2

Improving Health Through Genomics

As a foundation for the workshop, an overview of the increasing use of genomics in precision medicine was provided in a keynote address by Euan Ashley, associate dean of the School of Medicine, Roger and Joelle Burnell Professor of Genomics and Precision Health, and professor of medicine,

genetics, biomedical data science, and pathology at Stanford University.¹ A brief discussion followed, moderated by Sarah Wordsworth, professor and university lecturer in the Health Economics Research Centre and Nuffield Department of Population Health at the University of Oxford.

A VISION FOR GENOMICS IN HEALTH CARE

The aspiration for genomics is to be able to provide a genome “for anyone who wants it…when they want it,” Ashley said. To achieve the goal of better care through genomics, an individual’s genome could be a useful part of their medical record so it is readily available to warn of the risk for thousands of diseases and to inform prescribing by providers. In addition, millions of individual genomes from around the globe could be “seamlessly and securely connected” and linked to a “constantly updated anthology of medical evidence connecting genetic variation and risk of disease,” he said.

PROGRESS TOWARD THE GOAL OF BETTER CARE THROUGH GENOMICS

The progress thus far toward realizing the potential of genomics in precision care across the four key areas of cost, accuracy, speed, and implementation and integration was reviewed by Ashley.

Cost

The cost of a high-depth genome over time shows a clear downward trend,² which has been compared to the graph for Moore’s Law.³ When sequencing of the human genome was nearing completion, the cost of a complete sequence was roughly $100 million. Decreasing costs tracked with Moore’s Law until about 2008, when the cost of sequencing a genome began to drop rapidly until leveling off around 2014, at around $500 per genome, he said. However, in the months before the workshop, several companies announced they would soon be able to offer human genomes for around $100.⁴ This trajectory of cost reduction was “unprecedented”

___________________

¹ See all speakers’ full biographies in Appendix B.

² For more information on the cost of genome sequencing over time, see https://www.genome.gov/about-genomics/fact-sheets/Sequencing-Human-Genome-cost (accessed January 9, 2023).

³ Moore’s Law, published by Intel cofounder Gordon Moore, predicted the exponential growth of computer processing speed, specifically, that the number of transistors on a microchip would double about every 18 months, while costs would decline (Moore, 1965).

⁴ See https://www.science.org/content/article/100-genome-new-dna-sequencers-could-be-game-changer-biology-medicine (accessed February 2, 2023).

in the history of technology and medicine, said Ashley. To illustrate just how striking this reduction in cost is, said Ashley, the cost of a Ferrari 458 Spider in 2013 was about $400,000. If the price of this Ferrari mirrored the price of a human genome, the car would currently cost about 5 cents, and would soon be 1 cent.

While the costs quoted for providing a high-depth genome have dropped dramatically, there are significant capital acquisition costs to be considered for any company or organization entering the field. Some sequencing instruments cost more than $1 million, and some companies do include amortized capital costs into the cost of a genome, Ashley noted. Other factors that affect overall cost include technology, applications, and ancillary costs. The recent expiration of the patent for the sequencing-by-synthesis⁵ approach used in clinical sequencing has allowed multiple new companies to enter the field, creating a more competitive pricing environment. Long-read sequencing⁶ technologies are becoming available, which are advantageous for accurate germline and somatic sequencing, noted Ashley. These approaches have the potential to illuminate the “dark elements” of the genome that short-read approaches find challenging to assemble.

There are also counting applications (e.g., liquid biopsy, noninvasive prenatal testing) for which short-read sequencing of cell-free DNA or RNA is more cost-effective (cost per gigabase⁷ of data). Ancillary costs, including analysis costs and cloud compute costs, have not decreased to the extent that sequencing costs have. Human curation costs have remained steady; however, the time required of a human curator has decreased. Most clinical genomes are currently billed in the range of $3,000 to $12,000, Ashley added. Overall, the progress by the genomics community in reducing the cost of providing a genome deserved a “gold medal,” he said.

