4 DISCUSSION
GS offers comprehensive variant detection in coding and non-coding
regions of the genome. Thus, GS may provide diagnostic benefit for
conditions with high degrees of genetic and phenotypic heterogeneity,
such as GDD/ID. In this study, pediatric patients with unexplained
GDD/ID despite prior genomic testing (i.e., CMA and/or ES) received GS
resulting in an overall diagnostic yield of 23% (23/100). When compared
to previous testing, the reasons for missed diagnoses included small
variants undetectable by CMA (9/23, 39.1%), CNVs missed by ES (4/23,
17.4%), and variants in new disease-causing genes (4/23, 17.4%).
For the families with previous CMA testing only, the yield after GS was
high (9/14, 64.3%) and reanalysis of the CMA data did not provide any
diagnoses. Though studies suggest that all previously reported genetic
testing data should be periodically
reinterpreted(SoRelle, Thodeson, Arnold,
Gotway, & Park, 2019; Sun et al.,
2020), the present data demonstrate that there is little benefit in
reanalysis of CMA data. Rather, our data suggest that patients with
negative CMA results should be offered GS to increase the likelihood of
reaching a diagnosis.
Previous work has shown that periodic reanalysis of ES data would
benefit some patients (Wenger et al.,
2017; Xiao et al., 2018). Here, we show
that 8.1% (7/86) of the patients with ES data could be solved by
reanalysis. Although the study design included a reanalysis step, it was
difficult to account for newly reported disease-causing genes that
emerged during the time interval between study enrollment and study
conclusion. Periodic reanalysis of ES has some economic advantages over
GS(Koriath et al., 2021), however this is
based on the assumption that the pathogenic variants are located in the
coding region of the human genome (representing approximately 1% of the
entire genome) and not in non-coding regions where ES is unable to
interrogate.
Our data show that 15 cases could not be solved by reanalysis including
seven that received previous ES. Due to technical limitations, CMA was
unable to detect small variations, while ES could not detect CNVs when
the probe information was unknown. This was observed in scenarios 1 and
3 (above). If many samples are captured by the same known exome probe
set, new tools allow CNV detection by ES
data(Fromer et al., 2012;
Talevich, Shain, Botton, & Bastian,
2016). However, the samples in this study were obtained from different
hospitals. Thus, we could not perform CNV analysis to avoid scenario 3.
Furthermore, the algorithms were unable to detect single exon
CNVs(Sun et al., 2020). In this study,
iw098 of Family WGSI032XH carried an exon 1 deletion of ARID1Bgene. Even if we had reanalyzed the CNV data, the family would remain
undiagnosed. GS was able to overcome this limitation by detecting
different variant types of different sizes in a single test.
ES uses probes to capture the ROI primarily in coding regions. For the
WGSI078XH family, patient iw243 carried a deletion in the 3’UTR ofIGF2 and was diagnosed with Silver-Russell syndrome 3 (OMIM
616489). As there is no probe that covers the deleted region, it was not
detected by ES. Moreover, probe capture is not perfect. The highly
variable region is not well captured because the variant allele might
have low affinity to the probe which could result in allele dropout
(scenarios 7 and 8). Family WGSI072XH (Figure 3a ) and iw028 of
WGSI010XH represent examples of these scenarios (Figure 3b ). GS
simplifies this process by sequencing everything in the isolated DNA,
thereby overcoming the aforementioned probe issues encountered with ES.
Taken together, GS demonstrated a diagnostic advantage over CMA and ES
in large part due to their technical limitations. A bonus is that GS was
able to detect DNA contamination or blood infection. With the ability to
detect pathogenic variants in coding and noncoding regions of the genome
and the steady decline in sequencing costs, GS has significant potential
to eliminate or reduce the burden of a lengthy diagnostic odyssey in
GDD/ID patients.
GS has shown great promise for identifying targeted treatments in NICU
and pediatric intensive care unit (PICU) patients with its ability to
achieve rapid diagnosis (French et al.,
2019; Wang et al., 2020). Here, we show
the clinical utility of GS in GDD/ID with 39.1% (9/23) of those
diagnosed reporting changes to medical management. Potential treatments
were identified for four families and two families stopped unnecessary
medications. Moreover, so far, at least five families have changed their
reproductive plans, resulting in the birth of healthy children. For
undiagnosed GDD/ID, the treatment is limited unless there is an
underlying metabolic disorder, which is reflected in our study cohort.
Although no effective therapies were found for some families, four
families (WGSI023XH, WGSI028XH, WGSI036XH and WGSI052XH) reported that
GS helped to relieve stress. Two of them (WGSI023XH and WGSI052XH)
joined internet patient groups where they could communicate the disease
courses with other families affected by the same disease. In total, 12
out of the 17 families who participated in the phone interview commented
on the positive aspect brought by GS diagnosis.
Though GS might have advantages in structural variation (SV) detection,
we did not identified patients with pathogenic SV in this study. This
might be due to the limitation of short reads from next generation
sequencing (NGS). For the patients remain undiagnosed, long read
sequencing might help finding the causative variations. Furthermore, the
families in this study were enrolled in a consecutive fashion and not at
the same time. Thus, the time between receipt of the CMA/ES data and the
study conclusion was relatively long. Thus, four cases were solved by
reanalysis because of newly reported genes.
In conclusion, GS is a comprehensive method to detect different types of
variants in genomic and mitochondrial DNA, which could reduce the burden
of test selection faced by clinicians. This study suggests that GS has
clinical advantages for undiagnosed GDD/ID patients even with
reanalysis. Finally, because GS provides the most comprehensive level of
coverage, it is possible to refer back to the original GS dataset to
integrate newly discovered gene-disease associations and update clinical
presentations that would further enhance the diagnostic yield.