Low-coverage Whole-genome Sequencing Data Research Articles

Imputation strategies for low-coverage whole-genome sequencing data and their effects on genomic prediction and genome-wide association studies in pigs

The uncertainty resulting from missing genotypes in low-coverage whole-genome sequencing (LCWGS) data complicates genotype imputation. The aim of this study is to find out an optimal strategy for accurately imputing LCWGS data and assess its effectiveness for genomic prediction (GP) and genome-wide association study (GWAS) on economically important traits of Large White pigs. The LCWGS data of 1 423 Large White pigs were imputed using three different strategies: (1) using the high-coverage whole-genome sequencing (HCWGS) of 30 key progenitors as the reference panel (Ref_LG); (2) mixing HCWGS of key progenitors with LCWGS (Mix_HLG) and (3) self-imputation in LCWGS (Within_LG). Additionally, to compare the imputation effects of LCWGS, we also imputed SNP chip data of 1 423 Large White pigs to the whole-genome sequencing level using the reference panel consisting of key progenitors (Ref_SNP). To evaluate effects of the imputed sequencing data, we compared the accuracies of GP and statistical power of GWAS for four reproductive traits based on the chip data, sequencing data imputed from chip data and LCWGS data using an optimal strategy. The average imputation accuracies of the Within_LG, Ref_LG and Mix_HLG were 0.9893, 0.9899 and 0.9875, respectively, which were higher than that of the Ref_SNP (0.8522). Using the imputed sequencing data from LCWGS with the Ref_LG imputation strategy, the accuracies of GP for four traits improved by approximately 0.31–1.04% compared to the chip data, and by 0.7–1.05% compared to the imputed sequencing data from chip data. Furthermore, by using the sequence data imputed from LCWGS with the Ref_LG, 18 candidate genes were identified to be associated with the four reproductive traits of interest in Large White pigs: total number of piglets born - EPC2, MBD5, ORC4 and ACVR2A; number of piglets born healthy - IKBKE; total litter weight of piglets born alive - HSPA13 and CPA1; gestation length - GTF2H5, ITGAV, NFE2L2, CALCRL, ITGA4, STAT1, HOXD10, MSTN, COL5A2 and STAT4. With the exception of EPC2, ORC4, ACVR2A and MSTN, others represent novel candidates. Our findings can provide a reference for the application of LCWGS data in livestock and poultry.

Accurate genotype imputation from low-coverage whole-genome sequencing data of rainbow trout.

With the rapid and significant cost reduction of next-generation sequencing, low-coverage whole-genome sequencing (lcWGS), followed by genotype imputation, is becoming a cost-effective alternative to single-nucleotide polymorphism (SNP)-array genotyping. The objectives of this study were 2-fold: (1) construct a haplotype reference panel for genotype imputation from lcWGS data in rainbow trout (Oncorhynchus mykiss); and (2) evaluate the concordance between imputed genotypes and SNP-array genotypes in 2 breeding populations. Medium-coverage (12×) whole-genome sequences were obtained from a total of 410 fish representing 5 breeding populations with various spawning dates. The short-read sequences were mapped to the rainbow trout reference genome, and genetic variants were identified using GATK. After data filtering, 20,434,612 biallelic SNPs were retained. The reference panel was phased with SHAPEIT5 and was used as a reference to impute genotypes from lcWGS data employing GLIMPSE2. A total of 90 fish from the Troutlodge November breeding population were sequenced with an average coverage of 1.3×, and these fish were also genotyped with the Axiom 57K rainbow trout SNP array. The concordance between array-based genotypes and imputed genotypes was 99.1%. After downsampling the coverage to 0.5×, 0.2×, and 0.1×, the concordance between array-based genotypes and imputed genotypes was 98.7, 97.8, and 96.7%, respectively. In the USDA odd-year breeding population, the concordance between array-based genotypes and imputed genotypes was 97.8% for 109 fish downsampled to 0.5× coverage. Therefore, the reference haplotype panel reported in this study can be used to accurately impute genotypes from lcWGS data in rainbow trout breeding populations.

