted to be heterozygous in the fetus and not present in the mother.

[0161] Exclude regions with less than 10x coverage in the confirmation sample.

[0162] Bold indicates values highlighted in text.

[0163] Table 7. Genotyping performance and coverage based on paternal gDNA ES

[0164]

[0165]

[0166] Table 8. Genotyping accuracy of maternal heterozygous variants

[0167]

[0168] Evaluate the accuracy of fetal genotypes assigned by NIFS for all sites that are heterozygous in the mother (as determined by maternal gDNA ES).

[0169] Bold indicates values highlighted in text.

[0170] Table 9. Genotyping accuracy of maternal and fetal genotypes

[0171]

[0172]

[0173]

[0174]

[0175] Table 10. Fetal site-level variant detection

[0176]

[0177]

[0178] # Evaluate the presence or absence of all sites in the confirmation sample that are on the filtered list of sites detected by NIFS, regardless of genotype. Exclude sites that are only maternal (as determined by the confirmation sample and maternal gDNA sample) and regions with less than 10x coverage in the confirmation sample from the evaluation.

[0179] * All sites in the unfiltered list of sites detected by NIFS were evaluated for the presence or absence of the site in the confirmatory sample, regardless of genotype. Sites that were only maternal (as determined by the confirmatory sample and maternal gDNA sample) and regions with less than 10x coverage in the confirmatory sample were excluded from evaluation.

[0180] Bold indicates values highlighted in text.

[0181] Table 11. Maternal variant detection and genotyping performance for germline maternal ES

[0182]

[0183]

[0184] * Maternal accuracy was significantly reduced when fetal fraction was close to 50% because the maternal and fetal unique allele fraction was equivalent. Genotype accuracy was calculated by comparing the maternal genotype assigned at each site by NIFS to the genotyping from the maternal gDNA ES.

[0185] Table 12. Clinical information for samples

[0186]

[0187]

[0188] * Excluded from clinical evaluation because patient never received follow-up testing

[0189] Table 13. Clinically relevant variants

[0190]

[0191]

[0192] * Paternal genotype derived from a separate collection of DNA that underwent exome sequencing; see Table 1

[0193] # Twin loss detected by NIFS during sex inference; see Figure 4

[0194] & Note that these breakpoints are the minimum breakpoints defined by the missing exons identified

[0195] Table 14. Maternal carrier variants

[0196]

[0197]

[0198] Note, 16 / 28 (57.1%) samples had at least one maternal carrier variant.

[0199] References:

[0200] 1. Lowther C, Valkanas E, Giordano JL, et al. Systematic evaluation of genome sequencing for the assessment of fetal structural anomalies [Internet]. bioRxiv. 2020; Available from:

[0201] biorxiv.org / content / 10.1101 / 2020.08.12.248526.abstract

[0202] 2. Talkowski ME, Ordulu Z, Pillalamarri V, et al. Clinical diagnosis by whole-genome sequencing of a prenatal sample. N Engl J Med 2012;367(23):2226-32.

[0203] 3. Tolusso LK, Hazelton P, Wong B, Swarr DT. Beyond diagnostic yield: prenatal exome sequencing results in maternal, neonatal, and familial clinical management changes. Genet Med 2021;23(5):909-17.

[0204] 4. Gregg AR, Skotko BG, Benkendorf JL, et al. Noninvasive prenatal screening for fetal aneuploidy, 2016 update: a position statement of the American College of Medical Genetics and Genomics. Genet Med 2016;18(10):1056-65.

[0205] 5. American College of Obstetricians and Gynecologists’ Committee on Practice Bulletins—Obstetrics, Committee on Genetics, Society for Maternal-Fetal Medicine. Screening for Fetal Chromosomal Abnormalities: ACOG Practice Bulletin, Number 226. Obstet Gynecol 2020; 136(4):e48-69.

[0206] 6. Bianchi DW, Parker RL, Wentworth J, et al. DNA sequencing versus standard prenatal aneuploidy screening. N Engl J Med 2014; 370(9): 799-808.

[0207] 7. Norton ME, Jacobsson B, Swamy GK, et al. Cell-free DNA analysis for noninvasive examination of trisomy. N Engl J Med 2015; 372(17): 1589-97.

[0208] 8. Yatsenko SA, Peters DG, Saller DN, Chu T, Clemens M, Rajkovic A. Maternal cell-free DNA-based screening for fetal microdeletion and the importance of careful diagnostic follow-up. Genet Med 2015; 17(10): 836-8.

