June 21, 2023 | At the Bio-IT World Conference & Expo last month, members of the Genomics England team shared about the organization’s history, vision, and what it’s working on now.
Parker Moss, the Chief Partnerships Officer at Genomics England, recalled the group’s genesis with the 100,000 Genomes Project beginning in 2012 gathering genomes from cancer and rare disease patients. When that goal was met, Genomics England had to reconceive its vision and purpose, Moss said. The forward-looking structure for the organization is to create symbiotic research and clinical components.
For the clinic, Genomics England set up a UK National Health System clinical governance framework. A single consent form for every hospital in the country allows the organization to recruit from everywhere. Samples are gathered from the clinic, sequenced on the Human Genome Campus, and processed on custom bioinformatics pipelines. Lists of clinically-annotated variants are shared with NHS doctors and tumor boards to make treatment recommendations. This pipeline, Moss said, processes thousands of patients each month.
The clinical data not only go back into NHS for treatment decisions, they also flow into the research side of Genomics England, populating a huge research database that is in the process of moving to Amazon Web Services. Genomics England accredits and equips researchers to work with those data. Moss reported that ten of the fifteen top pharma are using the database for target discovery, patient stratification, and biomarker discovery.
All of these data need to be consumable, Moss knows. Genomics England has partnered with Amazon Web Services and LifeBit to build a platform that allows both technical and non-technical users to explore data, run pipelines, and create genotypic or phenotypic cohorts—all within the research environment. Users can take aggregate data out, but not patient-level data. Users’ can bring their own data in from AWS, and their own pipelines can be containerized and brought into the GEL workspace as well. Jupyter Notebooks or GitHub applications can also be used. “It’s an intuitive user interface,” Moss said.
But the “emotional beating heart” of the company, Moss emphasized, is when researcher findings get funneled back to the clinic. “This is more in rare disease than cancer, where rare disease diagnostics is such a challenge. We have academic and commercial partners around the world studying these variants of unknown significance in undiagnosed patients, and they can annotate and recommend diagnoses back into the clinic,” he said.
That process is mediated by Genomics England, Moss explained, checking that recommended diagnoses are ethical and accurate, but the system has provided diagnoses for thousands of patients—a 25% diagnostic yield for rare disease. “It’s really the hopes and dreams factory of these young parents. Imagine you have a young child with a severe rare disease. You’re really hoping that some kind of academic somewhere in the world is going to figure out the silver bullet and at least get them a diagnosis, hopefully a therapeutic.”
It’s an infinity loop, Moss said, by way of illustration. “The more clinicians that find that they are getting diagnoses from academics, the more they’re keen to refer their patients to the genomic medicine service that we operate with the NHS. That means we have more data, which attracts more researchers, and that increases the rate of diagnostic discoveries. It’s a self-reinforcing loop.”
Growing the Dataset: Digitizing Images
GEL’s expanding, growing dataset began with cancer and rare disease genomic data from the 100,000 Genomes Project. GEL has since added a large covid-19 dataset gathered from acutely and mildly ill patients, which now, Moss notes, serves as a nice control set to the rare disease and cancer data.
From there, these genomes—nearing 150,000—are paired with clinical data from the NHS including mental health data, phenotype data, radiology, pathology, treatment history, and both in-patient and out-patient hospital records. The UK’s National Pathology Imaging Co-operative is currently in the process of digitizing all of the images with pathology reports for cancer patients in the 100K Genomes Project—gathering glass slides from 34 NHS sites. The process is about one-third of the way through, said Daljeet Bansal, NPIC Operations Director. Slides are being physically shipped from hospitals all over the UK to NPIC’s facility in Leeds to be scanned and submitted the GEL Trusted Research Environment, she explained.
Prabhu Arumugam, Director Of Clinical Data And Imaging for Genomics England, has been focusing on standards for this type of digital images. The images NPIC is gathering are scanned on a single scanner (a Leica GT450) at 40x coverage for consistency, then a single slide from each participant is scanned by multiple scanners to give variability to any type of machine learning algorithm. “It is logistically a challenge,” Arumugam admits, “but we are setting the standards in terms of how we create a research archive and we are really driving toward that DICOM-compatible slide image.” The team is committed to publishing the standards as they go and Arumugam referred to a Nature Medicine publication last May (DOI: 10.1038/s41591-022-01798-z)
Key to projects like this is Genomics England’s broad consent process through which patients gives GEL all of their retrospect and prospective clinical data for their lives. They are consented for re-contact to gather additional samples or phenotype data over time.
In addition to this imaging work, Genomics England has been putting its 100 bioinformaticians to work special projects for both the UK government and others, Moss said, highlighting three in particular.
GEL plans to kick off a sequencing project encompassing 200,000 newborns, of which they expect to find about 1,000 babies with monogenic disease that is addressable, identifiable, and perhaps treatable by doing whole genome sequencing up front. “It’s our first all-comer cohort,” Moss said. “The purpose of it is to really test the hypothesis that there is a clinical—but also health economic—value of, for every 200 patients you sequence, finding one baby who has a rare disease that if you intervene early enough in the pathway, there are new modalities like antisense oligonucleotides and eventually gene therapies and gene editing approaches where you can—if you catch the child early—you can reverse the progression of that disease.”
GEL is also working closely with Oxford Nanopore to introduce long read sequencing in the cancer clinic to not only get much higher resolution of structural variation, but also methylation status analysis. “That’s a really big program, where we’re testing not only scientifically, but we’re testing the implementation science of running Oxford Nanopore machines much closer to the patient, closer to the cancer clinics rather than the centralized model that we run today with Illumina,” Moss said.
Finally, GEL recently secured funding for a diversity program to ensure that the Genomics England datasets reflects the ancestral diversity of the UK population. “We’re making a lot of efforts to sample in the minority communities.”