Emerging technologies for prenatal diagnosis: the application of whole genome and RNA sequencing

Abstract

DNA sequencing technologies for clinical genetic testing have been rapidly evolving in recent years, and steadily become more important within the field of prenatal diagnostics. This review aims to give an overview of recent developments and to describe how they have the potential to fill the gaps of the currently clinically implemented methods for prenatal diagnosis of various genetic disorders. It has been shown for postnatal testing that whole genome sequencing provides a set of added benefits compared to exome sequencing, and it is to be expected that this will be the case for prenatal testing as well. RNA-sequencing, already used postnatally, can provide valuable complementary data to DNA-based testing, and aid in variant interpretation. While not ready for clinical implementation, emerging technologies such as long-read and Hi-C sequencing analyses might add to the toolbox for interpreting the expanding genetic data sets generated by genome-wide sequencing. Lastly, we also discuss some more practical implications of introducing these emerging technologies, which generate larger and larger genomic data sets, in the prenatal field.

Introduction

Many studies have been published on the use of exome sequencing for prenatal diagnosis^1–4, with diagnostic yields ranging from 4% for cases with isolated increased nuchal translucency⁵ up to over 80 % in a highly selected population with ultrasound findings suggestive of a skeletal dysplasia⁶. Although professional guidelines on sequencing for fetal diagnosis currently do not recommend prenatal sequencing as a routine diagnostic test but recognize it may be indicated when multiple fetal anomalies are present^7–9, exome analysis has been increasingly adopted in the clinical prenatal setting. Studies on postnatal testing report that sequencing can lead to up to 40% incremental diagnostic yield in addition to the 25-40% by standard karyotyping and chromosomal microarray (CMA) analysis. In addition, prenatal sequencing has also been extremely helpful in elucidating and expanding fetal phenotypes of known disease-causing genes ^10–12.

Exome testing, however, only covers a fraction of the entire genome, and is not adequate for the detection of variants in promoters and other regulatory elements, intronic, and intergenic regions, and of small copy number variants that would go undetected by chromosome or CMA analysis. Hence, with technologies rapidly improving and decreasing sequencing costs, the use of prenatal whole genome sequencing (WGS) is being investigated, as discussed below. In addition, this review explores the potential value of RNA sequencing (RNA-seq) in the prenatal diagnostic setting. Transcriptome analysis by RNA-seq is a newer valuable player in the field of genetic testing that has gained a lot of interest in recent years, and can provide complementary information to DNA-based testing, both in terms of variant interpretation as well as revealing variants that would otherwise go undetected. Finally, we include a perspective on other emerging technologies including long-read sequencing, optical mapping and Hi-C analysis. This review discusses the utilization of these new technologies in the context of unbiased prenatal diagnosis where there is no clear indication based on the clinical information.

Whole Genome Sequencing

Initial attempts for prenatal utilization of whole genome sequencing

Whole genome sequencing enables the comprehensive analysis of almost all genetic information from an individual in one assay in a short turnaround time (Figure 1). Starting from various sample sources, an individual’s genomic DNA is extracted and fragmented, and serves as input material for further library preparation. These fragment libraries subsequently undergo cluster amplification and massively parallel sequencing on one of various platforms. The most commonly used platform currently is the “sequencing-by-synthesis” technology by Illumina¹³, which generates short individual reads of 25 to 300 bp. For traditional testing options such as panel or exome analysis, an enrichment step during library preparation needs to be included¹⁴, partly to circumvent the prohibitive cost of “sequencing everything without selection”. The targeted genomic sequences of interest are usually enriched by specific capture (removal of unselected sequences) or amplification (synthesis of the selected sequences). Because there is no sequence selection by design during WGS sample preparation, this step of enrichment is omitted. The availability of WGS for clinical diagnostic use was limited by its high cost, until thanks to technological advances within the last few years the cost of sequencing the human genome could be lowered to approach $1,000. One technical advantage associated with sequencing the entire genome is that the omission of the enrichment step results in a more uniform sequencing coverage compared to panel or exome sequencing. For the purpose of genetic diagnostic testing, for which most variants of interest are heterozygous, homozygous, or hemizygous germline variants that are rare in the population, it is often recommended that a mean coverage of 30X is adequate for a research grade WGS and a mean coverage >40X is needed for a clinical grade WGS¹⁵. In contrast, for most current exome capture processes, a mean coverage of 100X or higher is needed to ensure that there are enough sequencing reads in poorly covered genomic regions to reliably call heterozygous variants. An additional benefit of the uniformity in sequencing coverage from WGS is that copy number variants can be more reliably assayed; exomes, on the contrary, suffer from bias associated with the capture process and thus often present a challenge for CNV calling (Figure 1).