[0209] 9. Zhang J, Li J, Saucier JB, et al. Non-invasive prenatal sequencing for multiple Mendelian monogenic disorders using circulating cell-free fetal DNA. Nat Med 2019;25(3):439-47.

[0210] 10. Breveglieri G, D’Aversa E, Finotti A, Borgatti M. Non-invasive Prenatal Testing Using Fetal DNA. Mol Diagn Ther 2019;23(2):291-9.

[0211] 11. Dungan JS, Klugman S, Darilek S, et al. Noninvasive prenatal screening (NIPS) for fetal chromosome abnormalities in a general-risk population: An evidence-based clinical guideline of the American College of Medical Genetics and Genomics (ACMG). Genet Med 2023;25(2):100336.

[0212] 12. Rose NC, Barrie ES, Malinowski J, et al. Systematic evidence-based review: The application of noninvasive prenatal screening using cell-free DNA in general-risk pregnancies. Genet Med 2022;24(7):1379-91.

[0213] 13. Fan HC, Gu W, Wang J, Blumenfeld YJ, El-Sayed YY, Quake SR. Non-invasive prenatal measurement of the fetal genome. Nature 2012;487(7407):320-4.

[0214] 14. Provenzano A, Farina A, Seidenari A, et al. Prenatal Noninvasive Trio-WES in a Case of Pregnancy-Related Liver Disorder. Diagnostics (Basel) [Internet] 2021; 11(10). Available from: dx.doi.org / 10.3390 / diagnostics11101904

[0215] 15. Filer DL, Mieczkowski PA, Brandt A, et al. Noninvasive prenatal exome sequencing diagnostic utility limited by sequencing depth and fetal fraction. Prenat Diagn [Internet] 2021; Available from: dx.doi.org / 10.1002 / pd.6009

[0216] 16. Guo MH, Gregg AR. Estimating yields of prenatal carrier screening and implications for design of expanded carrier screening panels. Genet Med 2019;21(9):1940-7.

[0217] 17. Ben-Shachar R, Svenson A, Goldberg JD, Muzzey D. A data-driven evaluation of the size and content of expanded carrier screening panels. Genet Med 2019;21(9):1931-9.

[0218] 18. Saunders CJ, Miller NA, Soden SE, et al. Rapid whole-genome sequencing for genetic disease diagnosis in neonatal intensive care units. Sci Transl Med 2012;4(154): 154ra135.

[0219] 19. Balloux F, Brynildsrud O, van Dorp L, et al. From Theory to Practice: Translating Whole- Genome Sequencing (WGS) into the Clinic. Trends Microbiol 2018;26(12): 1035-48.

[0220] 20. Miller DT, Lee K, Abul-Husn NS, et al. ACMG SF v3.1 list for reporting of secondary findings in clinical exome and genome sequencing: A policy statement of the American College of Medical Genetics and Genomics (ACMG). Genet Med 2022;24(7): 1407-14.

[0221] 21. Monaghan KG, Leach NT, Pekarek D, Prasad P, Rose NC, ACMG Professional Practice and Guidelines Committee. The use of fetal exome sequencing in prenatal diagnosis: a points to consider document of the American College of Medical Genetics and Genomics (ACMG). Genet Med 2020;22(4):675-80.

[0222] 22. Van den Veyver IB, Chandler N, Wilkins-Haug LE, Wapner RJ, Chitty LS, ISPD Board of Directors. International Society for Prenatal Diagnosis Updated Position Statement on the use of genome-wide sequencing for prenatal diagnosis. Prenat Diagn 2022;42(6):796-803.

[0223] 23. McElrath TF, Lim K-H, Pare E, et al. Longitudinal evaluation of predictive value for preeclampsia of circulating angiogenic factors through pregnancy. Am J Obstet Gynecol 2012;207(5):407.e1-7.

[0224] 24. Li H. Aligning sequence reads, clone sequences and assembly contigs with BWA-MEM [Internet]. arXiv [q-bio.GN]. 2013 [cited 2019 Nov 21] ; Available from: arxiv.org / abs / 1303.3997

[0225] 25. Poplin R, Ruano-Rubio V, DePristo MA, et al. Scaling accurate genetic variant discovery to tens of thousands of samples [Internet]. bioRxiv. 2018 [cited 2019 Nov 21] ; 201178. Available from: biorxiv.org / content / 10.1101 / 201178v3.abstract

[0226] 26. O’Leary NA, Wright MW, Brister JR, et al. Reference sequence (RefSeq) database at NCBI: current status, taxonomic expansion, and functional annotation. Nucleic Acids Res 2016;44(D1):D733-45.