Marker Density and Models to Improve the Accuracy of Genomic Selection for Growth and Slaughter Traits in Meat Rabbits.

The selection and breeding of good meat rabbit breeds are fundamental to their industrial development, and genomic selection (GS) can employ genomic information to make up for the shortcomings of traditional phenotype-based breeding methods. For the practical implementation of GS in meat rabbit breeding, it is necessary to assess different marker densities and GS models. Here, we obtained low-coverage whole-genome sequencing (lcWGS) data from 1515 meat rabbits (including parent herd and half-sibling offspring). The specific objectives were (1) to derive a baseline for heritability estimates and genomic predictions based on randomly selected marker densities and (2) to assess the accuracy of genomic predictions for single- and multiple-trait linear mixed models. We found that a marker density of 50 K can be used as a baseline for heritability estimation and genomic prediction. For GS, the multi-trait genomic best linear unbiased prediction (GBLUP) model results in more accurate predictions for virtually all traits compared to the single-trait model, with improvements greater than 15% for all of them, which may be attributed to the use of information on genetically related traits. In addition, we discovered a positive correlation between the performance of the multi-trait GBLUP and the genetic correlation between the traits. We anticipate that this approach will provide solutions for GS, as well as optimize breeding programs, in meat rabbits.

Noninvasive early differential diagnosis and progression monitoring of ovarian cancer using the copy number alterations of plasma cell-free DNA

Ovarian cancer (OV) is the most lethal gynecological malignancy and requires improved early detection methods and more effective intervention to achieve a better prognosis. The lack of sensitive and noninvasive biomarkers with clinical utility remains a challenge. Here, we conducted a genome-wide copy number variation (CNV) profiling analysis using low-coverage whole genome sequencing (LC-WGS) of plasma cfDNA in patients with nonmalignant and malignant ovarian tumors and identified 10 malignancy-specific and 12 late-stage-specific CNV markers from plasma cfDNA LC-WGS data. Concordance analysis indicated a significant correlation of identified CNV markers between CNV profiles of plasma cfDNA and tissue DNA (Pearson's r = 0.64, P = 0.006 for the TCGA cohort and r = 0.51, P = 0.04 for the Dariush cohort). By leveraging these specific CNV markers and machine learning algorithms, we developed robust predictive models showing excellent performance in distinguishing between malignant and nonmalignant ovarian tumors with F1-scores of 0.90 and ranging from 0.75 to 0.99, and prediction accuracy of 0.89 and ranging from 0.66 to 0.98, respectively, as well as between early- and late-stage ovarian tumors with F1-scores of 0.84 and ranging from 0.61 to 1.00, and prediction accuracy of 0.82 and ranging from 0.63 to 0.96 in our institute cohort and other external validation cohorts. Furthermore, we also discovered and validated certain CNV features associated with survival outcomes and platinum-based chemotherapy response in multicenter cohorts. In conclusion, our study demonstrated the clinical utility of CNV profiling in plasma cfDNA using LC-WGS as a cost-effective and accessible liquid biopsy for OV.

Genotype error due to low-coverage sequencing induces uncertainty in polygenic scoring

Polygenic scores (PGSs) have emerged as a standard approach to predict phenotypes from genotype data in a wide array of applications from socio-genomics to personalized medicine. Traditional PGSs assume genotype data to be error-free, ignoring possible errors and uncertainties introduced from genotyping, sequencing, and/or imputation. In this work, we investigate the effects of genotyping error due to low coverage sequencing on PGS estimation. We leverage SNP array and low-coverage whole-genome sequencing data (lcWGS, median coverage 0.04×) of 802 individuals from the Dana-Farber PROFILE cohort to show that PGS error correlates with sequencing depth (p = 1.2×10-7). We develop a probabilistic approach that incorporates genotype error in PGS estimation to produce well-calibrated PGS credible intervals and show that the probabilistic approach increases classification accuracy by up to 6% as compared to traditional PGSs that ignore genotyping error. Finally, we use simulations to explore the combined effect of genotyping and effect size errors and their implication on PGS-based risk-stratification. Our results illustrate the importance of considering genotyping error as a source of PGS error especially for cohorts with varying genotyping technologies and/or low-coverage sequencing.