Figure 1. Comparison of variant detection for different variant types between chromosomal microarray, exome sequencing and whole genome sequencing.

(A) Microarray analysis is capable of detection copy number variants, such as a heterozygous deletion (indicated in blue). The detection of the deletion depends on the coverage and signal from probes laid out on the design of the microarray. A sufficient number of probes located within the genomic region of the copy number variant is necessary for robust detection of the variant. (B) Small variants such as SNPs or INDELs can be detected in coding regions from an exome sequencing assay. Variants located in noncoding regions are not captured and therefore will not be detected. Copy number variants can in theory be detected by analysis of the exome sequencing read count data; however, the robustness of such an analysis is limited. This challenge is illustrated in the figure by the unevenness of read depth (the number of reads covering each base averaged across a region) over the gene, which can potentially mask read depth alterations representing a real copy number variant. (C) Whole genome sequencing generates data that can support variant calling of coding and noncoding small variants, copy number variants, and repeat expansions. Copy number variants can be more readily detected in whole genome sequencing compare to exomes because of the more even baseline sequencing depth. Copy number variant calling is boosted by the higher probability to detect breakpoints by the split-read algorithm (as illustrated by the two broken sequencing reads connected by dotted lines) and the read-pair algorithm (the purple read pairs that are spaced further away compared to other read pairs) in the figure. In addition, the green split-read pairs illustrate a structural variant (SV) such as a translocation, with part of the read aligning to a location elsewhere in the genome. Created with BioRender.com.

The first report of genome sequencing applied in prenatal context came out in 2012 and describes how genome sequencing of an amniotic fluid sample was able to reveal and refine the breakpoints of a balanced translocation disrupting the CHD7 gene, leading to a prenatal diagnosis of CHARGE syndrome that would not have been established by another diagnostic method¹⁶. A second case report from 2015 described how prenatal genome sequencing was used to resolve the breakpoints of a complex chromoanasynthesis event, a rare phenomenon characterized by multiple catastrophic chromosomal rearrangements often detected from germline microarray analysis results in patients with developmental disorders ¹⁷; the prenatal WGS data revealed additional information on cryptic structural events¹⁸. Although in both cases long-read genome sequencing (discussed further) was used instead of the standard short-read workflow, they nevertheless illustrate that WGS can be performed on prenatal DNA specimens generating results similar to those one would expect from a postnatal WGS analysis. For both cases, the decision to perform WGS on a prenatal sample was prompted by the suspicion of a potential complex mutational event that would be challenging to resolve by other methods, showing that the utilization of WGS for prenatal diagnosis had thus far been investigational and focused on selected scenarios.

As pointed out above, one major advantage of WGS is the ability to detect variants in non-coding regions. Although the majority of all described disease-associated variants so far are located within exons or at exon-intron junctions, there are still a considerable number of causative variants reported outside of these regions (Figure 1), for example in regulatory regions (such as promoters, enhancers, and transcription factor binding sites, etc.) ^19,20 or deep intronic regions ²¹. Our understanding of the consequences of these types of variants and the non-coding genome organization in general is more limited than for exonic variants, but it is steadily expanding e.g. by continued mapping and functional characterization of regulatory elements ^22,23, expanded analysis of regulatory variants ^24–26, and improvements to prediction tools to aid in their interpretation are under development^27–29.

