Genomics in Clinical Practice: Lessons from the Front Lines

Science Translational Medicine  17 Jul 2013:
Vol. 5, Issue 194, pp. 194cm5
DOI: 10.1126/scitranslmed.3006468


The price of whole-genome and -exome sequencing has fallen to the point where these methods can be applied to clinical medicine. Here, we outline the lessons we have learned in converting a sequencing laboratory designed for research into a fully functional clinical program.


Three years ago, the Medical College of Wisconsin launched a Genomics Medicine clinic in collaboration with Children’s Hospital of Wisconsin and Froedtert Hospital with the goal of using whole-genome sequencing (WGS) to elucidate the etiology of undiagnosed diseases in patients who had exhausted all standard care options. We used exome sequencing for our first pediatric case, Nic Volker, and had to develop analytical tools to evaluate ~18,000 variants. Subsequently, 23 pediatric patients and 2 adult patients have had their genomes sequenced, requiring that we develop a strategy to manage >400,000 variants. So far, we have been able to obtain a definitive diagnosis in 27% of cases (7/26) (Table 1). Our initial concerns were cost and data accuracy, but the major challenges turned out to be the logistics of delivering genome sequence information to clinicians, how clinicians use the data, and how patients and their families deal with the secondary (incidental) findings. Here, we discuss the challenges of building our clinical genomic medicine program and the lessons we have learned. We hope this Commentary will help other groups to implement their own clinical programs or, at a minimum, help to initiate a conversation about the application of WGS in clinical care.

Table 1

Results of WGS at the MCW Genomics Medicine Clinic for 23 Pediatric (CHW) and 2 Adult (FH) Cases.

View this table:


Today, in clinical medicine, elucidating genetic causation is commonly pursued for rare diseases. In contrast, risk assessment for common diseases is based on family history even though this is often incomplete or inaccurate (1, 2). Genetic diagnosis and risk assessment can be achieved through traditional genetic testing (at the individual gene or gene panel level), but there are limitations. WGS is a more comprehensive approach but is not regularly used because testing is perceived as expensive, often is not covered by insurance companies, and is not always thought necessary based on clinical decision-making guidelines. As a result, most clinicians are unlikely to view WGS as a useful clinical tool. When considering WGS, clinicians may ask: If I find the gene mutation and there is nothing I can do that will change the clinical course, why do the test? In other words, are the results from gene tests or WGS clinically actionable?

The term “clinically actionable” has a wide range of definitions (3); this lack of clarity has the potential to limit viable clinical use and complicates the development of uniform policies. At the strict definitional level, actionable is “information that allows a decision to be made or action to be taken” ( In this regard, using WGS to make the diagnosis, or a better diagnosis, can be considered “actionable”. Our clinical program is built around the principal idea that making a diagnosis is essential, even in cases where the results may not lead to better treatment. The long-standing experience of clinical genetics has been that a diagnosis can change the management plan for a patient even in the absence of specific therapeutic choices (4).

For example, our group was considering a liver transplant for a critically ill child with severe liver disease (5). To aid in diagnosis, next-generation sequencing was performed and the clinical team learned that the child carried two mutations in the TWINKLE gene and that the liver manifestations were likely to be the result of a newly identified gene variant. TWINKLE mutations are known to cause progressive neurological deterioration (6), and the patient was beginning to show neurological symptoms. In consultation with the parents and the clinical team, a liver transplant was not performed, given the poor long-term prognosis, and the child died at 6 months of age (5). In this case, although there was no improved measured clinical outcome, the parents were reassured that the disease was incurable and that a liver transplant would not have prevented the child’s death. The potential donor liver was available for another child and our patient did not suffer from a futile liver transplant. In another case, WGS uncovered a rare autosomal recessive gastrointestinal (GI) syndrome and, although there is no cure for the disease, the diagnosis allowed us to improve the patient’s quality of life by providing anticipatory guidance. The information also enabled us to provide an accurate recurrence risk to the parents and options for prenatal testing, and the parents derived comfort from knowing the diagnosis. Although physicians and insurance companies have traditionally used clinical utility to determine the value of a test, the concept of personal utility is now being applied to genomic information (7) and, we believe, has a role in clinical medicine.