Imputation of low-coverage sequencing data from 150,119 UK Biobank genomes

The release of 150,119 UK Biobank sequences represents an unprecedented opportunity as a reference panel to impute low-coverage whole-genome sequencing data with high accuracy but current methods cannot cope with the size of the data. Here we introduce GLIMPSE2, a low-coverage whole-genome sequencing imputation method that scales sublinearly in both the number of samples and markers, achieving efficient whole-genome imputation from the UK Biobank reference panel while retaining high accuracy for ancient and modern genomes, particularly at rare variants and for very low-coverage samples.

Genome-Wide Analysis of Palindrome Formation with Next-Generation Sequencing (GAPF-Seq) and a Bioinformatics Pipeline for Assessing De Novo Palindromes in Cancer Genomes.

DNA palindromes are a type of chromosomal aberration that appears frequently during tumorigenesis. They are characterized by sequences of nucleotides that are identical to their reverse complements and often arise due to illegitimate repair of DNA double-strand breaks, fusion of telomeres, or stalled replication forks, all of which are common adverse early events in cancer. Here, we describe the protocol for enriching palindromes from genomic DNA sources with low-input DNA amounts and detail a bioinformatics tool for assessing the enrichment and location of de novo palindrome formation from low-coverage whole-genome sequencing data.

Contribution bias of parental genomes to the hybrid lineages of black Amur bream and topmouth culter revealed by low-coverage whole-genome sequencing

Interspecific hybridization has the potential to increase animal genetic diversity, enable the introgression of advantageous genetic variation, and even create hybrid lineages. Using the low-coverage whole-genome sequencing data of black Amur bream (Megalobrama terminalis [SJ]), topmouth culter (Culter alburnus [QZ]), and their hybrid progeny, we focused on the sequence variation and contribution bias to the genomic composition of the hybrid lineages of SJ♀ × QZ♂ [ZJ] and QZ♀ × SJ♂ [FJ]. Both mash-based genetic distance analysis and variant analysis based on various reference genomes revealed that the hybrid lineages' genomes are more prone to QZ composition. Furthermore, we assess the degree of subgenomic intermixing between parents QZ and SJ in the FJ and ZJ hybrid lineages. In the Jaccard index-zero genomic area, 89.41% of the genomic region in ZJ and 90.23% in FJ are entirely from one of the parental genomes. This genomic area is primarily from QZ, regardless of FJ or ZJ. In FJ, the proportion is 55.70%; in ZJ, the proportion is 54.60%. Using the reference-free variant discovery tool, we identified two InDel markers (Both were SSR-type InDel markers) and one (T/A)-type SNP marker that displayed differences in agarose electrophoresis. Our current findings suggest that parental genomic contribution bias leads to an imbalance in the genomic composition of hybrid lineages. The genomic composition bias is connected to species but not maternal effects, giving insights into the development of vertebrate genomes shortly after hybridization.

Cost-effectively dissecting the genetic architecture of complex wool traits in rabbits by low-coverage sequencing

BackgroundRabbit wool traits are important in fiber production and for model organism research on hair growth, but their genetic architecture remains obscure. In this study, we focused on wool characteristics in Angora rabbits, a breed well-known for the quality of its wool. Considering the cost to generate population-scale sequence data and the biased detection of variants using chip data, developing an effective genotyping strategy using low-coverage whole-genome sequencing (LCS) data is necessary to conduct genetic analyses.ResultsDifferent genotype imputation strategies (BaseVar + STITCH, Bcftools + Beagle4, and GATK + Beagle5), sequencing coverages (0.1X, 0.5X, 1.0X, 1.5X, and 2.0X), and sample sizes (100, 200, 300, 400, 500, and 600) were compared. Our results showed that using BaseVar + STITCH at a sequencing depth of 1.0X with a sample size larger than 300 resulted in the highest genotyping accuracy, with a genotype concordance higher than 98.8% and genotype accuracy higher than 0.97. We performed multivariate genome-wide association studies (GWAS), followed by conditional GWAS and estimation of the confidence intervals of quantitative trait loci (QTL) to investigate the genetic architecture of wool traits. Six QTL were detected, which explained 0.4 to 7.5% of the phenotypic variation. Gene-level mapping identified the fibroblast growth factor 10 (FGF10) gene as associated with fiber growth and diameter, which agrees with previous results from functional data analyses on the FGF gene family in other species, and is relevant for wool rabbit breeding.ConclusionsWe suggest that LCS followed by imputation can be a cost-effective alternative to array and high-depth sequencing for assessing common variants. GWAS combined with LCS can identify new QTL and candidate genes that are associated with quantitative traits. This study provides a cost-effective and powerful method for investigating the genetic architecture of complex traits, which will be useful for genomic breeding applications.

