Designing the Future of Multiomics Research

The omics revolution has transformed how we use biological information to understand human bodies and disease. Methods that use single-omics data, like genomics, proteomics and metabolomics, have been essential to these leaps forward. Now, researchers are realizing the value in combining these data through multiomics. These approaches better reflect how genes, proteins and metabolites interact and feedback on each other. To expand multiomics we will need bold new study designs, data analysis and collection methods.

The big data genomics revolution

Just over 20 years ago, geneticists made a breakthrough in our understanding of depression.¹ Using a technique called candidate gene analysis, researchers showed that variations in the serotonin transporter gene explained why some people developed depression in response to stressful life events and others didn’t. This seismic discovery was cited thousands of times. Unfortunately, it was also completely incorrect.

Huge replication studies involving tens of thousands of participants could not replicate the findings.² The debunking of this theory contributed to the wider abandonment of candidate gene studies altogether.³

Simply put, these studies had tried to stay small when they needed to go big. The authors of the debunked depression paper used a 1000-participant sample. Modern genome-wide association studies (GWAS), which ultimately replaced candidate gene studies, recognize the need to recruit more than one million participants to detect small effect sizes. This embrace of big omics research has been transformative, said Urko Marigorta, a genomics researcher at CIC bioGUNE, a molecular biology institute in Bilbao, Basque Country. “Nowadays, the replicability rate of genetic information must be 98 or 99%,” said Marigorta. Now, genomics research can build reliably on previous information, making each new GWAS study an extension of a strong scientific foundation rather than a standalone venture on shaky statistical ground.

GWAS studies clarified that the information encoded in our DNA only tells us so much. To better understand the human body in health and disease, multiomics approaches incorporating other biological information sources are needed.

Geneticists have reaped the rewards of large-scale omics because of genetic databases like the UK Biobank, which stores biological and medical data from 500,000 participants, including the world’s largest store of genomes. It’s easy to see a future in which proteomics and transcriptomics studies are all conducted on a similarly massive scale.

Multiomics analysis: Snapshots or videos?

Marigorta said that size might not matter for omics studies of our metabolome and proteome. Many cross-sectional analyses show differences in gene expression, proteins or metabolites between healthy people and people living with diseases, but it’s unclear how relevant these changes are to disease onset. “Most of these are not causal effects but are byproducts of disease,” Marigorta explained.

These differences simply represent snapshots of individuals’ constantly changing proteomes or metabolomes, said Dr. Mike Snyder, a professor of genetics at Stanford University.

Achieve Exceptional Longitudinal Performance in Clinical Plasma Proteomics

Learn how to achieve consistent protein identification rates across longitudinal measurements.
View Whitepaper

Advertisement

Snyder and Marigorta advocate for longitudinal studies, which take repeated measures from the same individual. The difference between cross-sectional and longitudinal multiomics approaches is like “the difference between a photograph and a video,” said Snyder. If studies resample from a single individual, their baseline measurements act as a control, he adds. While cross-sectional studies need huge samples to dampen statistical noise, he explained that the differences between one individual’s healthy and diseased states are lower than those between two individuals’ diseased states.

How longitudinal multiomics analysis works

A recent paper from Marigorta’s lab is an example of this approach to omics.⁴ This unusual study didn’t target a particular disease but instead recruited 162 healthy individuals. The team sampled their cohort’s whole-exome genomics, urine and serum metabolomics and lipoproteomics data at baseline. Importantly, they re-recorded the same information after 24 and 36 months. The study characterized this cohort into groups based on disease risk predictors. Individually, the omics analyses could not consistently separate the volunteers’ data. However, when the information was combined, the patients could be divided into clusters. One group had a blood profile that indicated higher risk factors for dyslipoproteinemias, disorders of blood lipoprotein levels. Those clusters persisted when the same metabolic information was plugged in two years later. Such findings could act as guides to preventive strategies for disease.

Snyder said these types of studies are being boosted by the advent of “wearable” technologies that provide real-time data on important biological metrics. Wearables – miniaturized electronics that record biomedical information through built-in sensors – have become more complex and capable over the last decade. Wearables’ continuous tracking capabilities are well-placed to record predictive biomarkers of disease, and have shown success in predicting cardiometabolic performance and anticipating diseases like atrial fibrillation.⁵

Importantly, many of these tracking technologies have been integrated into consumer devices, expanding their utility for the wider population. Snyder’s own research suggests the future of wearables might see them working in combination with microsampling approaches. A recent study combined wearable data from a smartwatch and continuous glucose monitor with a blood microsampling regimen that revealed how individuals’ metabolomes and inflammasomes responded to dietary interventions.⁶

Snyder is also conducting longitudinal research, and points to an example of a study participant who noticed an important biomarker double in value to a level that still registered as normal. The participant brought this increase to Snyder, who sent him for further tests that only then detected an above-range reading. “If you only have one measurement, in today's world, most physicians wouldn’t catch that,” explained Snyder.

Building a big multiomics data resource

Long-term approaches have one big barrier: cost. Repeated measurements and participant-tracking methods are involved and time-consuming approaches. But some ambitious omics initiatives still want to combine big science and longitudinal analysis.

Professor Tuuli Lappaleinen, genomics platform director at the KTH Royal Institute of Technology, spearheads one such project. Precision Omics Initiative Sweden (PROMISE) is a planned national multiomics data resource that aims to enhance research and healthcare across Sweden.⁷ PROMISE will amass data from between 100,000 and 500,000 volunteers – representing 1–5% of the Swedish population. This will include healthcare records and multiomic data such as volunteers’ genomes, proteomes and metabolomes. Lappaleinen said longitudinal data will be an important feature of PROMISE, as it aims to build a lifelong picture of health and disease. PROMISE will also use existing healthcare data, which Sweden records in registries that often go back several decades. “Everyone has an identity number that allows you to link the registries together,” Lappaleinen adds. 

Is precision health compatible with privacy?

Building a data resource is only half the battle. That data has to be shared with those who can use it effectively. PROMISE will standardize how data is transferred between healthcare and research settings in Sweden, paired with legislative changes that will make it easier to integrate data sets. PROMISE will need buy-in from participants, whose biological data could be widely shared. Lappeleinen said that Sweden benefits from high trust in public institutions and widespread technology adoption among the population. “Those kinds of foundations are very, very important when you ask people to participate,” she added. Additional practical challenges come from variation in how data is collected at its source – an example is the challenge in integrating written electronic health records. Lappaleinen said that initiatives like the Global Alliance for Genomics and Health, which aims to standardize data structure to maximize sharing between countries, have eased this process but acknowledged that complete data harmony between countries with different approaches to privacy and different health system architectures is unlikely.

As large-scale initiatives like PROMISE and small-scale interventions using precision wearables become more common, it is likely that the public will need to share their health data more widely, said Snyder. He acknowledges that this may risk some data being hacked: “I think the benefits outweigh the cost,” Snyder adds. He compares the use of wearables to credit cards – where the risk of being hacked or stolen is dwarfed by the convenience of not carrying bags of cash. Snyder said that multiomics projects should do all they can to protect data but that, ultimately, we may have to accept that precision health initiatives require us to cede some control over our proteomic and metabolic information. “Nothing's really private anymore,” Snyder concludes.