Note: This story has been updated to clarify which version of HL7 FHIR Epic Systems currently supports.
CHICAGO – Although the US Food and Drug Administration (FDA) provided some long-sought clarity in 2022 on how it would regulate clinical decision support and in vitro diagnostic software, technology developers and healthcare organizations still struggled with how to integrate genomics data into clinical practice.
It will likely take more progress in standards development and harmonization — not to mention cooperation by healthcare providers and software vendors — to achieve the potential of precision medicine. It may also require additional government funding like that provided by the European Union late last year to standardize next-generation sequencing workflows in hopes of making precision medicine a standard of care in oncology.
Or it could take a disruptive force from outside the traditional world of bioinformatics, which is what Amazon Web Services is attempting to be with the November launch of Amazon Omics.
No path will be easy, but the recently concluded year saw significant progress in the bioinformatics world in both the clinical and research realms.
Data standards
A longstanding hurdle to integrating genomics into clinical practice that continued in 2022 is lack of clarity when it comes to data standards — or even lack of standards in the first place. A key challenge is matching genotypes with phenotypes in EHRs, according to Charles Jaffe, CEO of standards development organization Health Level Seven International (HL7).
While HL7 has taken on genomics by including an add-on to its Fast Healthcare Interoperability Resources, or FHIR, standard since 2019, FHIR Genomics has not been widely or uniformly adopted, some critics have said.
Nephi Walton, associate medical director of precision genomics at Intermountain Healthcare in Salt Lake City, said that the FHIR Genomics module component still is not mature and has "significant error in it." He took issue with how the standard classifies genotypes and its inability to handle compound heterozygotes and de novo variants.
Compounding that problem is that major EHR vendor Epic Systems initially adopted a variation of FHIR that HL7 later abandoned and therefore is no longer supported by others. Epic has since updated it software to support FHIR version 2.5.1 interfaces, but a useful standard, Walton said, must be able to keep up as genomics evolves.
The standards world did take a step forward at midyear when the International Organization for Standardization (ISO) officially approved Phenopackets, a standard for sharing disease and phenotype information for diagnosing and treating rare and hereditary diseases including cancer. Phenopackets may prove to be an important tool in representing phenotypes and genotypes in machine-readable form in EHRs so the information can be better interpreted, but it remains to be seen how widely this standard will be adopted.
However, perhaps the most important medical coding system in the US, because it is so closely tied to billing and reimbursement, is proving to be inadequate for the age of precision medicine. Many genetic and rare diseases do not have corresponding International Classification of Diseases 10th Revision (ICD-10) entries, making it difficult for healthcare providers to bill for genetic testing or for treatment of certain rare genetic diseases.
Coders often will pick the ICD-10 code that will produce the highest allowable reimbursement for a given patient's insurance plan, but that might not be medically adequate.
"ICD-10 is really a classification, and it's not intended to specifically identify every clinical concept," explained Sue Bowman, senior director of coding policy and compliance at the American Health Information Management Association (AHIMA).
"It's not going to have a unique code for every medical concept because classifications, by definition, group things into buckets," Bowman said. Many rare conditions do end up in what ICD-10 considers "residual" categories.
Clinical decision support
Somewhat related to data standards, but with its own unique challenges, is the issue of clinical decision support.
In the US, the October guidance from the FDA discussed which kinds of software the agency will consider medical devices and what types are exempt, thanks to changes the 21st Century Cures Act made to section 520 of the Food, Drug, and Cosmetic (FD&C) Act. While the document is nonbinding, it indicates current FDA thinking on how it will regulate clinical decision support software and what kinds of technology should be submitted to the agency for 510(k) review.
Regulatory attorneys suggested that more makers of clinical decision support software in genomics and in vitro diagnostics may now have to submit their products for FDA review. Kelly Sager, VP and general manager of clinical decision support solutions at Beckman Coulter, said that the guidance "provides yet another level of clarity when it comes to CDS software function."
Across the Atlantic is a program called Integrated and Standardized NGS Workflows for Personalized Therapy (Instand-NGS4P). The European Commission late last year awarded 11 contracts collectively worth nearly €4.8 million ($5.1 million) for the second of three phases of this program that aims to integrate NGS-based precision medicine into oncology practice across Europe.