CMDB: the comprehensive population genome variation database of China.

A high-quality genome variation database derived from a large-scale population is one of the most important infrastructures for genomics, clinical and translational medicine research. Here, we developed the Chinese Millionome Database (CMDB), a database that contains 9.04 million single nucleotide variants (SNV) with allele frequency information derived from low-coverage (0.06×-0.1×) whole-genome sequencing (WGS) data of 141 431 unrelated healthy Chinese individuals. These individuals were recruited from 31 out of the 34 administrative divisions in China, covering Han and 36 other ethnic minorities. CMDB, housing the WGS data of a multi-ethnic Chinese population featuring wide geographical distribution, has become the most representative and comprehensive Chinese population genome database to date. Researchers can quickly search for variant, gene or genomic regions to obtain the variant information, including mutation basic information, allele frequency, genic annotation and overview of frequencies in global populations. Furthermore, the CMDB also provides information on the association of the variants with a range of phenotypes, including height, BMI, maternal age and twin pregnancy. Based on these data, researchers can conduct meta-analysis of related phenotypes. CMDB is freely available at https://db.cngb.org/cmdb/.

Development and Characterization of 20 Genomic SSR Markers for Ornamental Cultivars of Weigela.

Weigela (Caprifoliaceae) is a genus of ornamental plants popular for its phenotypic variation and hardiness, that includes species hybridized to produce the commercially available cultivars. Despite its popularity, limited genetic resources exist for the genus. Twenty genomic simple sequence repeat (gSSR) markers distributed across the genome were developed using low coverage whole-genome sequencing data of Weigela Spilled Wine®. A cross-amplification evaluation with these 20 gSSR markers on a collection of 18 Weigela cultivars revealed a total of 111 unique alleles, including 36 private alleles. A diagrammatic key was constructed to identify cultivars using only six of the gSSR markers, demonstrating the newly developed gSSR markers are immediately useful for cultivar identification. Future uses could include breeding with marker-assisted selection, determining the history of hybridization of the current cultivated lines, aiding in the construction of genetic maps, and assessing the patterns of population genetic structure of Weigela spp.

Comparison of Genotype Imputation for SNP Array and Low-Coverage Whole-Genome Sequencing Data.

Genotype imputation is the term used to describe the process of inferring unobserved genotypes in a sample of individuals. It is a key step prior to a genome-wide association study (GWAS) or genomic prediction. The imputation accuracy will directly influence the results from subsequent analyses. In this simulation-based study, we investigate the accuracy of genotype imputation in relation to some factors characterizing SNP chip or low-coverage whole-genome sequencing (LCWGS) data. The factors included the imputation reference population size, the proportion of target markers /SNP density, the genetic relationship (distance) between the target population and the reference population, and the imputation method. Simulations of genotypes were based on coalescence theory accounting for the demographic history of pigs. A population of simulated founders diverged to produce four separate but related populations of descendants. The genomic data of 20,000 individuals were simulated for a 10-Mb chromosome fragment. Our results showed that the proportion of target markers or SNP density was the most critical factor affecting imputation accuracy under all imputation situations. Compared with Minimac4, Beagle5.1 reproduced higher-accuracy imputed data in most cases, more notably when imputing from the LCWGS data. Compared with SNP chip data, LCWGS provided more accurate genotype imputation. Our findings provided a relatively comprehensive insight into the accuracy of genotype imputation in a realistic population of domestic animals.

