Genomics and Biodata in 2020: the future is here!

I have been working in the field of bioinformatics/computational biology for 10 years now and have always held a strong belief that the future of healthcare and life sciences as a whole would be data driven. Having attended the Festival of Genomics & Biodata 2020 conference a couple of weeks ago (self-proclaimed “largest genomics event in the UK”), I came out with a buzzing feeling that the future is here and now. Here is why…

State-level initiatives

The UK has been at the forefront of genomics research initiatives including the sequencing of the human genome, and more recently the 100,000 genomes project completed in 2018. In keeping with this tradition, multiple state-level initiatives have been implemented to ensure that the UK remains at the leading edge of the genomics health revolution. For example, the UK Biobank initiative (which involves the collection of rich data from a prospective cohort of 500,000 people recruited via the NHS across the UK, more on this in my previous post) is, as Nature puts it “pav[ing] the way to precision medicine”.

Meanwhile, the NHS Genomic Medicine Service (NHS GMS) has an ambitious plan to deliver high-quality, standardised genomic analysis across the NHS, soon. This involves the creation of 7 genomic laboratory hubs across the UK that will provide consolidated testing and interpretation across a series of tests in a test directory currently covering 180 common cancer indications and 300 rare disease indications. At present, three cancer clinical indications (sarcoma, acute myeloid leukaemia and all childhood tumours) are moving to whole genome sequencing as standard care practice. Further selected clinical indications will be labelled for whole genome sequencing in the future. In October 2019, the NHS GMS launched a programme of rapid exome sequencing for NICU/PICU (neonatal and paediatric intensive care) patients. This programme aims to quickly diagnose acutely unwell children with likely monogenic disorders, and decide whether to refer them to a specialist, schedule annual screenings or implement an early redirection to palliative care.

Another crucial player in the UK genomics ecosystem is Genomics England, originally set up to the 100k genomes project. Now that this project is complete, the mission of Genomics England is to focus on access and outcomes of genomic data research, to ensure a clinical return from the data. In practice this includes at least in part providing a platform to aggregate, de-identify and manage access to data, both genomic and clinical, generated as part of clinical care, to researchers. The platform enables feedback of the information from the research into the clinic, including at an individual level by re-engaging with a patient when new findings emerge. In other words, the platform sits at the interface of genomic healthcare and genomic research, and enable them to feed into each other.

Finally, all of these initiatives are supported by a favourable wind from Westminster. In the words of Baroness Nicola Blackwood, the UK government is committed to England remaining a global leader in genomic medicine, and wants us to have one of the most pro-innovation health care system in the world. The Life sciences business strategy (written in 2016, updated in 2020) and the yet to be published Genomics healthcare strategy outline how we will achieve this by improving collaborations between the NHS and life sciences companies, supporting and incentivising investment into the life sciences sector, increasing the UK’s investment in R&D, support clinical research via the NIHR, and improving the access to rich research-ready data from the NHS and other sources.  

Big players go big (bio)data, small players hit the big stage

One cannot talk about genomics and big data without talking about Illumina, whose technology generates the vast majority of the world’s sequencing data. In recent years, the company has made efforts towards moving their technology to clinical settings and actually improving health outcomes. This includes investigating how to remove barriers to clinical adoption, and identifying new clinical applications for sequencing technologies. A key component of the latter is the creation of two Illumina Accelerators (in the bay area and in Cambridge, UK). These host, support and coach genomics start-ups like Juno Bio using metagenomics to investigate the vaginal microbiome (which has been associated with conditions ranging from STIs to fertility problems).

Similarly, another life science giant is banking on collaborations with disruptors to make the most of their massive resources: Bayer’s lifehubs initiative (with hubs in Lyon, Berlin, Singapore, Osaka, Boston, California, and recently an AI-focussed hub in Reading UK) aims to provide spaces to connect with the community to “explore impactful outcomes” through collaborations. An example of a data-driven solution developed through these initiatives is an AI pre-screening tool that is able to predict whether a patient is likely to have a mutation of interest by analysing early biopsy slides and radiology images. These patients can then be selected for sequencing, thereby realising potential saving in the genomics screening phase supporting the implementation of a personalised therapy.

Traditional big pharma players are also moving to a drug discovery and development model that has genomics data at its core. One example of this significant investment in genomics is AstraZeneca’s 2 million genomes initiative (discussed in my previous post). The aim is for this dataset to be used by AZ’s R&D teams to identify new targets and improve patient stratification and diagnostic. In the meantime, AZ researchers have already created an internal dataset of gene-phenotype associations comprising 12,000 phenotypes, using whole exome data from UK biobank, and this resource is already in use internally by the company. 