[0227] 27. Benjamin D, Sato T, Cibulskis K, Getz G, Stewart C, Lichtenstein L. Calling Somatic SNVs and Indels with Mutect2 [Internet]. bioRxiv. 2019 [cited 2022 Apr 12]; 861054. Available from: biorxiv.org / content / 10.1101 / 861054vl

[0228] 28. Karczewski KJ, Francioli LC, Tiao G, et al. The mutational constraint spectrum quantified from variation in 141,456 humans. Nature 2020; 581(7809):434-43.

[0229] 29. Bekker J, Davis J. Learning from positive and unlabeled data: a survey. Mach Learn 2020; 109(4):719-60.

[0230] 30. Li H. Toward better understanding of artifacts in variant calling from high-coverage samples. Bioinformatics 2014; 30(20):2843-51.

[0231] 31. Bingham E, Chen JP, Jankowiak M, et al. Pyro: Deep Universal Probabilistic Programming. J Mach Learn Res 2019;20(28):1-6.

[0232] 32. Fu JM, Satterstrom FK, Peng M, et al. Rare coding variation provides insight into the genetic architecture and phenotypic context of autism. Nat Genet 2022;54(9):1320-31.

[0233] 33. Cleary JG, Braithwaite R, Gaastra K, et al. Joint variant and de novo mutation identification on pedigrees from high-throughput sequencing data. J Comput Biol 2014;21(6):405-19.

[0234] 34. Cleary JG, Braithwaite R, Gaastra K, et al. Comparing Variant Call Files for Performance Benchmarking of Next-Generation Sequencing Variant Calling Pipelines [Internet]. bioRxiv. 2015 [cited 2023 Jun 15]; 023754. Available from: biorxiv.org / content / 10.1101 / 023754v2

[0235] 35. Manichaikul A, Mychaleckyj JC, Rich SS, Daly K, Sale M, Chen W-M. Robust relationship inference in genome-wide association studies. Bioinformatics 2010;26(22):2867-73.

[0236] 36. Danecek P, Bonfield JK, Liddle J, et al. Twelve years of SAMtools and BCFtools. Gigascience [Internet] 2021; 10(2). Available from: dx.doi.org / 10.1093 / gigascience / giab008

[0237] 37. Wang K, Li M, Hakonarson H. ANNOVAR: functional annotation of genetic variants from high-throughput sequencing data. Nucleic Acids Res 2010; 38(16): el64.

[0238] 38. Ioannidis NM, Rothstein JH, Pejaver V, et al. REVEL: An Ensemble Method for Predicting the Pathogenicity of Rare Missense Variants. Am J Hum Genet 2016; 99(4): 877-85.

[0239] 39. Landrum MJ, Lee JM, Benson M, et al. ClinVar: improving access to variant interpretations and supporting evidence. Nucleic Acids Res 2018; 46(D1): D1062-7.

[0240] 40. Amberger JS, Bocchini CA, Schiettecatte F, Scott AF, Hamosh A. Omim.org: Online Mendelian Inheritance in Man an online catalog of human genes and genetic disorders. Nucleic Acids Res 2015; 43(Database issue): D789-98.

[0241] 41. Jaganathan K, Kyriazopoulou Panagiotopoulou S, McRae JF, et al. Predicting Splicing from Primary Sequence with Deep Learning. Cell 2019; 176(3):535- 548.e24.

[0242] 42. Gregg AR, Aarabi M, Klugman S, et al. Screening for autosomal recessive and X-linked conditions during pregnancy and preconception: a practice resource of the American College of Medical Genetics and Genomics (ACMG). Genet Med 2021; 23(10): 1793-806.

[0243] 43. Harrison SM, Biesecker LG, Rehm HL. Overview of Specifications to the ACMG / AMP Variant Interpretation Guidelines. Curr Protoc Hum Genet 2019; 103(1):e93.

[0244] 44. Richards S, Aziz N, Bale S, et al. Standards and guidelines for the interpretation of sequence variants: a joint consensus recommendation of the American College of Medical Genetics and Genomics and the Association for Molecular Pathology. Genet Med 2015; 17(5):405-24.

[0245] 45. Riggs ER, Andersen EF, Cherry AM, et al. Technical standards for the interpretation and reporting of constitutional copy-number variants: a joint consensus recommendation of the American College of Medical Genetics and Genomics (ACMG) and the Clinical Genome Resource (ClinGen). Genet Med 2020;22(2):245-57.

[0246] Other embodiments

[0247] It is to be understood that while the application has been described in conjunction with the detailed description thereof, the foregoing description is meant to illustrate and not limit the scope of the application. Other aspects, advantages, and modifications are within the scope of the following claims.

