How “-omics” will impact healthcare
Emerging laboratory technology and the forefront of biological study have begun to amass large reservoirs of data, newly referred to as “-omics” – genomics, transcriptomics, proteomics, metabolomics. The list is increasing. So, just what are these things and how will they impact the delivery of healthcare in the near and distant future?
Genomics, and in particular, personal genomics, refers to the sequencing, analysis and interpretation of an individual’s entire DNA content – their genome. Automated high-capacity sequencing has made this a commoditized endeavor. It has become a commercial activity, not just in the research labs at universities. By determining an individual’s genome, and comparing it to known, published sets of data, the expression of various traits and risk of disease can emerge.
We are also at a time when clinical data from Electronic Health Records (EHRs) is becoming increasingly interconnected. We are seeing health data move from institution-centered silos (albeit connected in more significant ways than ever before) to fully unified aggregated data stores. These data stores are starting to become organized into a Medical Knowledge Graph. This Medical Knowledge Graph contains the longitudinal clinical history of large numbers of individuals – the phenotypic data, the observed phenomena of large populations. Once might consider this, if we are using “-omics” terminology, as the “phenomenome” of an individual’s story.
Combining this clinical history of an individual, and of large populations of individuals, with genomic data is staggeringly powerful. There is the research application of this, where diseases and genetic characteristics can be understood in more depth than ever possible before. There is also the individual application of this, where for a given individual with a given set of conditions, lab tests, medications, demographics, and genetic information, personalized recommendations can be made for health that will work right the first time. This is the goal of Personalized Medicine.
Additional “whole body” assessments of individuals and populations are also emerging. Metabolomics is the systematic study of metabolic processes, and the unique chemical fingerprints that specific cellular processes leave behind. Proteomics is the large-scale study of proteins, particularly in their structures and functions. These kinds of studies, especially when combined with genomic data as well as “phenomenomic” data from clinical records, are poised to unlock our understanding of vexing medical issues, such as autoimmune diseases, cancers, allergies, obesity, addictive disorders, and many others. Why do some individuals, when exposed to the same external circumstances, develop disease, while others do not? In cancer treatment, which individuals will respond to a given treatment regimen, while others will respond better to some other one? The answers to these kinds of questions will emerge from combining these massive data sets and learning from the patterns seen.
The role of AI in putting this all together
Artificial Intelligence (AI) is a set of algorithms that can sift through large amounts of data and identify patterns. The biggest limitation in using AI in healthcare isn’t so much the sophistication of the algorithms – it’s more the limitations of data available from which to learn. Health data has been historically fragmented into proprietary silos, generally centered around the institutions (hospitals, doctor’s offices) that created them. But with the emergence of cross-institution aggregated data stores, and its organization into a Medical Knowledge Graph, AI has the capacity to search through that data and identify patterns that may not have been previously apparent.
When this pattern-recognition ability is applied beyond just the clinical data (the “phenomenome”) and also includes analyzing genome data, proteome data, metabolome data, and others as these technologies emerge, then the insights become very deep indeed.
Clinical research can be guided by the patterns that emerge from AI’s application to these large cross-domain data sets. Targeted medical treatment, improved drug development, better treatment pathways – all these advances will be the result of such understanding.
Further, for the individual, a personalized set of recommendations for improved health can be offered. The right medications to use, the right diagnostic tests that should be done to better clarify the best way forward, the impact of life choices and nutrition – all these things result from such an understanding as well. With health plans, authorization pathways can be streamlined and unblocked, so that the best (and least duplicative, least expensive) path to a desired outcome for a specific individual can be what is automatically authorized.
The results from such activity can change the delivery of healthcare in a dramatic way. In order to build that future, we need to aggregate data and let AI algorithms help us find the patterns generally and specifically for an individual. Clinical data (the “phenomenome”) needs to be aggregated into a universal Medical Knowledge Graph. Genomic data, metabolomics data, proteomic data, and all the emerging “-omic” data need to be correlated with each other and with the Knowledge Graph, so that AI will learn from much broader data sets than what is available now. This is the future towards which we are building.