Accuracy

“Precision medicine needs to be accurate medicine,” Ashley said. As discussed, there are limitations to the ability of short-read sequencing to fully characterize the genome, and many of the actionable disease gene variants identified by the American College of Medical Genetics and Genomics were being missed in standard clinical genome sequencing (Ashley, 2016; Dewey et al., 2014). Over the past decade, however, there have been sig-

___________________

⁵ Sequencing by synthesis utilizes fluorescently labeled nucleotides for all four nucleotide bases in DNA to identify the sequence of a piece of DNA as it is being replicated, or synthesized, within tens of millions of clusters parallel. The DNA is sequenced in small reads, typically 300-500 base pairs (Slatko et al., 2018).

⁶ Long-read sequencing identifies the sequence from single fragments of DNA in separate wells which allows for reads of 30,000 to 50,000 base pairs (Slatko et al., 2018).

⁷ One gigabase equals 1 billion bases (Shchelochkov, 2023).

nificant improvements in the ability to quantify the genome accurately. As examples, Ashley showed the accuracy statistics (F1 scores of 0.9 or higher) from the most recent sequencing challenge hosted by the Genome-in-a-Bottle Consortium of the National Institute of Standards and Technology, and accuracy statistics for three of the top sequencing platforms, noting that, while the technologies each have different strengths, overall accuracy is greatly improved. There are efforts to accurately characterize the full set of more than 6,000 medically relevant genes (Kim et al., 2022). Accuracy is vital for diagnosis; most rare disease genome studies have a diagnosis rate of 30 to 50 percent for previously undiagnosed disease, and most of these findings are actionable (Splinter et al., 2018), Ashley explained. Although the accuracy of genomes has come a long way and continues to improve, there is still more to be done, Ashley said, and he awarded the genomics community a “silver medal” for accuracy.

Speed

The speed of completing a genome has decreased dramatically since the early genomes that took more than 1,000 hours to assemble. In 2012, Kingsmore and colleagues demonstrated the ability to return genomic findings to patients in the neonatal intensive care unit (NICU) in 50 hours (Saunders et al., 2012). In 2014, Ashley and colleagues accomplished the task in 42 hours (Priest et al., 2014). By 2015, Kingsmore had reduced the time needed to 26 hours and was awarded a Guinness World Record for fastest genetic diagnosis (Miller et al., 2015), and in 2018 Kingsmore beat that record with a time of 19.5 hours.⁸ Most recently, Ashley and colleagues described returning genomic findings to patients in critical care settings in less than 8 hours, with the most rapid diagnosis made in 7 hours and 20 minutes (Gorzynski et al., 2022). It is likely that the process can be completed even faster, Ashley said. Using the car analogy again to illustrate, Ashley noted that the top speed of the 2013 Ferrari 458 Spider is 199 miles per hour. If the increase in the car’s top speed mirrored the increase in speed to deliver genome results the Ferrari would go at 32 million miles per hour. It is now possible to “not just sequence the genome, but return a diagnosis, within one nursing shift,” Ashley said, further noting that this is “a major achievement for [the genomics] community” and worthy of a “gold medal.”

___________________

⁸ See https://www.rchsd.org/about-us/newsroom/press-releases/new-guinness-world-records-title-set-for-fastest-genetic-diagnosis/ (accessed December 14, 2022).

Implementation and Integration

Working with the technology company SAP, Ashley and colleagues proposed a framework for using natural language processing to integrate genomes and data from wearable sensors into patients’ electronic health records (EHRs).⁹ As an example, he showed the framework dashboard for a patient in which data from the patient’s genome about a cardiac genetic mutation was available together with the patient’s diagnosis of cardiomyopathy. While wearable devices have been broadly available to consumers and data have just begun to be integrated into the EHR, the ability to integrate genomes into the EHR and make genomic data accessible is very limited because the genomic data remain largely in PDF form. The EHR provider, Epic, is working to enable the exchange of genetic testing results between laboratories and the Epic system, he noted.

Stanford University has launched an in-house whole-genome sequencing service and provides “a whole-genome backbone for all genetic testing,” said Ashley. This allows for in silico genetic panels and improved detection of structural variants, he said. Having the full high-depth genome also facilitates the automatic generation of polygenic risk scores. For example, clinical providers will soon be able to easily access a patient’s integrated clinical and genomic risk scores for cardiovascular disease. Similarly, pharmacogenomics is beginning to be integrated into the EHR.

