De novo Mutations in Developmental Disorders

August 17, 2017 by Dan Koboldt

I’m an avid supporter of large-scale genomics studies, which have yielded some of the most important resources in human genetics and cancer genomics. With my new focus on inherited disorders that affect children, I have a new appreciation for initiatives that tackle not-so-common diseases.

One such effort is the UK-based Deciphering Developmental Disorders (DDD) study, which recruited 13,000 patients with undiagnosed developmental disorders from 2011 to 2015. Their penultimate findings, published earlier this year in Nature, have implicated 14 new candidate genes for DDs and provided some fascinating insights into their genetic underpinnings. Yet I’m also discussing this publication because it illustrates some important considerations for sequencing studies of genetic disorders.

The DDD study performed exome sequencing on 4,293 families, predominantly as classic trios of an affected individual and his or her parents. Most affecteds (81%) had been screened for large duplications and deletions, which is typically done by array and captures events measuring hundreds of thousands or millions of base pairs. The authors also collected the family history and detailed clinical information (e.g. growth measurements, developmental milestones, and HPO terms).

De Novo Mutation Calling

The authors placed a primary emphasis on de novo mutations (DNMs) present in the affected child but not his or her parents. Well-established de novo mutation rates in humans predict that newborns harbor, on average, 0.5 DNMs in coding sequences. The authors used DeNovoGear and identified ~1.95 coding DNMs per proband. They acknowledge that this is too high, but argue that they were aiming for sensitivity and can address the lower specificity by independent validation.

DNM Filtering Stringency (DDDS, Nature 2017)

Their filtering strategy for candidate DNMs appears technically sound. Using a set of 1,133 trios for which all candidate DNMs underwent validation (3730 sequencing), the authors chose a posterior probability threshold leaning towards sensitivity (see right). They also applied some common sense filters:

Requiring that GATK call the variant in the child but neither parent
Removing DNMs with strand bias (Fisher p<10^-3) in supporting reads
Removing DNMs at multiallelic sites or with 1+ supporting reads in both parents

If we assume a conservative validation rate of 80%, that would yield about 1.5 DNMs per proband (a 3-fold enrichment over expectation). Yet this may be the true underlying biology: by enrolling DD patients with predominantly healthy parents, they’ve almost certainly enriched their cohort for individuals harboring coding DNMs.

DD Gene Discovery

Of the 4,293 individuals in this study, 23% had “probably pathogenic” missense or truncating DNMs within a clinically curated set of dominant DD genes. To systematically search for frequently mutated genes in DDs, the authors combined their 8,361 candidate DNMs with 4,224 DNMs (3,287 individuals) from 13 published sequencing studies. Some 93 genes achieved genome-wide significance, of which 80 had previous evidence of association with developmental disorders.

DD Associated Genes (DDDS, Nature 2017)

The authors performed some interesting analyses that I won’t cover in-depth here, such as:

Creating composite face images from clinical photos of patients harboring mutations in the same gene to highlight common facial dysmorphology. It’s essentially the converse of the Face2Gene project. Honestly, the composites all look the same to my untrained eye; I find the example photos in Smith’s to be more useful.
Integrating phenotypic data into disease gene association. Combining genetic evidence with HPO term similarity increase the significance for some genes, but decreased significance for a larger number of genes that are associated with DD and broader phenotypes.

About 20% of individuals in the cohort had clinical evidence of seizures and/or epilepsy, so the authors compared the analysis of this subgroup compared to the entire cohort for seizure-associated genes. Fifteen such genes achieved significance in both the whole cohort and the clinical seizure subset, but another 9 genes were significant only in the cohort analysis.

Furthermore, of the 285 individuals with pathogenic DNMs in seizure-associated genes, only about half (56%) had clinical evidence of seizures. The authors conclude that a large sample size easily outweighs phenotypic similarity for achieving statistical power to discover new disease genes.

Exome or Genome Sequencing?