Claims

1. A computer-implemented method for assigning maternal or fetal origin to one or more genetic variations in cell-free DNA (cfDNA) from a sample of a pregnant mammal, preferably a pregnant human, the method comprising: (a) accessing from a memory a probabilistic model for assigning maternal or fetal origin to genetic variation in DNA from a sample obtained from a pregnant mammal, wherein the model assigns maternal or fetal origin based on a combination of fetal fraction and / or DNA fragment size and other sequencing features; (b) inputting into the model a set of values ​​representing one or more genetic variants detected in cfDNA from a peripheral blood sample of a pregnant mammal, wherein the values ​​include empirically determined sequence information, e.g., the proportion of different bases in the reads, and DNA fragment size information, e.g., a rank sum statistic for each genetic variant; and (c) assigning maternal or fetal origin to one or more genetic variants using the model.

2. The method of claim 1, wherein the genetic variation comprises a single nucleotide variation (SNV), an insertion or deletion, and / or a copy number variation (CNV).

3. The method of claim 1 , wherein the initial set of values ​​representing one or more genetic variations is obtained by a method comprising: aligning raw sequencing reads derived from the cfDNA to a reference genome sequence; converting the raw sequencing reads into consensus reads; realigning the consensus reads with the reference genome sequence, thereby generating a set of aligned consensus reads; identifying consensus reads that differ from the reference genome; assigning consensus reads that differ from the reference genome as alternative alleles and assigning consensus reads that match the reference genome as reference alleles, and determining a fragment size rank sum statistic representing a distribution of estimated fragment sizes of reads supporting the reference allele compared to a distribution of fragment sizes of reads supporting the alternative allele, An initial set of values ​​representing sequence identity and DNA fragment size rank sum statistics for one or more genetic variants is thereby obtained.

4. The method of claim 3 , wherein each of the raw sequencing reads comprises a unique molecular identifier (UMI); and the method comprises converting the raw sequencing reads into a single consensus read for each UMI.

5. The method of claim 1 , further comprising selecting a set of candidate variants prior to step (b) by a method comprising: accessing a machine learning classifier, optionally a random forest-based model, from a memory, wherein the machine learning classifier is trained using a set of predetermined filtering criteria and a subset of sites present in the sample or in a reference sample to identify potential false positive (FP) sites; inputting the initial set of variants into the machine learning classifier; and A set of candidate variants is selected from the initial set by filtering using a trained machine learning classifier to remove sets of variants enriched in false positive (FP) sites.

6. The method of claim 1, wherein the probability model is a Bayesian mixture model that simultaneously estimates the fetal fraction and assigns fetal or maternal origin to each variant site in the set.

7. The method of claim 6, wherein the Bayesian mixture model is a Bayesian Gaussian mixture model constrained by variant allele fractions and fragment size rank sum statistics.

8. The method of claim 6, wherein the fetal fraction of the sample is modeled as a latent variable (f), and the mean of the variant allele fraction distribution is set for each component based on f.

9. The method of claim 6, wherein the fetal fraction is estimated based on a reference fetal fraction determined based on clustering derived from VAFs across sites.

10. The method of claims 1-9, further comprising outputting a list of one or more genetic variants identified as having a fetal origin and / or one or more genetic variants identified as having a maternal origin.

11. The method according to claims 1-9, further comprising: comparing the genetic variant to a database containing a list of genetic variants and information regarding the variants' potential medical relevance to the fetus or mother; Identify potentially medically relevant fetal or maternal variants; and A list of potentially medically relevant one or more genetic variants identified as having a fetal origin and / or one or more genetic variants identified as having a maternal origin is output.

12. The method of claim 11, wherein the method further comprises recommending further testing based on the presence of a potentially medically relevant variant.

13. The method of claim 12, wherein the further testing comprises amniocentesis or chorionic villus sampling (CVS); further monitoring of the fetus via ultrasound; or genetic testing of the mother.

14. The method of claims 1-13, further comprising using high-throughput sequencing on cfDNA extracted from a single sample of peripheral blood from the mother, optionally wherein exome capture is performed prior to said sequencing.

15. The method of claim 14, wherein adapters having universal PCR primer sequences and unique molecular identifiers (UMIs) are ligated to the cfDNA, and PCR amplification is performed prior to the sequencing.

16. The method of claim 14, further comprising enriching the sample for fetal DNA, optionally by contacting the cfDNA with a plurality of oligonucleotides that bind to portions of the fetal genome, optionally including fetal protein-encoding genes or other regions of the fetal genome that may be relevant to clinical interpretation or variant identification.

Images