We are working to evolve healthcare from its current focus on reactively diagnosing and treating people after they have become ill to a new focus on proactively monitoring and protecting people’s healthy state.^1,2,3,4 This involves defining people’s baseline healthy state, developing a personalized, comprehensive disease risk profile, and innovating methods to rapidly detect shifts away from people’s healthy baseline – ideally at onset of illness and presymptomatically. To obtain a clearer picture of people’s healthy state and predict their risks for disease, our laboratory uses “big data” approaches: deep omics profiling, such as genome sequencing, transcriptomics, proteomics, metabolomics, lipidomics, and microbiome measurements; synthesis of wearables data; integration of subject questionnaires; and extensive clinical tests. This deep, individualized profiling creates a much clearer picture of one’s health—much like examining most of the pieces of a jigsaw puzzle gives a clearer picture than just examining a few.

Many aspects of present-day health checkups are suboptimal. They involve traveling to a clinic – which is inconvenient and potentially exposes one to germs from folks who may be ill; physiologic measurements are performed in an unfamiliar setting; blood collection by venipuncture may provoke anxiety and pain; and relatively few (10-15) measurements are made. Starting six years ago, we set out to determine if we could reproducibly measure a much larger number of molecules from tiny droplets of blood (10 microliters) that could be obtained with the same lancets used for blood sugar tests by persons with diabetes. These small volume blood samples were mailed to a central analysis laboratory (Figure 1). Initially, this process presented many challenges in terms of reproducibility and stability of the analytes, however we developed new biochemical tricks to overcome these issues.

Figure 1. Microsampling multi-omics workflow.

We compared many different substrates for the blood collection step and settled on the Mitra^® microsampling device from Neoteryx^® which gives defined volumes (typically 10 microliters) that are fairly reproducible (within 10%), which we found to be extremely valuable for the analytical assays (we now also use Tasso devices). We spent considerable effort optimizing the extraction protocol to obtain a diverse array of molecules and recover as many metabolites, lipids, and proteins as possible. These were subjected to different assay methods: mass spectrometry for proteins, metabolites, and lipids; and targeted Luminex^® immunoassays for specific proteins and other high interest biochemical markers (e.g., cortisol). In total, from just 2 ten-microliter blood droplets, we were able to measure over 2,000 known biochemical markers associated with health and disease.

Once we settled on a collection method and protocol, we examined the stability of these analytes with regard to storage duration, temperature, and the combined effects of both (referred to as the interaction effect). Much to our delight, we found that nearly all proteins were stable, as were the vast majority of metabolites. Lipids were less stable overall, but there are many specific lipids that are stable and as such can be identified and quantified. Presumably the less stable analytes can be quantified from their by products and knowledge of their storage conditions.

As a proof of concept, we performed two studies that would be difficult to do with existing methods and nearly impossible in an individual’s normal environment. The first was to examine different people’s responses to a standard food. Previous work has suggested that people’s glucose levels respond differently to different foods⁵, but studies that examined detailed responses to complex foods were more limited. We chose the Ensure^® workout shake, which is commonly available and represents a complex mixture of lipids, carbohydrates, and proteins. By collecting small amounts of blood at different time points after people drank their Ensure shakes, we found that everyone responded differently and that we could group individuals into 5 classes based on their response patterns (Figure 2). Some people showed a strong increase in amino acids in their blood, others experienced increases in free fatty acids (which are involved in metabolism and immunity⁶) and still others increased in carbohydrates. Importantly, some individuals had a strong inflammatory response suggesting they were reacting to something in the shake, whereas others showed a decrease in relative cytokines levels, suggesting the shake was suppressing inflammation. These personalized responses could all be measured using our simple method and our approach thus paves the way to improvements in personalized diets.

Figure 2. Individual responses to a complex Ensure shake. Gray circles indicate the average value. Colored circles representing different classes of molecules show increased or decreased levels relative to the average.