Haplotype-aware inference of human chromosome abnormalities

Extra or missing chromosomes-a phenomenon termed aneuploidy-frequently arise during human meiosis and embryonic mitosis and are the leading cause of pregnancy loss, including in the context of in vitro fertilization (IVF). While meiotic aneuploidies affect all cells and are deleterious, mitotic errors generate mosaicism, which may be compatible with healthy live birth. Large-scale abnormalities such as triploidy and haploidy also contribute to adverse pregnancy outcomes, but remain hidden from standard sequencing-based approaches to preimplantation genetic testing for aneuploidy (PGT-A). The ability to reliably distinguish meiotic and mitotic aneuploidies, as well as abnormalities in genome-wide ploidy, may thus prove valuable for enhancing IVF outcomes. Here, we describe a statistical method for distinguishing these forms of aneuploidy based on analysis of low-coverage whole-genome sequencing data, which is the current standard in the field. Our approach overcomes the sparse nature of the data by leveraging allele frequencies and linkage disequilibrium (LD) measured in a population reference panel. The method, which we term LD-informed PGT-A (LD-PGTA), retains high accuracy down to coverage as low as 0.05 × and at higher coverage can also distinguish between meiosis I and meiosis II errors based on signatures spanning the centromeres. LD-PGTA provides fundamental insight into the origins of human chromosome abnormalities, as well as a practical tool with the potential to improve genetic testing during IVF.

Validation of low-coverage whole-genome sequencing for mitochondrial DNA variants suggests mitochondrial DNA as a genetic cause of preterm birth.

Preterm birth (PTB), or birth that occurs earlier than 37 weeks of gestational age, is a major contributor to infant mortality and neonatal hospitalization. Mutations in the mitochondrial genome (mtDNA) have been linked to various rare mitochondrial disorders and may be a contributing factor in PTB given that maternal genetic factors have been strongly linked to PTB. However, to date, no study has found a conclusive connection between a particular mtDNA variant and PTB. Given the high mtDNA copy number per cell, an automated pipeline was developed for detecting mtDNA variants using low‐coverage whole‐genome sequencing (lcWGS) data. The pipeline was first validated against samples of known heteroplasmy, and then applied to 929 samples from a PTB cohort from diverse ethnic backgrounds with an average gestational age of 27.18 weeks (range: 21–30). Our new pipeline successfully identified haplogroups and a large number of mtDNA variants in this large PTB cohort, including 8 samples carrying known pathogenic variants and 47 samples carrying rare mtDNA variants. These results confirm that lcWGS can be utilized to reliably identify mtDNA variants. These mtDNA variants may make a contribution toward preterm birth in a small proportion of live births.

Rapid genotype imputation from sequence with reference panels.

Inexpensive genotyping methods are essential to modern genomics. Here we present QUILT, which performs diploid genotype imputation using low-coverage whole genome sequence data. QUILT employs Gibbs sampling to partition reads into maternal and paternal sets, facilitating rapid haploid imputation using large reference panels. We show this partitioning to be accurate over many megabases, enabling highly accurate imputation close to theoretical limits and outperforming existing methods. Moreover, QUILT can impute accurately using diverse technologies, including using long reads from Oxford Nanopore Technologies, and a novel form of low-cost barcoded Illumina sequencing called haplotagging, with the latter showing improved accuracy at low coverages. Relative to DNA genotyping microarrays, QUILT offers improved accuracy at reduced cost, particularly for diverse populations that are traditionally underserved in modern genomic analyses, with accuracy nearly doubling at rare SNPs. Finally, QUILT can accurately impute (4-digit) HLA types, the first such method from low-coverage sequence data.

