A team led by researchers from the University of California, San Diego has found that sequencing the exomes of a cohort of patients with a genetic disease led to a change in diagnosis for almost 10 percent of them. The analysis also identified likely disease-causing genes for almost 20 percent of patients, and a small list of candidate variants for about three quarters of the group.
The study, published last week in Science Translational Medicine, is among the first to examine the utility of exome sequencing in a cohort of patients, rather than presenting the results of a single successful case where exome sequencing led to a diagnosis.
Because of the nature of the cohort — the study only included consanguineous families with recessive disease, and only those families that had tested negative for known candidate genes — the results can't be extrapolated to exome sequencing in disease diagnostics in general, but it proves its utility for this type of patient group.
"The main finding of this work is that whole-exome sequencing is beneficial over individual candidate gene sequencing in identifying mutations in genes not previously suspected in a given patient," the authors wrote.
According to Joe Gleeson, a professor in the department of neuroscience and pediatrics at UCSD and the senior author of the study, cost reasons alone make it likely that exome sequencing will replace candidate gene sequencing soon and become a standard clinical diagnostic test. "If you are going to order genetic testing, you may as well start with the exome," he said.
Gleeson's lab has been studying the genetics of pediatric-onset neurodevelopmental diseases, which include intellectual disability, epilepsy, autism, structural brain diseases, and neuromuscular disorders. Only a small percentage of patients currently receive a genetic diagnosis, which "really limits care," he said. "We started this study with the goal of finding new genes."
The researchers are focusing on consanguineous families with recessive disease so they only need to consider homozygous variants and can exclude the large number of heterozygous variants that exome sequencing identifies, a challenge with this method.
So far, they have recruited about 2,000 consanguineous families from the Middle East, and last week's publication presents the results of their pilot project.
For that project, the researchers started with 188 families, each with at least two children affected by neurodevelopmental disease. They excluded 40 families because they found them to have mutations in known disease-causing genes associated with the patients' initial diagnosis.
Testing for mutations in those known genes took two staff members approximately one year, Gleeson said, a huge and costly effort that can no longer be justified in the face of falling prices for exome sequencing. At the time of the pilot study, sequencing an exome cost about $5,000 per patient, he said, which has now dropped to $1,000, while the cost of candidate gene sequencing has remained the same.
Thus, Gleeson's lab has now completely stopped candidate gene sequencing, viewing it as "complete waste." It now generally performs exome sequencing on one affected family member even when a mutation in a known disease gene is suspected. In case of a negative result, the exome of a second affected family member is sequenced.
Of the remaining 148 families in the pilot study, the team excluded another 30 because a genome-wide parametric linkage analysis revealed a single linkage peak for them, suggesting the use of a more direct method than exome sequencing to identify the mutant gene.
They then sequenced the exome of one patient in each of the 118 remaining families, obtaining an average read depth of more than 10x for 96 percent of the exome. Sequencing was conducted at the Broad Institute using the Agilent SureSelect All Exome 50 Mb kit and the Illumina GAIIx or HiSeq 2000 with 75-base paired-end reads and 30x sequencing depth per exome.
On average, the researchers identified 26,000 variants per patient, which were filtered and prioritized according to the assumed recessive disease model, the type of mutation, amino acid conservation, predicted protein damage, and relevance of the gene to the disease. Variants that made it onto the final list — between four and 21 per family — were validated by Sanger sequencing and tested for segregation in the family.
In 10 patients, or 8 percent, the researchers detected mutations in genes already known to cause a neurodevelopmental disease, which they had missed earlier because the patients' symptoms did not suggest testing for those genes. In these 10 cases, the finding changed their initial diagnosis, either modifying or correcting it.
Having a correct diagnosis helped all 10 families at some level, Gleeson said, for example by providing a better understanding of possible future problems or enabling better genetic counseling, carrier testing, and prenatal testing. In two cases, the new diagnosis also led to a change in medication.
In 22 of the patients, or 19 percent, the researchers identified a single homozygous variant in a gene that had not been previously associated with disease. For two of these patients — one with microcephaly, simplified gyral pattern, and insulin-dependent diabetes; the other with Joubert syndrome — they showed that the genes they identified — GFM2 and EXOC8 — are "rational candidates" for the diseases.
Gleeson said that for some of the other 20 patients, the researchers have now found additional patients with mutations in the same genes, making it likely that they are causative. Researchers are now increasingly sharing their results in collaborative networks, he said, increasing the odds to find unrelated families with mutations in the same genes.
In 86 patients, or about three quarters, the researchers identified between two and 10 variants, "some of which are good disease-causing candidates," they wrote. Gleeson said they are now pursuing a number of strategies to whittle these down to a single disease-causing variant.
The most straightforward way is to sequence the exome of a second affected family member, which eliminates the vast majority of variants. "That's been our most effective strategy," he said, increasing the percentage of families with "a very good single candidate gene" from 20 percent to about 80 percent.
Another strategy is to filter the data against public databases, such as dbGAP, to further eliminate variants that are not disease-causing.
The third approach is to look for segregation of the mutations in the family, which is possible to do for a couple of variants, but not feasible for dozens of variants, Gleeson said.
Overall, the study resulted in a change in diagnosis for only 8 percent of cases, which might not be seen as a great success. But the results are skewed because the study only included cases where candidate gene sequencing had been unsuccessful. "If we had included those 40 cases where we found mutations [in candidate genes], the numbers would have looked totally different," Gleeson said.
Also, a correct diagnosis for 8 percent of patients could still make the case for exome sequencing as a diagnostic tool. "If the medical community feels that that kind of outcome is worthwhile, then I think it would support exome sequencing in the clinic currently," according to Gleeson.
Many doctors are already enrolling their patients — especially "medical enigmas" where all existing tests have come back negative — into exome sequencing research studies, he said.
The data interpretation is still a bottleneck, preventing exome sequencing from becoming a routine diagnostic tool. "It's not hard to generate the data," Gleeson said. "The harder part is the interpretation." For example, even if mutations fall into known disease genes, caution is needed because many previously reported mutations are false positives.
"Over time, we will see the research and clinic become ever more closely aligned," Gleeson said. "This is an instance where the research will lead the clinic, much in the way that [copy number variation] analysis was initially started as a study and has turned into a widely used clinical diagnostic tool."
In a few families the researchers have studied so far, exome sequencing did not turn up any candidate variants, and for those families, whole-genome sequencing will be the next step. While it will be difficult to analyze variants in introns and regulatory elements for some time to come, "I think that the software will improve as we understand the epigenome and regulatory elements better in the coming years," he said. Another reason to switch to whole-genome sequencing is that exome sequencing does not reveal copy number variations, he added.
Gleeson's lab is currently applying exome sequencing to the remainder of its cohort, while wrapping up its recruitment of consanguineous families.
It is also now using exome sequencing to look for somatic mutations in patients with neurodevelopmental diseases. "There is a big interest now in de novo mutations that are zygotic — the mutation has occurred in the egg or the sperm — but also post-zygotic, which is after the embryo has started to develop, and I think exome sequencing can play a huge role there," he said.