The authors felt that the large number of significant genes provided an opportunity to compare exome and genome sequencing, with respect to their power to identify genes associated at genome-wide significance. They laid out the following assumptions:

Genome sequencing costs $1,000 per sample.
Exome sequencing costs 30-40% of that cost per sample
Genome sequencing uncovers 5% more coding variants
Project budgets is $1 million, $2 million, or $3 million

Under these assumptions, the authors found that “exome sequencing detects more than twice as many genome-wide significant genes.” Well, sure: if you can sequence 3 times as many samples with 95% sensitivity, you obviously have far more statistical power. Sample numbers always trump slight increases in variant detection sensitivity.

But let’s not forget that this is a gene-centric analysis focused on coding DNMs, which are accessible to both exome and WGS strategies. It ignores the classes of pathogenic variants that are refractory to detection by exome sequencing, such as noncoding and structural variants. These are harder to find and to distinguish from benign variation than coding variants. But if we don’t look for them, we’ll never fully understand the genetic architecture of inherited diseases.

Previous Testing and the Falling Hit Rate

The authors also noted that, although their power calculations estimated they’d identify about half of all haploinsufficient DD-associate genes at genome-wide significance. Empirically, they identified slightly less than half (47%). They hypothesized that a priori genetic testing may have depleted the cohort of easy-to-diagnose disorders. Indeed, an assessment of the DNMs in the most clinically recognizable DD genes showed a three-fold lower enrichment for truncating mutations than other DD genes. When these disorders were removed, they identified 55% of all haploinsufficient genes. Right on the money.

This example illustrates an idea about the feedback loop between clinical diagnostic sequencing and research discovery sequencing. In the first several years after exome sequencing was applied to rare disease diagnosis in appreciable numbers, researchers gradually came to recognize that the industry-wide “hit rate” — the proportion of sequenced cases achieving a molecular diagnosis — is typically 25-35% across centers and studies. Clinical and diagnostic laboratories have often adopted this as an informal benchmark for the success of their pipelines.

As this study illustrates, the proportion of cases that achieve a molecular diagnosis depends considerably on the extent of genetic (and biochemical) testing already performed. For completely naive (untested) individuals, we expect that hit rate to increase over time as new assays, methods, and knowledge improves our ability to find the underlying cause of disease.

Yet the improvement and widespread adoption of genetic testing means that such naive samples are increasingly rare. Most patients with a rare disease, particularly a recognizable one, will undergo a barrage of gene or panel testing (and often a copy number array) as part of the diagnostic process. As we find and establish new disease genes, that barrage becomes ever more effective at providing diagnoses before exome or genome sequencing are necessary.

In summary, we should not be alarmed (or self-critical) if the “hit rate” of exome and genome sequencing studies decreases over time. That is the expected outcome. In settings like ours, where research aims to inform clinical care, it’s actually a sign that we’re moving in the right direction.

Clinical Sequencing: Addressing the Dreaded VUS

July 19, 2017 by Dan Koboldt

The adoption of exome and whole genome sequencing as frontline genetic tests has definitively proven to increase the speed and success rate of molecular diagnosis. These improvements should not only help end the diagnostic odyssey for many patients, but also allow faster intervention and genome‐informed clinical care.

Yet the task of interpreting exome‐ or genome‐wide variants of possible clinical relevance remains daunting for geneticists and genetic counselors. Unbiased sequencing yields a substantial number of variants, many of which are private to an individual or family and whose pathogenicity is difficult to assess. As a result, genetic testing reports increasingly contain variants of unknown significance (VUS).

The Problem with VUS

VUS are a growing problem in this era of widespread clinical exome sequencing. They’re generally included in clinical reports when associated with a relevant patient phenotype. Yet they’re also the most time-consuming class of variants to interpret for both the clinical laboratory personnel and the clinician receiving the report. I imagine they’re difficult to explain to patients as well (“Yes, you have a variant. No, we don’t know what it means.”).

A VUS classification simply means that a variant didn’t meet the ACMG-defined criteria to be classified as pathogenic/likely-pathogenic or benign/likely-benign. Sometimes it’s due to information that’s missing, such as the variant’s mutation status (i.e. is it de novo) or its effect on protein function/splicing. Other times, the information at hand is conflicting: computational algorithms have different estimates of pathogenicity, or the variant has been reported in both disease patients and healthy individuals.