Clinical decision-making is especially challenging when there are no data to support the role of a specific genetic variant in a particular disease phenotype (so-called variants of uncertain significance, VUS). Our first pediatric patient, Nic Volker, had severe GI disease and had undergone multiple procedures under anesthetic. Whole-exome sequencing revealed a mutation in the XIAP gene. XIAP mutations are associated with an increased risk of lymphoproliferative disease (LPD) (8), but no scientific evidence linked these mutations to Nic’s GI symptoms. Although reconstituting Nic’s immune system with a bone marrow transplant (BMT) would reduce the risk of LPD, this intervention carried significant risks and it was not clear whether it would improve Nic’s GI disease. Fortuitously, while we and the family were considering the decision, molecular research was published supporting an association between XIAP and a gene in a signaling pathway known to be altered in GI disease (8). We proceeded with a cord blood transplant that resulted in resolution of Nic’s GI symptoms. It is uncertain what course we would have taken if there had not been the compelling need for a BMT based on the molecular diagnosis or what other clinicians would have done with the same data we had. There have been two other cases in which a VUS was found to be a clear contributor to disease etiology after the publication of new research data. With WGS, we have improved our rate of diagnosis to 27% for rare or undiagnosed diseases (Table 1). We wonder how many of the 73% of unsolved cases after WGS have the correct VUS but no supporting data to prove causality. There is an urgent need to solve this problem. One simple solution, at least for rare diseases, is to increase sharing of sequenced human genomes (7). A recent White Paper from the Broad Institute creates a potential framework ( for making this happen. Of course, many complex issues remain to be resolved, including release of clinical data from hospital sites, compliance with HIPAA regulations, and informed consent. An alternative would be to share data in a federated rather than a centralized manner, which would make it easier for hospitals and clinics to share sensitive patient data.


A key consideration is what types of data will be returned to the clinician and to the patient (or patient’s family). Once the clinic or laboratory determines what data they will return, an analysis pipeline (consisting of different analysis tools, algorithms, and computational steps) can be built. However, because the data sets are so large (up to 1 terabyte per genome) the bulk of the work must be automated (computer-assisted analysis, CAA). Currently, there is no standard platform, thus requiring institutions to establish their own CAA or link together a variety of commercial and open-source software packages and data sets that contain information about disease-causing variants. Genomics laboratories new to the clinical arena need to identify state and local laws and hospital guidelines that affect storage of clinical and laboratory data and samples and the types of genomic sequence data that need to be disclosed to the patient.

There remains no clear consensus about what data must be stored from WGS and exome sequencing, nor are there clear guidelines about the storage of secondary (incidental) or tertiary data. Several groups are working on guidelines. The American College of Medical Genetics (ACMG) recently published guidelines regarding the return of secondary (incidental) findings to patients and their physicians (; see also Editorial by S. Kingsmore, this issue). These guidelines define what results a clinical laboratory should return to a physician ordering a test for whole-exome sequencing or WGS. The guidelines state that certain secondary results should be reported “without reference to patient preferences” (9). These guidelines have not been uniformly accepted (1012) but do set the standard that certain secondary findings should be returned to the patient’s physician and patient. In our opinion, the issue of the return of secondary results should be determined by the physician and patient before ordering the test.

