Genomics vs. Metabolomics: who will win?
Big data that includes metabolic data may provide us with new intermediate phenotypes for testing preventive interventions early in the causal pathway of disease and allow us to prevent future disease
In general, I am rather a big data sceptic because the results can rarely be applied to individuals. However, this Israeli study looking at laboratory results of 2.8 million adults is telling us that what we consider as a normal laboratory result may not necessarily be normal and when integrated with other data may actually be quite a good predictor of future disease.
Am I surprised by these results? No, I shouldn’t be. Although reductionists like to define life as being self-replicating DNA or RNA code, i.e informatics, it can only self-replicate via metabolism. Although metabolism is essentially the execution of DNA-encoded algorithms these have to integrate and respond to the micro-environment and the lived-in macro-environment of the organism. This means life is metabolism; i.e. the interactions of the environment with informatics. So a diseased life or a life that will be diseased in the future will have to have a signature of this disease in its metabolic profile. This explains why there have to be signatures present in laboratory results, which are snapshots of the body’s metabolism at a point in time.
This study clearly challenges our rigid and clearly flawed methodology of defining what is a normal laboratory result. In general, we go to a so-called healthy population and measure metabolite X. We then generate a distribution of results and set the normal value at being within two standard deviations of the mean. So anyone with a result outside the normal range is considered abnormal. This study is telling us that even if you have a normal result within the normal range this may indicate disease or future disease.
Clearly, laboratory results should be treated as continuous interacting variables and incorporated into future prediction models rather than being used and independent categorical or binary variables. For example, mild red blood cell macrocytosis within the normal range and in association with normal haemoglobin levels, normal AST and ALT levels, but an AST:ALT ratio close to one may be telling us something about that patient’s environment. And when this is linked to genomic data it could be telling us that this patient is at very high risk of future liver disease. When this is combined with genomic, other metabolic data and other data it could also indicate this patient is at very high risk of developing dementia in the future.
It is clear that big data sets that include metabolic data that can be integrated with genomic, demographic and other phenotypic data will lead to improved disease risk prediction. This type of data analysis and modelling may provide us with new intermediate phenotypes for testing preventive interventions early in the causal pathway of disease and allow us to prevent future disease. What we now need to do is to do the work.
There is little doubt that we are living in exciting times.
Cohen et al. Personalized lab test models to quantify disease potentials in healthy individuals
Netta Mendelson. Nat Med. 2021 Sep;27(9):1582-1591.
Standardized lab tests are central for patient evaluation, differential diagnosis and treatment. Interpretation of these data is nevertheless lacking quantitative and personalized metrics. Here we report on the modeling of 2.1 billion lab measurements of 92 different lab tests from 2.8 million adults over a span of 18 years. Following unsupervised filtering of 131 chronic conditions and 5,223 drug-test pairs we performed a virtual survey of lab tests distributions in healthy individuals. Age and sex alone explain less than 10% of the within-normal test variance in 89 out of 92 tests. Personalized models based on patients' history explain 60% of the variance for 17 tests and over 36% for half of the tests. This allows for systematic stratification of the risk for future abnormal test levels and subsequent emerging disease. Multivariate modeling of within-normal lab tests can be readily implemented as a basis for quantitative patient evaluation.
MS Research MS-Selfie Newsletter
General Disclaimer: Please note that the opinions expressed here are those of Professor Giovannoni and do not necessarily reflect the positions of the Barts and The London School of Medicine and Dentistry nor Barts Health NHS Trust and are not meant to be interpreted as personal clinical advice.