CNV detection in whole genome sequencing

A recent study applying trio low-pass genome sequencing for CNV detection on amniotic fluid, chorionic villi and fetal cord blood samples demonstrated a diagnostic yield of 12.4% (ranging from aneuploidies to CNVs down to <100 kb), based on a sample series of 315 pregnancies³⁰. This illustrates a second benefit of genome sequencing: the ability to detect CNVs as an alternative methodology aside from CMA. Two developments, one on the wet-bench side and the other on the dry-lab side, were critical in bringing the CNV detection accuracy by WGS on par with CMA. Omission of the PCR amplification step in library preparation ensured that minimum artificial bias is introduced to the raw data generation for CNV analysis³¹. Meanwhile, although an extensive list of CNV calling tools have been developed and are available for use³², several clinical laboratories decided to prioritize the implementation of one algorithm as they introduced CNV analysis to clinical WGS: the read-depth calling algorithm, i.e. the counting application for the number of sequencing fragments mapping to a fixed genomic interval to infer whether an excess (copy number gain) or a shortage (copy number loss) of reads are present (Figure 1) ^33,34. Implementation of these two developments for postnatal diagnostic WGS yielded promising results showing that CNV detection from WGS can be as robust as CMA ^33,34.

WGS provides a higher resolution for CNV detection compared to CMA (Figure 1). The raw data from WGS carry read count information at each of the three billion bases in the human genome; in contrast, CMA is physically limited in its data content because a maximum of only one to four million oligonucleotide probes can be printed on a microarray slide. Thus, it is possible for WGS to detect CNV missed on a CMA because either the microarray does not have adequate oligo density to support a sensitive detection for small CNVs, or the genomic region of interest is not interrogated well during the array design because other genomic loci were prioritized. Moreover, CNV calling from WGS can be achieved beyond the conventional “counting paradigm” as is done in microarrays: precise or estimated breakpoints of genomic rearrangements can be retrieved from WGS data (Figure 1) to enhance the CNV calling by read-depth analysis³⁵ which can benefit evaluation of challenging CNVs such as low-level mosaic changes. Obviously, balanced copy number neutral structural variants will not be revealed by CMA, but can be captured by WGS as reported above^16,18. Ultimately, it is preferred that algorithms leveraging both the read-depth and breakpoint data are integrated to improve the copy number and structural variant calling³⁶. It is critical to understand the strengths and limitations of each variant calling algorithm to make appropriate decisions regarding when different algorithms can work in synergy or in complement.

Nevertheless, there have yet to be systematic efforts in testing the robustness of these processes in a prenatal setting. In particular, a commonly encountered challenge in prenatal sampling is low DNA input amount, which may limit the utilization of a no-PCR library preparation protocol, potentially compromising CNV calling quality. Notwithstanding, WGS might serve as a diagnostic approach for which less DNA input is needed in general: a recent study comparing genome sequencing to sequential CMA and exome sequencing describes a WGS workflow for which only 100 ng input material is required, as opposed to 400 ng for the sequential test protocol³⁷. Another study from 2018 investigated whole genome analysis of both fetal cells and fetal cell-free DNA obtained from amniotic fluid, and found that both sources are comparable in terms of quality and results. This suggests that amniotic fluid supernatant could also be used as a source of fetal DNA, without having to sacrifice any of the cell pellet going through the standard prenatal diagnosis workflow³⁸. Additional investigations are warranted to explore a protocol to obtain sufficient DNA at an early timepoint in the prenatal testing and generate WGS data with uniform, unbiased sequencing coverage.

Detection of other variant types by whole genome sequencing

Recent WGS pipelines also include a dedicated module for the analysis of deleterious repeat expansions (Figure 1) ³⁹. Although most repeat expansion disorders will only present at later age, prenatal manifestations are possible, for example myotonic dystrophy type I which can present with reduced fetal movements, club feet, polyhydramnios, etc.⁴⁰. For such cases with non-specific prenatal presentations, whole genome sequencing could provide a solution that has the potential to incorporate all diagnostic analyses in a single test, including coding SNVs, CNVs, intronic variants, regulatory region variants, and repeat expansions (Figure 1). In addition, mitochondrial DNA (mtDNA) analysis can be included, although many prenatal cases for which mtDNA testing is performed, have a test indication of a positive family history⁴¹, in which case known familial variant testing rather than WGS would be the more appropriate test to order. Nevertheless, the ability to have all these different analyses incorporated into one workflow, can also considerably aid in getting valuable and clinically relevant results faster compared to sequential testing, or concurrent performing several different tests, each with their own turnaround time. This is of particular importance in the prenatal setting, for which only a short time frame to decide on and prepare for clinical management is available. This was for example illustrated in a 2021 study by Zhou et al., in which the authors reported a median turnaround time of 18 days for genome sequencing versus a median 10+21 day turnaround time for sequential CMA followed by WES³⁷.