The VUS problem is not limited to clinical sequencing, by the way. Researchers conducting studies of rare/Mendelian disorders also struggle with this uncertainty when trying to uncover new disease-causing genes and mutations.

VUS-busters at ASHG

The bad news is that VUS are a significant issue. The good news is that many groups are working on methods, tools, databases, and other resources to shift would-be-VUS into a more useful category on a clinical report. I worked with several of them to propose an invited session on VUS-busting for the American Society of Human Genetics Meeting, which takes place in Orlando in October. I’m pleased to report that the programming committee responded with enthusiasm, and selected our invited session as a clinical spotlight!

Session Title: VUS-busters: Cutting-edge Strategies for Interpreting Variants in Clinical and Research Sequencing

Date: Thursday, October 19, 2017

Time: 4:15-6:15 p.m.

Moderators: Dan Koboldt (Nationwide Children’s Hospital) and Aaron Quinlan (University of Utah)

Description: This session brings together a diverse panel of experts in clinical genomic medicine to address issues of central importance to both researchers and clinicians. They will describe state‐of‐the‐art approaches for alleviating the interpretation bottleneck, including aggregated genome databases, comprehensive variant annotation, and phenotype‐driven analysis. They will also describe best practices for solving difficult‐to‐ diagnose cases: disorders arising from unusual inheritance models, somatic and germline mosaicism, pathogenic noncoding variants, structural variants, and multifactorial genetic bases. Together, they will provide cohesive guidance for improving the speed and success of variant interpretation.

I’m pleased to announce these four speakers, each of whom will bring a unique perspective to the session:

Anne O’Donnell-Luria (Boston Children’s Hospital), on using large-scale, diverse reference databases to improve variant interpretation
Kim McBride (Nationwide Children’s Hospital), on identifying oligogenic causes of congenital heart disease.
Nara Sobreira (Johns Hopkins University), on novel analytic approaches used to solve unsolved whole exome sequencing data.
Peter Robinson (Jackson Laboratory), on phenotype-driven analysis of exome and genome data.

Special thanks to Anne, Kim, Nara, Peter, and Aaron for helping me prepare a competitive invited session proposal, and to the programming committee for supporting our vision. See you in Orlando this October!

RNA-Seq for Mendelian Disease Cases

May 26, 2017 by Dan Koboldt

Next-generation sequencing technologies have transformed the way we study rare genetic disorders. Exome sequencing for Mendelian disease is the poster child for success in this arena. However, despite the frenzied pace of disease gene discovery and the growth of public sequence databases, the diagnostic/success rates for exome sequencing have remained somewhat constant over the past several years. Depending on the type of disorder, available samples, mode of inheritance, and other factors, one can expect that ~35-50% of clinical cases may be solved by exome sequencing.

On the research side, the success rates for rare disease studies have often gone down, for the simple reason that as we improve the catalogue of disease-causing genes/mutations, genetic tests become increasingly effective. In other words, easy-to-solve cases never make it to a research study. The cases that do make it through, while valuable for gene discovery, will almost certainly have a low success rate.

Knowing this intellectually does not make it any less frustrating when you can’t solve a rare disease patient or family where there’s clearly something to find. When an exome is negative, what recourse do we have? Whole-genome sequencing is often what researchers consider pursuing next — in theory, it enables the detection of variants that are not detectable by exome sequencing (e.g. in the 5-10% of coding bases that exome sequencing does not sufficiently cover, or structural variants with noncoding breakpoints). Yet the “rescue rate” of WGS for negative exome cases has been disappointing.

Transcriptome sequencing offers a possible alternative to further investigate negative exome cases. For certain categories of variants (e.g. nonsense, frameshift, and splice site mutations), RNA-seq can provide a molecular confirmation of the predicted pathogenic effect. This is especially helpful in the gray area of splice-region variants, the effect of which is difficult to assess from genomic information alone.