The following questions are particularly important to consider as they will have a profound impact on how the genomic medicine program and CAA pipelines will be developed and on cost. What types of data will be returned to the patient or to the parents in the case of children? For example, should parents receive information about their child’s status for the Alzheimer’s disease-associated allele ApoE4? If the only data to be returned are directly related to the disease for which the patient is being treated, then CAA pipelines and data storage are simplified. In our clinic, the patient or parents (in the case of minors) are given the choice regarding what type of results they would like disclosed in addition to data directly related to the patient’s illness (7). We acknowledge that this approach, to incorporate patient preferences into data return, is not universally supported (13). We also offer to provide an annual follow-up clinical visit, where the patient returns and we reanalyze the genomic data. If patients and parents express an interest, we will confirm the types of results that they are interested in and reanalyze the data from the nucleotide-called sequence. These choices have had a significant impact on our CAA pipelines and data storage decisions and have increased our overhead costs. Many have suggested that the easy alternative is to resequence the patient’s genome in the future as sequencing costs decrease and analytical tools improve, but genome resequencing is not without its problems.


Technology is not without its limitations. There are errors or uncertainties introduced into the sequence from the sequencing machines and analytical algorithms, and there are errors in the reference data we align the new sequence against. Rosenfeld et al. report that differences in the generation and analysis of genome sequences result in a 4–14% range in the number of variants called in the same sample (14). This has implications for clinical use. If the original sequence data are not stored, it will be impossible to determine the source of error in the future. Hopefully, future storage formats will retain key data elements, enabling far less data per sample to be stored. Given the differences among variant calling programs, some groups have chosen to use a combination of programs. But what should be done with the variants called by a single program, many of which are likely to be accurate (14)? In a clinical setting, each analytical process requires a documented validation step (15); whenever any process is altered, there has to be revalidation. Our tertiary analysis platform developed in house, CarpeNovo, is updated and revalidated every 6 months (a process that takes 4 to 6 weeks). We had to develop and write standard operating procedures (1192 pages) for all parts of the process in order to obtain CLIA (Clinical Laboratory Improvement Amendment) certification and CAP (College of American Pathology) accreditation, a legal requirement for the reporting of clinical results.

Given the challenges outlined, it may seem that WGS should be outsourced to a few expert centers. However, as the cost of WGS decreases and the speed of sequencing and analysis increases, it is likely that most hospitals and diagnostic laboratories will implement WGS. Local implementation has the advantages of speed, compliant data storage, easier interactions with the clinical team, and the availability of other types of data—e.g., gene expression (RNAseq) data and genomic sequencing of the gut and skin microbiota.


There is a critical, unmet need for a comprehensive clinical and phenotype profile for each patient that can be integrated with the variants discovered in the patient’s genome. Current electronic health records (EHRs) are not designed to handle the complex medical data for patients with rare or undiagnosed diseases, who have visited many different clinicians and hospitals and have numerous clinician reports on paper, laboratory printouts, and other communications. We have developed an in-house software (ClinMiner) to integrate data from multiple sources, both paper and electronic, and to standardize the data using SNOMED CT, LOINC, and RxNorm ontologies. Data entry, query systems, and patient reports with timelines and charts provide the clinician with easy access to comprehensive clinical profiles. Until EHRs are designed to meet these needs, sequencing groups along with their clinical partners will need to develop similar tools.

After variants are called, the genomics data must be integrated into the clinical workflow (Fig. 1). Simple questions become incredibly complex, for example: (i) What results will go into the medical record? (ii) How will the report be generated? (iii) What can be billed and how will this be paid for? (iv) What is the clinical oversight for a program that incorporates genetic counselors, clinical geneticists, pathologists, clinical laboratory personnel, bioinformaticians, data analysts, and the ordering clinician? In our opinion, the solution is to develop a service line that crosses departments and integrates the various personnel with the appropriate expertise to improve patient care. Developing a reimbursement strategy is also a major challenge, as Medicaid, Medicare, and many insurance companies will not pay for WGS, leaving patients or their providers with the bill. How each provider manages this challenge in the face of an evolving health care system requires careful consideration. We have had some success by demonstrating that the cost of sequencing one whole genome is more economical than ordering multiple genetic tests.