The diagnostic yield of whole genome sequencing

Considering the overall benefits of WGS, it is likely to offer the prenatal field a diagnostic yield similar to postnatal testing. Lionel et al. reported clinical WGS findings from 103 patients with pediatric genetic disorders⁴². A molecular diagnosis was achieved in 41% of the cohort, including diagnoses that would not have been made by an exome such as structural variants, non-exonic variants, and mitochondrial variants. The UK 100,000 Genome project performed a similarly comprehensive WGS analysis on over 2000 families with genetic disorders⁴³. The molecular diagnostic yield varied among the type of disorders, with cases likely having a monogenic cause higher than cases with likely a complex cause (35% versus 11%). Multiple reports have been published in the meantime on the utility of (rapid) genome sequencing in the neonatal and pediatric intensive care setting, which was already illustrated back in 2015 by Willig et al.⁴⁴. For a series of 35 critically ill infants, the authors reported a diagnostic yield of 57%, including two variants that would not have been identified on standard exome sequencing: one was an exonic deletion in trans with an SNV, hence genome sequencing provided the full diagnosis, and one was a mitochondrial variant. More recent studies have emphasized these results with diagnostic rates ranging from 20 to 45%, and affecting clinical management (including subspecialty referrals, medication, procedures) in a significant number of cases^45–48. One publication coming out of Project Baby Bear, with recruitment of 184 infants in five of California’s children’s hospitals, highlights the clinical utility of such rapid turnaround time testing and the considerable cost savings in terms of medical care that can go along with that⁴⁹. From these encouraging results we predict that WGS will also play an important role in prenatal diagnosis, especially as prenatal imaging methods are rapidly improving and knowledge on prenatal phenotyping is ever expanding.

RNA-sequencing

Application of RNA-sequencing in postnatal Mendelian diagnostics

Complementary to DNA sequencing, transcriptome analysis plays a key role in the diagnosis of genetic disorders. Two major approaches have been widely applied, gene expression microarray and RNA-seq. Gene expression microarray has been the traditional method utilized in various research studies. Arrays are adequate to characterize expression abundance changes of pre-defined transcripts, but are not suited for the detection of abnormal splicing events or exon-scale gene expression alterations⁵⁰. As a consequence, microarray data often fall short in the diagnosis of single gene disorders, and have mostly been used in profiling and biomarker identification involving the gestalt of multiple genes, including studies focusing on prenatal biology^51–53.

In the past few years, transcriptome analysis by RNA-seq has been increasingly applied in the context of Mendelian diagnostics for pediatric and adult disorders, either focusing on specific disease groups^54–57, or on a broader, diverse set of genetic conditions^58–60. These studies report that the addition of RNA-seq data facilitated an incremental molecular diagnostic rate ranging from 10% to 36%. However, RNA-seq has not yet been applied for clinical diagnosis of prenatal genetic disorders, despite its growing popularity in postnatal diagnostics. A common principle guiding the analysis of RNA-seq data for clinical diagnostics is to identify within the transcriptome, the individual transcriptional alteration that causes the Mendelian disorder. Gestalt analysis, i.e. profiling of an expression signature that includes multiple genes from a pathway, is not usually performed. Three data analysis modalities are generally applied to RNA-seq data for clinical diagnosis of single gene disorders: [1] the detection of gene expression outliers compared to a control group, [2] illegitimate splice junctions not represented in control databases, and [3] monoallelic expression at biallelic sites.