353 ASAS-EAAP Talk: Low-coverage whole-genome sequencing in local livestock breeds

Abstract Current European regulations for autochthonous livestock breeds put a special emphasis on pedigree completeness, which requires laboratory paternity testing by genetic markers in most cases. This entails significant economic expenditure for breed societies and precludes other investments in breeding programs, such as genomic evaluation. Within this context, we developed paternity testing through low-coverage whole-genome data in order to reuse these data for genomic evaluation at no cost. Simulations relied on diploid genomes composed by 30 chromosomes (100 cM each) with 3,000,000 SNP per chromosome. Each population evolved during 1,000 non-overlapping generations with effective size 100, mutation rate 10–4, and recombination by Kosambi’s function. Only those populations with 1,000,000 ± 10% polymorphic SNP per chromosome in generation 1,000 were retained for further analyses, and expanded to the required number of parents and offspring. Individuals were sequenced at 0.01, 0.05, 0.1, 0.5 and 1X depth, with 100, 500, 1,000 or 10,000 base-pair reads and by assuming a random sequencing error rate per SNP between 10–2 and 10–5. Assuming known allele frequencies in the population and sequencing error rate, 0.05X depth sufficed to corroborate the true father (85,0%) and to discard other candidates (96,3%). Those percentages increased up to 99,6% and 99,9% with 0,1X depth, respectively (read length = 10,000 bp; smaller read lengths slightly improved the results because they increase the number of sequenced SNP). Results were highly sensitive to biases in allele frequencies and robust to inaccuracies regarding sequencing error rate. Low-coverage whole-genome sequencing data could be subsequently integrated into genomic BLUP equations by appropriately constructing the genomic relationship matrix. This approach increased the correlation between simulated and predicted breeding values by 1.21% (h2 = 0.25; 100 parents and 900 offspring; 0.1X depth by 10,000 bp reads). Although small, this increase opens the door to genomic evaluation in local livestock breeds.

Non-invasive prediction of fetal growth restriction by whole-genome promoter profiling of maternal plasma DNA: a nested case-control study.

ObjectiveTo predict fetal growth restriction (FGR) by whole‐genome promoter profiling of maternal plasma.DesignNested case–control study.SettingHospital‐based.Population or Sample810 pregnancies: 162 FGR cases and 648 controls.MethodsWe identified gene promoters with a nucleosome footprint that differed between FGR cases and controls based on maternal plasma cell‐free DNA (cfDNA) nucleosome profiling. Optimal classifiers were developed using support vector machine (SVM) and logistic regression (LR) models.Main outcome measuresGenes with differential coverages in promoter regions through the low‐coverage whole‐genome sequencing data analysis among FGR cases and controls. Receiver operating characteristic (ROC) analysis (area under the curve [AUC], accuracy, sensitivity and specificity) was used to evaluate the performance of classifiers.ResultsThrough the low‐coverage whole‐genome sequencing data analysis of FGR cases and controls, genes with significantly differential DNA coverage at promoter regions (−1000 to +1000 bp of transcription start sites) were identified. The non‐invasive ‘FGR classifier 1’ (CFGR1) had the highest classification performance (AUC, 0.803; 95% CI 0.767–0.839; accuracy, 83.2%) was developed based on 14 genes with differential promoter coverage using a support vector machine.ConclusionsA promising FGR prediction method was successfully developed for assessing the risk of FGR at an early gestational age based on maternal plasma cfDNA nucleosome profiling.Tweetable abstractA promising FGR prediction method was successfully developed, based on maternal plasma cfDNA nucleosome profiling.

Statistical Mitogenome Assembly with RepeaTs.

By using next-generation sequencing technologies, it is possible to quickly and inexpensively generate large numbers of relatively short reads from both the nuclear and mitochondrial DNA (mtDNA) contained in a biological sample. Unfortunately, assembling such whole-genome sequencing (WGS) data with standard de novo assemblers often fails to generate high-quality mitochondrial genome sequences due to the large difference in copy number (and hence sequencing depth) between the mitochondrial and nuclear genomes. Assembly of complete mitochondrial genome sequences is further complicated by the fact that many de novo assemblers are not designed for circular genomes and by the presence of repeats in the mitochondrial genomes of some species. In this article, we describe the Statistical Mitogenome Assembly with RepeaTs (SMART) pipeline for automated assembly of mitochondrial genomes from WGS data. SMART uses an efficient coverage-based filter to first select a subset of reads enriched in mtDNA sequences. Contigs produced by an initial assembly step are filtered using the Basic Local Alignment Search Tool searches against a comprehensive mitochondrial genome database and are used as "baits" for an alignment-based filter that produces the set of reads used in a second de novo assembly and scaffolding step. In the presence of repeats, the possible paths through the assembly graph are evaluated using a maximum likelihood model. Additionally, the assembly process is repeated for a user-specified number of times on resampled subsets of reads to select for annotation of the reconstructed sequences with highest bootstrap support. Experiments on WGS data sets from a variety of species show that the SMART pipeline produces complete circular mitochondrial genome sequences with a higher success rate than current state-of-the-art tools, particularly for low-coverage WGS data sets.