The second study was to determine if repeated profiling of an individual could detect novel personalized patterns related to their health, including patterns recurring on a 24-hour cycle (circadian patterns) and other associations. During the waking hours, samples were collected from a single individual every 1-2 hrs for just over 7 days—a total of 98 microsamples—and deep metabolomics, lipidomics, proteomics, and targeted assays including cytokines and hormones were performed. We found hundreds of molecules that exhibited changes in level on a circadian pattern, including novel ones: some were due to foods, but many appeared independent of food composition, suggesting these may be intrinsically cycling. Interestingly, we showed that we could follow the kinetics of drug levels (aspirin) at an individual level. One valuable observation was that even though the participant stopped drinking coffee before noon, there was still an inverse relationship between caffeine levels and sleep. As a consequence the participant has reduced his caffeine intake and generally stops drinking coffee even earlier in the day. Another important observation was an increase in several cytokines over three days in the absence of symptoms, suggestive of a subclinical inflammatory event—perhaps a viral infection that was eliminated by the immune system before becoming symptomatic. Similar subclinical inflammatory events have been noted previously³ and have important implications given the ongoing COVID-19 pandemic, where detection of asymptomatic events can be useful for self isolation and testing.^7,8

Additional insights can be obtained by combining biochemical profiling and wearable data such as smartwatch tracking and continuous glucose monitoring. This is an area our laboratory has pioneered for novel kinds of health monitoring.^3,4 We found hundreds of correlations between metabolites, lipids and proteins with heart rate, steps, glucose levels, and other physiological measures. Importantly, we can quantify these at a personal level (Figure 3). Even known relationships such as insulin response to glucose have important implications when considering diabetes risk or treatment. A slow response might mean poor release from the pancreas, which can be mitigated with a specific treatment, e.g. prandin derivatives.⁹ Other observations were made such as the increase in glucose ~50 minutes after an increase in α-synuclein, whose levels have been linked to both stress and circadian patterns (Figure 3)¹⁰. Identifying individuals with higher-risk biochemical patterns may help in tailoring interventions to mitigate disease impact and preserve quality of life.

Figure 3. Lagged correlations between glucose levels as measured by continuous glucose monitoring (CGM) and important biomarkers measured at the individual level. α-synuclein peaks before glucose by ~55 min; C-peptide (derived from insulin) and FLT3L follow glucose by 10 min and 15 min respectively. Levels of TGF-β rise and fall at approximately the same time as glucose levels.

Overall, our studies present new methods for home and remote monitoring using biochemical assays and wearables. We believe this type of approach will become extremely valuable for health monitoring and we hope that—with further development and clinical validation—remote sampling will soon become a routine part of healthcare. Thus, like many other lifestyle activities such as shopping and entertainment which are initiated at home, home health monitoring can become more frequent and accessible to everyone, enabling individuals to track their health, improve their lifestyle, catch diseases early, and lead long healthy lives.

References

1. Chen, R. et al. Personal omics profiling reveals dynamic molecular and medical phenotypes. Cell 148, 1293–1307 (2012).
2. Schüssler-Fiorenza Rose, S. M. et al. A longitudinal big data approach for precision health. Nat. Med. 25, 792–804 (2019).
3. Li, X. et al. Digital Health: Tracking Physiomes and Activity Using Wearable Biosensors Reveals Useful Health-Related Information. PLoS Biol. 15, e2001402 (2017).
4. Hall, H. et al. Glucotypes reveal new patterns of glucose dysregulation. PLoS Biol. 16, e2005143 (2018).
5. Zeevi, D. et al. Personalized Nutrition by Prediction of Glycemic Responses. Cell vol. 163 1079–1094 Preprint at https://doi.org/10.1016/j.cell.2015.11.001 (2015).
6. Boden, G. Obesity and Free Fatty Acids. Endocrinology and Metabolism Clinics of North America vol. 37 635–646 Preprint at https://doi.org/10.1016/j.ecl.2008.06.007 (2008).
7. Mishra, T. et al. Pre-symptomatic detection of COVID-19 from smartwatch data. Nat Biomed Eng 4, 1208–1220 (2020).
8. Alavi, A. et al. Real-time alerting system for COVID-19 and other stress events using wearable data. Nat. Med. 28, 175–184 (2022).
9. Balfour, J. A. & Faulds, D. Repaglinide. Drugs Aging 13, 173–180 (1998).
10. Scudamore, O. & Ciossek, T. Increased Oxidative Stress Exacerbates α-Synuclein Aggregation In Vivo. J. Neuropathol. Exp. Neurol. 77, 443–453 (2018).