Perspectives on RNA-sequencing utilization in prenatal diagnostics

We anticipate that depending on the clinical scenarios, transcriptome analysis by RNA-seq might be applied in prenatal genetic diagnostics for three major reasons (Figure 2). First, to help resolve the pathogenicity of candidate diagnostic variants of uncertain clinical significance (VUS) that are identified by DNA testing methods, such as panels, exome or WGS (Figure 2A). When WGS data are analyzed nowadays, most pipelines focus on the analysis of the “enhanced exome”, i.e. coding regions (pure exome) plus selected noncoding regions (such as intronic regions near exon/intron junctions, deep intronic/noncoding regions previously reported to be potentially clinically significant, or deep intronic/noncoding regions predicted by computational algorithms, such as spliceAI²⁹, to be potentially intolerant to variation). When candidates are identified from the “selected noncoding regions”, further evidence is usually needed to clarify their molecular deleteriousness. Deeper analysis of this type of candidate variants is beginning to be performed in clinical diagnostic settings. Often a targeted confirmatory assay is required to establish the validity, such as PCR assays that are specifically designed to amplify the hypothesized abnormal transcript²¹ or the expected allele with monoallelic expression⁶¹. The process to design and interpret the targeted confirmatory assay, however, is usually manual and laborious. Thus, it can hardly be scaled up in a clinical laboratory. RNA-seq would be easier to implement in a busy/fully active diagnostic lab workflow, as it can be more automated and applicable to all samples, avoiding the need to develop customized assays. Furthermore, if done in parallel with the DNA sequencing, it creates the opportunity to generate the necessary expression data at the same time of, or even before a candidate variant is nominated from the DNA analysis. These factors are critical for timely turnaround for a prenatal testing result. Furthermore, such evidence-based upgrading or downgrading of the potential pathogenicity of an identified VUS may dramatically alter the scope of the reporting content, especially considering that laboratories may have their own policy to exclude (or include) VUSs for prenatal exome or genome reporting⁸.

Figure 2. Three proposed scenarios for RNA-sequencing to be utilized in prenatal diagnostics.

In all three scenarios, interpretation of the whole genome is sliced into two sections: the “enhanced exome” (pure exome plus selected noncoding regions, as explained in the main text) and the remaining noncoding regions. The three scenarios differ from each other by the degree of thoroughness that the RNA-seq analyses are performed. (A) In the first scenario, RNA-seq data are used to clarify significance of a small number of variants identified from the “enhanced exome” analysis in a “targeted” manner. The majority of the RNA-seq data are not analyzed. The DNA sequencing data beyond the “enhanced exome” region are not analyzed with the RNA-seq data in an integrated manner. (B) In the second scenario, an additional paradigm of the RNA-seq data analysis is introduced, in which the RNA-seq data can be analyzed independent from the DNA data to generate gene expression outliers and cryptic splicing events. However, the DNA sequencing data beyond the “enhanced exome” regions are still not analyzed with the RNA-seq data in an integrated manner. (C) The third scenario, compared to the two previous scenarios, involves an additional paradigm of analysis to integrate the ensemble of other noncoding DNA data and RNA-seq data, generating the highest potential for discovery among the three scenarios.

Second, an alternative, and perhaps even more exciting, goal of RNA-seq analysis is to identify the diagnostic abnormal transcript agnostically, independent from knowledge gained from DNA-based test results (Figure 2B). There is growing evidence that agnostic transcriptome-driven diagnostic approaches can be successfully implemented in a postnatal scenario^57,58,60. Challenging variants may evade detection from genome- or exome-based analysis, but can be more readily identified from RNA-seq analysis. For example, a decrease in gene dosage may be caused in theory by a regulatory variant that is not currently linked to the gene of interest identified during WGS analysis, or by a cryptic splicing event that was not predicted by computational programs. Despite successes from research efforts to identify molecular diagnoses by RNA-seq, there is currently no consensus regarding best practices for systematic filtration, prioritization, and interpretation of RNA-seq results. Furthermore, it is critical that a control database comprising normal expression profiles and splice junctions be established, analogous to gnomAD ⁶² serving as a control database for germline WGS analysis, to facilitate filtering out common, benign variants in the agnostic RNA-seq interpretation pipeline.