Instand-NGS4P is intended to develop workflows for integrating, standardizing, and analyzing data from cancer gene testing, pharmacogenomics testing, and medication databases to support clinical decision support at the bedside. The current Phase 2 is the proof-of-concept stage, following design proposals in Phase 1. The third and final phase, to start in March 2024, will involve implementation and testing of technology in real-world clinical settings.
"Our goal [is] to really cover the complete diagnostic workflow, starting from the sample collection from the patient to isolation of nucleic acids to library preparation, sequencing, bioinformatics, and then up to producing the proper reports for decision-making at the bedside," said the program's director, Kurt Zatloukal, professor of pathology at Medical University of Graz in Austria.
The Austrian medical school is one of seven Central European hospitals that make up a buying coalition that are already employing NGS for research and diagnostics. The group also includes patient advocacy groups and participants in programs such as the European Life Sciences Infrastructure for Biological Information (ELIXIR).
The hospitals are hoping that the contractors develop technologies they and other hospitals want to purchase in the future. Because the technology will be used in clinical settings, it must comply with the EU's In Vitro Diagnostic Regulation (IVDR).
Contractors, including BC Platforms and Congenica, expect that their development within the parameters of this program will help them achieve IVDR compliance, and thus find their products ready to go to market by the time Instand-NGS4P ends in 2025.
New players
A disruptor like Amazon might be able to push bioinformatics forward, or it might not, given the track record of Big Tech in life sciences and healthcare. Google and its parent company, Alphabet, have found a modicum of success with Verily Life Sciences, but Google, Microsoft, and Apple have all had high-profile failed ventures that involved healthcare data. Amazon itself shut down its Amazon Care telemedicine service in 2022.
With this hindsight, Taha Kass-Hout, chief medical officer of AWS, frequently alluded at the time of the Amazon Omics launch last November to taking the "heavy lift" of technical processes off the backs of bioinformaticians and researchers.
"We take on the onus of provisioning, managing, scaling, and securing the entire infrastructure," said Kass-Hout, who also serves as VP of machine learning for AWS. "Taking all that out of the equation lets scientists be scientists and work on and focus on what they [know] best, such as scientific discovery and delivering better care for patients and innovating on therapeutics and diagnostics."
Amazon Omics is made up of three components: "omics-aware" object storage for raw sequencing data; Amazon Omics Workflows for processing raw sequences from FASTA, FASTQ, BAM, and CRAM files; and Amazon Omics Analytics, which adds structure and annotations to plain text formats like VCF to simplify variant and mutation queries, according to Kass-Hout. Customers can use the pieces either individually or all together.
Users can combine their own data with dozens of publicly available datasets. Amazon Omics supports workflows written in Nextflow and Workflow Description Language (WDL) and follows the HL7 FHIR standard.
Also worth watching are some 2022 startups. Nest Genomics is developing genetics-focused clinical decision support software that links into existing electronic health record (EHR) systems. Form Bio is a spinoff from Colossal Biosciences to commercialize the latter's computational biology technology platform.
$500M for NIH's BICAN
Not strictly a bioinformatics program, but worth watching for the sheer size and scope is the US National Institutes of Health's (NIH) BRAIN Initiative Cell Atlas Network (BICAN). This $500 million initiative, launched in September and funded through a series of grants over a five-year period, is meant to support the creation of cell atlases of human and nonhuman brains and maps of cell interactions to inform research into neurological diseases. Awardees will build the atlases and maps through single-cell sequencing, noninvasive medical imaging, and advanced bioinformatic analysis.
Part of the National Institute of Mental Health's ongoing Brain Research Through Advancing Innovative Neurotechnologies (BRAIN) Initiative, BICAN is the third phase of an ongoing effort to create a cellular map of the brain, following a pilot phase and the BRAIN Initiative Cell Census Network (BICCN), which kicked off in 2017.
The largest share of the money, $173 million in five grants, is going to a coalition led by the Allen Institute to build the first-ever complete cell atlas for the human brain, as well as atlases for the brains of marmosets and macaques. Another group, led by the Salk Institute for Biological Studies, was awarded $126 million to create the Center for Multiomic Human Brain Atlas, which will attempt to map the human brain on a cellular level to understand how neurotypical brains operate and age.
The Center for Multiomic Human Brain Atlas, led by Joseph Ecker, director of the Salk Institute's Genomic Analysis Laboratory, will create new analytics software, incorporating new as well as existing data pipelines from the participating institutions including pipelines created for the earlier BICCN project.