A recent paper from the MacArthur lab demonstrates the utility of RNA-seq in patients with genetically undiagnosed rare muscular disorders. They applied RNA-seq to muscle tissue from 50 unsolved cases. Four of these had a splice region variant of uncertain significance. Another 12 had a strong candidate gene based on clinical presentation and/or the presence of a single pathogenic variant in a known recessive gene. The majority (34) had undergone exome/genome sequencing but yielded no candidates.

Pathogenic Extended Splice Variants (2 cases)

Exon extension caused by a donor +3 A>C extended splice site variant (B. Cummings et al, Sci Transl Med 2017)

First, the straightforward analysis. In the four cases with candidate variants in the extended splice region (i.e. not the canonical splice site, but close to it), RNA-seq confirmed pathogenicity in 2 of them. This is a small sample size, so all it really does is highlight the difficulty of predicting variant effects outside of the canonical splice site.

Unrecognized/Undetected Splice Variants (4 cases)

Interestingly, RNA-seq enabled the diagnosis of a three cases in which a splicing variant was not suspected:

1 with a splice-region variant was pathogenic but had been missed by exome sequencing
1 with a missense variant that alters splicing
2 synonymous variants that alter splicing

The missense variant is interesting: it creates a stronger splicing motif than the wild-type sequence, leading to an exonic splice gain:

Exonic splice gain caused by a C>T donor splice site–creating variant in patient N22

Deep Intronic Variants that Alter Splicing

The real power of RNA-seq in cases like these is to identify novel splicing events that are not due to obvious splice site/region variants. In seven patients, the authors of this study identified aberrant splicing caused by deep intronic variants in known disease genes (DMD and COL6A1).

Intronic splice gain in patient N33 caused by a C>T donor splice site–creating deep intronic variant.

RNA-seq is really the only approach that can identify an event like this. Exome sequencing doesn’t cover this region. Whole-genome sequencing does, but undoubtedly will return dozens of intronic variants, most of which have no effect on splicing.

Rescue Rate for RNA-Seq

Overall, transcriptome sequencing enabled the diagnosis of 17 patients, or 34%. This rate is skewed a bit high by cases with candidate splice region variants (2/4 solved) and cases in which there was already a strong candidate gene (66%). Among cases without such strong leads, the rescue rate was 21%. This is a not-inconsiderable improvement to the overall diagnostic rate, and a strong argument that RNA-seq should be considered as a possible next step when exome sequencing alone does not achieve a diagnosis.

Importance of the Right Tissue

It should be noted that muscular disorders were well-suited to a transcriptome approach, because the relevant tissue is both accessible and often already banked due to the fact that muscle biopsy is a routine (if painful) diagnostic procedure. For other types of rare disorders, one may not have this luxury. In retinitis pigmentosa, for example, the relevant tissue is the retina. Good luck getting any RP patients to volunteer any of their diminishing retinal cells. Developmental timing is also a consideration: for some disorders, like birth defects, the genetic aberration affects a certain phase of development that is no longer ongoing when the patient is in the hospital.

You’ll note that the figures shared in this post also included, in blue, the splice pattern observed in normal individuals. These were gleaned from a subset of the skeletal muscle samples sequenced by the GTex consortium that were carefully matched to the patient specimens. A tissue-matched normal is incredibly powerful for analyses such as this one, and reinforces the need for more large-scale efforts to put transcriptome data in the public domain.

Structural Variation’s Role in Gene Expression

April 7, 2017 by Dan Koboldt

One significant challenges accompanied by advances in high-throughput genomic technologies (i.e. high-density arrays and next-gen sequencing) is that the speed at which we can detect candidate variants outstrips our ability to interpret their effect on phenotypes. In other words, genomic discovery is fast, but functional validation is slow. Also, the relationship between a sequence variant and an observable human trait tends to be complex, making it difficult to elucidate the mechanisms underlying an association.