Third, RNA-seq data can potentially be integrated into the review process of whole genome sequencing data, which can enable the organic interpretation of intronic variants genome-wide (Figure 2C). Current clinical interpretation pipelines utilized in postnatal WGS diagnostics have not yet incorporated processes to interpret the entire genome, but rather focus on an enhanced digital exome – intronic and intergenic regions are usually excluded from analysis or carved by heavily filtered pre-computed prediction scores^42,43. Development of a pipeline that leverages RNA-seq data to guide the filtration of WGS-detected intronic variants is necessary to enable systematic review of all intronic variants while maintaining the prioritized variants under a manageable number.

The matter of temporal and spatial specific expression in transcriptome analysis

Tissue and temporal specificity are inherent properties that should be considered in any gene expression analysis on an organismal level. On the one hand, tissue- and temporal- specific gene expression enriches the information that can be extracted from the transcriptome data, facilitating the diagnostic process. For instance, a Mendelian disorder presented by structural brain abnormality detected from a fetal ultrasound is expected to be caused by defects in a gene expressed in the fetal brain during the early embryological development time window; variants affecting genes not expressed in this tissue and time period can be deprioritized in the data interpretation. Currently, extensive resources are available from the GTEx consortium regarding the atlas of adult gene expression across dozens of tissue types⁶³. On the contrary, characterization of the human embryological and fetal gene expression atlas has not been done on a similar scale. Recent advances from the single cell perspective are starting to fill this gap in the prenatal period. Cao et al. profiled 4 million single cells of 15 fetal organs from 121 human fetal samples with 72 to 129 days post-conception, and constructed an annotated expression atlas of 657 human fetal cell types⁶⁴. Further expansion, detailed delineation, and a thorough clinical understanding of resources like this⁶⁵ is needed to guide the development of a well-informed clinical transcriptome interpretation pipeline.

On the other hand, tissue- and temporal- specific gene expression creates multiple layers of dimensions in transcriptome data on an organismal level, making RNA-seq data drastically more difficult to deconvolute and dissect compared to WGS data. The degree of variability of gene expression patterns across different tissues has been well appreciated. Particularly relevant to the practice of rare disease molecular diagnostics is the dilemma that the clinically accessible tissues often do not match the tissue of disease pathology. Although for pediatric or adult genetic testing the clinically accessible tissue, for example whole blood, has been demonstrated to serve as an effective proxy for transcriptome analysis⁵⁸, further computational analysis showed that as high as 40.2% of genes can be inadequately represented from RNA-seq if the disease tissue does not match the tested tissue⁶⁶. Knowledge regarding gene expression patterns in the prenatally clinically accessible tissues, namely chorionic villi and amniotic fluid cells, is even more scarce compared to the adult counterparts. In this context, gene expression patterns for the human placenta are being explored at the single cell level⁶⁷.

Cell-free RNA analysis in prenatal diagnostics

Since the initial report in 2005, amniotic fluid cell-free RNA (AF cfRNA) has been a popular source material for clinical researchers with an interest in understanding the dynamics of the developing fetal transcriptome^68,69. AF cfRNA is clinically accessible, albeit that it is often discarded in current fetal genetic test protocols that utilize DNA extracted from amniotic cells. Over the years, several studies have repeatedly illustrated new knowledge regarding maternal and fetal biology gleaned from AF cfRNA, such as those pertaining to normal fetal development^68,70, presentation of genetic disorders^71–73, and pregnancy complications^52,74,75. The primary technology used in most of these studies were microarrays. RNA-seq has been shown to generate gene expression data comparable to microarray, in addition to detecting several expected alternative splicing events including H19 and IGF2 for Russell-Silver/Beckwith-Wiedemann syndromes; however, it was reported that the degraded nature of AF cfRNA may present a technical challenge for RNA-seq⁷⁶. As discussed above, implementation by RNA-seq instead of microarray is critical in the genetic diagnostics application. Thus, further investigation is necessary to evaluate the suitability of AF cfRNA sequencing in a prenatal genetic diagnosis workflow.