DpGMM: A Dirichlet Process Gaussian Mixture Model for Copy Number Variation Detection in Low-Coverage Whole-Genome Sequencing Data

Comprehensive identification and cataloging of copy number variation (CNVs) are essential to providing a complete view of human genetic variation and to finding diseased genes. Due to the large-scale sequencing and cost control whole-genome sequencing (WGS) data, low-coverage data is favorably disposed towards CNV identification. However, such low-coverage data is sensitive to noise and sequencing biases, which results in low resolution of CNV detection in past experimental designs for WGS datasets. In this paper, we present a control-free Dirichlet process Gaussian mixture model (dpGMM) based approach, to analyze the read depth (RD) of low-coverage WGS datasets for CNV discovery. First, noise and biases of the RD signals are corrected through the preprocessing step of dpGMM. Then we assume that RD signals across genomic regions follow a Gaussian mixture model (GMM) in which each Gaussian distribution is followed by a copy number state. Without requiring the number of Gaussian distributions, dpGMM builds a Dirichlet process (DP) GMM for RD signals and further uses a DP prior to infer the number of Gaussian models. After that, we apply dpGMM to simulation datasets with different coverages and individual datasets, and compare ours to three widely used RD-based pipelines, CNVnator, GROM-RD, and BIC-seq2. Simulation results demonstrate that our approach, dpGMM, has a high F1 score in both low- and high- coverage sequences. Also, the number of overlaps between CNVs detected in real data by ours and the standard benchmark is twice as much as that detected by other tools such as CNVnator and GROM-RD.

Computational framework for targeted high-coverage sequencing based NIPT.

Non-invasive prenatal testing (NIPT) enables accurate detection of fetal chromosomal trisomies. The majority of publicly available computational methods for sequencing-based NIPT analyses rely on low-coverage whole-genome sequencing (WGS) data and are not applicable for targeted high-coverage sequencing data from cell-free DNA samples. Here, we present a novel computational framework for a targeted high-coverage sequencing-based NIPT analysis. The developed framework uses a hidden Markov model (HMM) in conjunction with a supplemental machine learning model, such as decision tree (DT) or support vector machine (SVM), to detect fetal trisomy and parental origin of additional fetal chromosomes. These models were developed using simulated datasets covering a wide range of biologically relevant scenarios with various chromosomal quantities, parental origins of extra chromosomes, fetal DNA fractions, and sequencing read depths. Developed models were tested on simulated and experimental targeted sequencing datasets. Consequently, we determined the functional feasibility and limitations of each proposed approach and demonstrated that read count-based HMM achieved the best overall classification accuracy of 0.89 for detecting fetal euploidies and trisomies on simulated dataset. Furthermore, we show that by using the DT and SVM on the HMM classification results, it was possible to increase the final trisomy classification accuracy to 0.98 and 0.99, respectively. We demonstrate that read count and allelic ratio-based models can achieve a high accuracy (up to 0.98) for detecting fetal trisomy even if the fetal fraction is as low as 2%. Currently, existing commercial NIPT analysis requires at least 4% of fetal fraction, which can be possibly a challenge in case of early gestational age (<10 weeks) or high maternal body mass index (>35 kg/m2). More accurate detection can be achieved at higher sequencing depth using HMM in conjunction with supplemental models, which significantly improve the trisomy detection especially in borderline scenarios (e.g., very low fetal fraction) and enables to perform NIPT even earlier than 10 weeks of pregnancy.