A challenge for the integration and implementation of genomics is that diagnostics remain significantly undervalued relative to therapeutics, Ashley said. He observed that the COVID-19 mRNA vaccines were the fastest medicines ever developed, but COVID diagnostics lagged far behind. Another challenge is determining who should pay for genomic diagnostics: health care systems, payers, pharmaceutical companies, or self-pay. Finally, there has been limited integration of genomic data with health care data in the delivery of care.

Overall, Ashley awarded the genomics community a “bronze medal” for efforts thus far to integrate and implement genomics in precision care. There are still challenges to address, particularly regarding inclusion and diversity in clinical genomic research (especially rare disease studies) and the obligations of researchers to study participants who remain undiagnosed when a clinical genomics study concludes (Halley et al., 2022a,b).

GENOMIC MEDICINE: PAST, PRESENT, AND FUTURE

Historically, genomic medicine was limited to cancer applications (genetic testing for inherited risk) and rare diseases (molecular classifica-

___________________

⁹ See https://www.biorxiv.org/content/10.1101/039651v5.full (accessed December 14, 2022).

tion of known disease and undiagnosed disease), Ashley said. The current landscape is much broader and includes early detection and tumor sequencing for cancer; analysis of circulating cell-free DNA for noninvasive prenatal testing, transplant rejection, and liquid biopsy; selective panels for pharmacogenomics; and organism sequencing for infectious diseases. He said the future holds the promise of

• Mendelian, oligogenic, and polygenic architectures for rare diseases;
• circulating cell-free RNA in addition to DNA;
• full comprehensive tumor immunogenomics and detection of minimal residual disease for cancers;
• EHR integration of pharmacogenomic data;
• full microbiome sequencing; and
• integrated polygenic risk scores for common diseases.

“We have come a long way in a short time,” Ashley concluded. An individual’s genome can now be completed in a matter of hours for the cost of several hundred dollars. Accuracy has improved significantly, owing in part to long-read sequencing, and there are numerous potential near-term applications. However, there is still work to be done to realize the full potential of genomics in precision health care. First, for genomic medicine to be equitable, he said, individuals need to be empowered to obtain their genomic information when they want it, and the diversity of clinical cohorts must be improved so the genetic diversity of the global population is represented. Second, patient-centered planning and decision making could be useful for providing equitable precision health care. Third, genomics could be integrated at the point of care. Finally, payers need to recognize the health benefits and cost savings of implementing genomics in the delivery of health care, he said.

DISCUSSION

“The genome is not one size fits all,” said a workshop participant, noting that while the information gleaned from the genome may be useful for different purposes across the life span, everyone does not necessarily need genome sequencing right now. Perhaps, the participant suggested, the focus should be on when genome sequencing is needed and figuring out how to implement it. From a public health perspective, Ashley said, the right approach is probably a stepwise approach where what could be done is thought about first as well as who would get the most benefit. Taking a system perspective, he further clarified that if the field’s responsibility is to do the most good, it should work from there to something aspirational like a genome for everyone who wants it to empower those who want to have access to this type of health information.

David Ledbetter, chief clinical and research officer at Unified Patient Network, Inc. and professor in the department of psychiatry at the University of Florida, commented that the United Kingdom (UK) seems to be well ahead of the United States on the implementation of genomics in clinical care. For example, the UK recently announced the launch of a 5-million-person polygenic risk score clinical study.¹⁰ The UK is at the forefront of clinical genomics, and this is attributable in part to the fact that the UK has a single-payer system, Ashley suggested. Having a national health service allows for the uptake of genomics rapidly and at scale. The UK’s progress in implementing genomics also stems from the efforts of individuals who have championed genomics, whether at the grassroots level, in political or governmental leadership, or leaders of the medical community, he added.

In response to a workshop participant’s question, Ashley said that even as genomic technology advances, the patient’s personal family medical history will always remain very important across all aspects of care. A genome can help clarify elements of family history that are forgotten or not known and better empower patients to understand risk.

___________________

¹⁰ https://www.genomeweb.com/microarrays-multiplexing/uk-researchers-aim-discern-new-polygenic-risk-scores-5-million-genotypes#.Y63ndnbMI2w (accessed December 29, 2022).

This page intentionally left blank.