Other emerging ‘omics technologies

Several new sequencing technologies have been on the horizon for diagnostic utilization. Many of those technologies aims to tackle a limitation of the Illumina sequencing method: short sequencing reads present a challenge in the mapping of homologous or difficult-to-sequence regions, as well as complex breakpoints of structural variants. As discussed earlier, long-insert WGS had been successfully applied to prenatal cases previously known to carry cryptic complex rearrangements^16,18. Similarly, optical mapping, a long-read genome fragment analysis technology, has been used prenatally to characterize a known familial disease variant for facioscapulohumeral muscular dystrophy, an autosomal dominant disease that can be caused by a contraction of the D4Z4 repeat elements at chromosome 4q35⁷⁷. Optical mapping served as an easier workflow in estimating the D4Z4 repeat number compared to the canonical, labor-intensive genotyping approach by pulsed-field gel electrophoresis and Southern blot analysis. Using the Oxford Nanopore platform, Miller et al. investigated individuals with unresolved postnatal genetic disorders by targeted long-read sequencing ⁷⁸. Complex molecular diagnosis such as repeat expansions and cryptic structural variants were revealed. In the same study, long-read sequencing was also shown to be instrumental in providing variant phasing information and detecting abnormal methylations, facilitating diagnostic interpretation. More recently, converged efforts of genomics, engineering, and informatics illustrated that it is possible to optimize the Nanopore long-read sequencing technology to achieve ultra-rapid genetic diagnosis by WGS in sub-10 hours⁷⁹.

High-throughput chromosome conformation capture (Hi-C) WGS is a technique to analyze spatial genome organization and map 3D chromosome folding⁸⁰. Hi-C has recently been applied for diagnosis of Mendelian disorders. It was shown that Hi-C can leverage the unique biochemical data generated on DNA contact points (e.g. of enhancer-promotor interactions) to inform mapping of structural variants⁸¹. Importantly, the mapping information provided by Hi-C can be used not only for the calling of copy number and structural variants, but also for the interpretation of their functional consequence and pathogenicity in terms of gene expression alterations due to changes in the spatial organization of the 3D genome^82,83. In a prenatal diagnostic scenario, it is conceivable that Hi-C WGS can aid in the interpretation of copy number or structural variants of unknown significance if expression dysregulation due to incorrect/abnormal spatial placement is suspected.

Although these emerging DNA genomic analysis technologies all have their own advantages to the regular WGS sequencing, they often involve dedicated sequencer platforms and informatic workflows that are different from what most large diagnostic operations are currently committed to for ongoing large-scale postnatal diagnostics. As the short-read WGS sequencing platforms have been evolving into an umbrella test that exemplifies great potential in detecting almost all major disease-causing variant types, transitioning these alternative technologies into routine diagnostics will be challenging unless a dramatic higher overall diagnostic yield can be achieved or a specialized use is established. Nevertheless, these new technologies are anticipated to play critical roles in their respective niche areas advancing our basic understanding of the prenatal genome and revealing novel genetic mechanisms that can cause abnormal prenatal phenotypes.

New interpretation and clinical reporting considerations associated with the advent of the new ‘omics data

The field of genetics and genomics has witnessed such a rapid development in sequencing technologies that data are being generated at a rate that is exceeding our ability to fully interpret them, especially in the application of clinical diagnostics. From our experience of postnatal genome-scale analysis and prenatal exome analysis, it has become gradually clear that three issues need to be addressed for data interpretation and clinical reporting. We anticipate that these issues may evolve as the new technologies discussed in this review will eventually be applied to prenatal diagnostics.