In his talk, Josh Lauer, Global Head of Personalised Healthcare Market Development at Roche outlined the company’s strong belief in the potential of meaningful data at scale, and made a compelling point for the adoption of full genomic profiling as standard of care for cancer. As we know, every cancer (even when considered “common”) has some aspects of a rare disease, because of the heterogeneity of genomics characteristics between patients with nominally the same condition, and between cells of a single cancer. As such, it would seem that personalised medicine is likely to be the only productive way forward to treat cancer. However, there are two very practical problems with this: (i) the target population (read “market”) for each therapy-stratification criterion pair is likely to be small, and (ii) finding the patients that would benefit from each therapy would be too expensive to justify the cost when measured against the benefits of a single therapy (e.g. we may have to sequence hundreds of people to find the handful of patients that could benefit from the therapy). Problem (i) could potentially be solved by recognising that therapy-stratification criteria pairs may actually apply across multiple nominally different cancers. Problem (ii) is only likely to be solved if full genomic profiling is part of the standard of care for each cancer patient, such that the cost of profiling can be amortised against all of the possible therapy-stratification criteria that will eventually be available. In the short term, this means that the cost effectiveness of therapies has to be evaluated independently of the profiling cost¹, to make it viable for personalised therapies to come to market. This ties in nicely with the NHS GMS initiative above, which together with an internal push at some of the pharma giants, could (finally) make the personalised medicine dream a reality for cancer patients.

Other big data genomics products becoming a commercial reality include Google’s Cloud Life Science (adopted by the Broad Institute to analyse all of their data - which according to Google brought down their analysis costs from $45/genome to $5/genome - with some optimisation), SevenBridges’ Graph Genome Suite (which is able to improve the accuracy of variant calling by moving away from the single linear reference genome, taking into account variability in a general or personalised way), SOPHiA’s advanced genomics analysis platform (which is now used by >1000 institutions in the world in 82 countries, including public and private hospitals, diagnostic labs and universities, and offers a common, easy to use interface for genomics data analytics), and of course consumer genetics tests ranging from ancestry to disease risk tests (by the likes of 23andMe, Invitae, Myriad, etc.) By all accords, the market for these tests is poised to grow, with companies like ixlayer even providing a platform to enable anyone to launch a precision health test.

Genomics is starting to pay off

All of these efforts are starting to pay off: according to Google’s Cloud Life Science marketing team, 19 of 59 drugs approved by the FDA in 2018 were genome-informed targeted oncology drugs. Further, a study published in Nature last year demonstrated that drug side effects are more likely to occur in organ systems where there is genetic evidence of a link between the drug target and a phenotype involving that organ system, compared to when there is no such genetic evidence – suggesting that genetic data should be used to predict safety issues associated with drug targets. The Open Targets initiative (more on that in my previous post), has identified multiple targets through their CRISPR-Cas9 synthetic lethal screen in cancer cell lines, some of which are already being pursued according to Philippe Sanseau, head of the Computational Biology and Stats Department in Target Sciences at GSK and the GSK lead for Open Targets.

Another major success story of the last few years lies in the application of liquid biopsies for pre-natal testing. Tests like Harmony (by Ariosa Diagnostics, now part of Roche), which analyse cell-free DNA in expectant mother’s blood to identify chromosomal and genetic abnormalities, are now offered to patients over the world, including in the UK. Liquid biopsy applications are also showing great promise in the field of oncology, with companies such as Inivata already providing commercially available tests for genes relevant to the care of patients with lung cancer (a type of cancer that has been a particularly successful example of personalised therapy).  Promising research (see e.g. this research at CRUK) is ongoing to increase the sensitivity of detection of circulating tumour DNA (ctDNA) from liquid biopsies, proving that cell free DNA is likely to be a crucial part of future cancer care. 

It is clear that challenges still remain - we are only at the start of the journey towards a fully genomics-enabled healthcare system. However, with the hard evidence of products hitting the market, funds being committed to R&D to support the next steps, a vibrant ecosystem including collaborations between players of all sizes, and a clear willingness from the UK government to embrace the technology, it is hard not to believe in the potential of genomics to improve healthcare. I have always been a believer in the promises of the science, and it’s thrilling to see it become reality.

¹Despite the promise of the “1,000$ genome”, profiling costs are still a significant concern for clinical adoption, as outlined by a recent study on the true cost of whole genome sequencing in cancer and rare disease in an NHS lab. The study pointed out the need to keep in mind that we often need to sequence more than one sample per patient (such as e.g. matched tumour and healthy samples for cancer, or triads for rare diseases), as well as include the costs of analysis, storage and reporting. The study concluded that, in the setup investigated, the cost of genome sequencing for a cancer case was almost £7k, and just over £7k for a rare disease trio. The highest cost related to sequencing (acquisition and maintenance of the sequencing machine, and most importantly, consumables). There is of course a potential economy of scale there – and much discussion was also had during the conference as to whether labs (and even research institutions) actually need their own sequencers. Why not let a “sequencing factory” do the sequencing? Considering that the consumables represent a very large part of the cost of sequencing, this is unlikely to completely solve the problem in my opinion, but it seems likely to be part of the solution.  James Hadfield, Director at AstraZeneca’s Precision Medicine Laboratories (though seemingly speaking from experience as a former head of CRUK-CI’s Genomics Core): “buy sequences, not sequencers!”…A sentiment echoed by an audience member (clearly not an audience member of the artisan sourdough generation): “I can buy sliced bread from Tesco’s, why would I make it?”