Fig. 1

The long and winding road. Shown is the linear flow for a patient undergoing WGS in our genomic medicine clinic. Patients can enter the program at several points. A typical patient is referred by their clinician or is already in our hospital system. The first steps are collecting data from their records, a visit (outpatient or inpatient), genetic counseling, and discussion and determination of what data they would like returned to them. Once the patient has consented to clinical care, their genome is sequenced. Subsequent analysis focuses on the primary clinical reason the patient was admitted. Any secondary findings that the patient has requested are returned later. We then invite patients to come back annually for clinical follow-up. Ongoing care depends on the clinical presentation, severity of symptoms, and the outcome of our assessment. We also allow patients to enter the system with their own genome sequence in hand as long as the sequence was generated in a CLIA/CAP-accredited laboratory.



For rare diseases and undiagnosed diseases, the economic case for WGS is relatively easy, as the patient and their family have gone from clinician to clinician, hospital to hospital, looking for an answer, while accumulating huge bills. What about the use of WGS for common diseases such as type 2 diabetes, heart disease, cancer, or for pharmacogenomic testing to determine which patients will have a beneficial or adverse reaction to a drug? Given the potential development of disease in everyone and the probabilistic nature of genomic data, there is a concern that follow-up analysis for secondary findings will increase the cost of clinical WGS. As was the case for MRI, the clinical teams, the hospitals, and the insurance companies will need to learn to balance deployment of new technologies, including WGS (16). Economic considerations should not be the dominant reason for delaying implementation of WGS in the clinic.

What about WGS for preventive care? Disease prevention is not considered economical because many people have to be screened to find the few at risk (17). Because everyone has a risk for developing at least one disease and WGS can expand the accuracy of family history, we think that WGS is economical in preventive care. Every genome sequenced offers value, as it provides a reference for an individual’s family, as well as for the general population. In our opinion, as medical and genomic information becomes better integrated, the combined data set will be of even greater value to the patient, their family, and society. However, the economic advantages of WGS still need to be clearly demonstrated, and the delay may have spurred the appeal of direct-to-consumer genetic testing.


We need to continue to explore and test how WGS findings can be conveyed to patients and their families and how follow-up studies should be conducted. The ethical, legal, and social implications could be limiting factors for the general uptake of genomic medicine and require considerable attention. Early reports from direct-to-consumer genetic testing (18) show that there has not been extensive psychological harm done by the information provided to customers. But will these findings hold, particularly for people who have not actively pursued results from their genome, unlike those using direct-to-consumer testing. Considering the known gaps in existing legislation (19), uncertainty remains about the ways in which life, disability, and long-term care insurers might use genetic information. Finally, privacy concerns legitimized in recent publications (20) require additional attention when contextualizing genomic data as “anonymous.” Engaging the public is an essential element for the success of genomic medicine. Public confidence regarding the management of genomic data and clarity regarding how such data can be used and shared will continue to be crucial as adoption of WGS increases.

Becoming a fully integrated genomic medicine clinic has many steps within care delivery that are specific to each institution, such as departmental boundaries, local politics, or silos that could be the biggest barriers to implementation. Education of providers about the use of genomic data is also a challenge. Although we have established a fully integrated genomic medicine clinical program, not every patient with a rare disease will benefit. We have had our share of successes (27% successful diagnoses), potential diagnoses (34%), and failures (39%).

We believe that WGS will improve the practice of medicine. Yet, despite the clear advantages it provides over other tests, a genome sequence is still just another data set. What we can extract from this data set will require further integration of genomic and phenotypic data and clinical trial results. Our first 26 clinical cases have been vital for defining and refining our program. We encourage other institutions to set up their own genomic medicine clinics because there is clear medical value in the human genome sequence that can be excavated to improve the diagnosis and treatment of human disease. The use of WGS in the clinic may still be debated, but we challenge the fence sitters to do 20 cases of their own and see whether WGS adds value to clinical decision-making.


View Abstract

Navigate This Article