First, the challenge of picking the diagnostic “needle” becomes greater as the size of the “haystack” shifts from exome to the whole genome and beyond. The two milestones when we initially introduced genome-scale data interpretation into prenatal and postnatal diagnostics were CMA^84–86 and exome sequencing^87,88. Although it is recommended for both assays that the test indication be provided to the laboratory for clinical correlation of the genetic data^8,89, in practice, the number of variants that must be filtered out because they are unrelated to the patient’s phenotype is much greater for exome analysis than for CMA. The bulkier raw data set from an exome leads to the requirement that a detailed clinical phenotype of the patient should be provided to facilitate the combined phenotype- and genotype-driven diagnostic interpretation. When the scope of data analysis expands into the whole genome, the number of variants that must be interpreted is manifold larger and the need to filter down the variants with the help of more refined and faster bioinformatics tools, paired with more precise phenotype matching becomes even more imminent. The human phenotype ontology (HPO) database, the currently most widely used standardized ontology to describe phenotypic abnormalities, does not have as extensive representation for prenatal disease phenotypes compared to postnatal ones. Systematic characterization and curation efforts are needed to close this gap. Subsequently, extended efforts in mapping of prenatal specific ontologies to diseases and disease genes will be critical to enrich our knowledge base that can fuel bioinformatic tools designed to facilitate diagnostic variant prioritization.

Second, the fact that more data are being generated than what is interpretable, based on the knowledge available at the at the analysis, promises that later reanalysis or reinterpretation may reveal more molecular diagnoses. This has been demonstrated for postnatal diagnosis of genetic disorders by exome sequencing⁹⁰, where it was mainly attributed to the rapid discovery of novel disease genes. The same is expected for prenatal WGS analysis, where the incremental diagnostic yield of re-analysis will potentially be even larger because of the wider variant calling potential that WGS data can offer. Additionally, if postnatal reanalysis of prenatal sequencing data, when the newborn’s phenotypic information can be better defined, becomes more routine in the future, the choice of WGS in the initial prenatal experiment may prove more “future-proof” for increasing the probability to achieve a molecular diagnosis over time.

Third, it is inevitable that when genome-scale data are analyzed, incidental findings will be uncovered. In a prenatal scenario, incidental findings may include postnatally medically actionable variants⁹¹, other postnatal disorders without any prenatal symptoms, misattributed parentage, etc. Recommendations have been formulated regarding these issues⁸, which is not the focus of the current review. However, implementation of new technologies may reveal previously unrecognized incidental findings. For example, future pipelines analyzing prenatal RNA-seq data may suggest risk for nongenetic pregnancy complications. As we are exposed to new data modalities from novel technologies, we should prepare to deal with more unexpected findings and increasingly complex and time-consuming pre- and post-test counseling which will accompany these new opportunities to improve the overall prenatal clinical care.

Concluding remarks

The field of prenatal genetic diagnostics has not yet developed a consensus for any of the new technologies discussed in this review. Experience gained from postnatal genetic diagnostics is likely to inform the analytical expectation of the same technologies in a prenatal setting to a great extent. For example, if a clinical laboratory is well-equipped with experience in validating and producing WGS data for CNV detection on DNA extracted from a patient with a pediatric disorder, the laboratory is likely appropriately prepared to handle the setup for experimental and informatics processing of a prenatal DNA specimen, and able to call similar CNVs. Instead, the roadblock for a successful clinical implementation of a new technology would be a well-rounded understanding of the limitation of the assay, the complexity associated with interrogating the entire genome or transcriptome, and specific clinical and ethical considerations pertaining to the prenatal context. The field is not ready to adopt these new technologies when only a few interesting case studies are published, illustrating proof-of-concept that previously undetected molecular diagnoses can be achieved. The field will only be ready when multiple large cohort series are concluded, showing not only a satisfying range of positive diagnoses being made, but more importantly, a good understanding of the “unresolved” cases, as well as illustrations of unexpected results encountered as we start to democratize these new technologies into prenatal clinical care. Generation of these data will inform guidance by professional societies for their implementation into the evolving field of prenatal genetic diagnosis.

Bulleted statements:

What is already known about this topic?

Many studies have been published on the use and utility of prenatal exome sequencing, which is gradually being implemented in the standard workflow for prenatal diagnosis.

What does this review add?

This review discusses the potential value of applying newer technologies to the field of prenatal genetic testing, including whole genome sequencing, RNA-sequencing, and other emerging technologies such as long-read genome sequencing and Hi-C analysis. We also reflect on the considerations for data interpretation and reporting that come along with these new ‘omics data and are particularly important for the prenatal field.