Gene expression offers an appealing intermediate phenotype. It’s influenced by sequence variation, quantifiable, and can be assayed in high-throughput fashion. Early systematic efforts to identify expression quantitative trait loci (eQTLs) revealed some fascinating insights into the genetic regulation of gene expression, such as the fact that much of that genetic control occurs in noncoding regions. Also, most eQTLs were identified in proximity to the affected gene (cis-eQTLs), but some were millions of base pairs away or on other chromosomes entirely (trans-eQTLs).

Most eQTL surveys conducted thus far are limited to a certain cell type — most often, lymphoblastoid cell lines — and utilize genetic information from SNP arrays or low-pass whole genome sequencing (WGS). That’s why I’ve been excited to see the results of the Genotype-Tissue Expression (GTEx) project, which has applied RNA-seq to numerous different human tissues from donors that also underwent genome sequencing. In addition to a wide set of tissues profiled, their WGS data allows for a fairly comprehensive interrogation of genomic variation including large structural variants (SVs).

In the latest issue of Nature Genetics, members of the GTEx consortium led by several of my former WashU colleagues report the most extensive study to date of the impact of structural variation on gene expression. Their dataset comprises RNA-seq from 13 different tissues along with deep WGS (blood) from a total of 147 individuals.

Genome-wide Structural Variation

A strength of this study is that the authors have considerable expertise in identifying SVs from genome sequencing data. Across the 147 individuals, they identified 23,602 high-confidence SVs, including deletions, duplications, inversions, and other SVs.

SVs Detected (Chiang et al, Nat. Genet. 2017)

Common SVs Affect Gene Expression

Of the 9,000 common SVs (MAF>0.05) in their dataset, 1,634 were associated with gene expression variation in at least one tissue. Because an SV can influence gene expression in multiple tissues and vice-versa, their complete analysis uncovered 5,128 individual eQTLs affecting 2,064 different genes.

Some 11% of the eQTLs involved an SV that altered one or more exons of the affected gene. More than 90% of the time, the effect on gene expression is consistent with the type of SV involved, i.e. deletions tend to decrease expression and duplications tend to increase it:

Effect of SV on Gene Expression (Chiang et al, Nat. Genet. 2017)

That’s a nice result if not an altogether surprising one, since you’d hope that changing the genomic copy number of a gene would alter its expression in somewhat predictable ways.

Contribution of SVs to eQTLs

The authors also identified eQTLs involving SNVs and small indels, which are more numerous and generally easier to detect than SVs with current next-gen sequencing technologies. This offered a more complete picture of genomic variants affecting gene expression, and allowed them to characterize the relative contribution of each type of variant to eQTLs. Using fine-mapping approaches, Chiang et al found that for 3.5 to 6.8% of cases, a structural variant was the probable causal variant.

Heritability of each eQTL apportioned to the most significant SV in the cis window (X-axis) and the additive effect from the top 1,000 most significant SNVs and indels in the cis window (Y-axis) in joint eQTL (left) or mapping (right) analyses (Chiang et al, Nat. Genet. 2017)

This might seem like a trivial contribution, but remember that the typical human genome harbors ~4 million SNVs and indels, but just 5-10,000 SVs. On a per-variant basis, an SV was 28 to 54 times more likely to modulate gene expression than an SNV or indel. Moreover, this is probably an under-estimate of the contribution of SVs, since our ability to detect/genotype them with current WGS technologies is considerably less than our ability to detect smaller variants.

Noncoding SVs Affect Gene Expression

Interestingly, 88.3% of SVs predicted to cause an eQTL did not affect gene dosage or structure. Instead, they likely exert their effects via noncoding regulatory mechanisms. Cross-referencing them with several classes of predicted regulatory elements revealed that SV-eQTLs were significantly enriched in regions that were:

10kbp upstream or downstream of the gene transcript
Within 1 kbp of an enhancer
Predicted to bind transcription factors (by FunSeq)

Overall, this study beautifully illustrates the important role that structural variants have in modulating gene expression. It also lends additional evidence to what I’ve been saying for a long time: that noncoding regulatory variants probably underlie the majority of inter-individual differences in humans.

« Previous Page