Personalized
precision health

Your body is a biochemical machine. If genes are the blueprints of the machine, proteins (functional products of genes) are the machine itself. Dynamic changes in biomolecules such as proteins and metabolites directly reflect health and disease, with or without symptoms. Many such molecules are widely used in medical labs, but often on a reactive basis. Many more molecules have been repeatedly shown to be effective indicators, but broad adoption could take years or decades. We are developing technology to enable the profiling of hundreds of such molecules at once at a reasonable cost (e.g., using LC-MS/MS), and building an AI platform to translate the molecular profile into accurate and actionable insights, so that you can proactively measure and monitor your health and detect disease early.

Example Molecules

Physicians are already using many proteins (often in the form of a blood test) to assess general and specific health conditions. Common examples include hemoglobin A1c (HbA1c) to diagnose and monitor diabetes; liver panels testing ALT, AST, and ALP enzymes; and C-reactive protein (CRP) as a non-specific indicator of inflammation. The diagram below shows more such examples. Proteins with an underscore are not yet broadly available as lab tests, and may cost over $100 per test.

(Hover on each protein for details)

Liver Disease
Heart Disease
Muscle Damage
Diabetes
Inflammation
Cardiovascular Disease
Kidney Disease
Pancreatic Disease
GGT
ALT
AST
ALP
Aldolase
Troponin
BNP/NT-proBNP
CK-MB
Myoglobin
CK-MM
HbA1c
AHSG
Adiponectin
C reactive protein
Serum amyloid A
ApoC3
ApoA1
ApoB
Lp(a)
Cystatin C
B2M
PEDF
Alpha-1-microglobulin
Alpha-2-macroglobulin
Trypsinogen
Lipase
Amylase

As complicated as those examples may seem, these are just the tip of the iceberg: the great majority of such associations unearthed by scientists remain in paper, with some implemented only in a handful of labs across the country.


Science of Omics

In the meantime, life scientists continue to unravel the complicated relationships between biological measures (called biomarkers) and biological states. Genomics studies genes and their regulation and expression (i.e., making proteins). Proteomics studies how protein quantities and variations change in relation to biological functions and diseases. Metabolomics studies metabolites, namely small-molecule (intermediate and end) products of metabolism. Microbiome is the collection of microorganisms (e.g., bacteria and fungi) of a given community (e.g., colonies in the gastrointestinal tract).



Research in those omics (especially proteomics) has seen an explosion in the discovery of potential biomarkers for many complex diseases such as cancer, diabetes, and cardiovascular diseases. Such studies are often accompanied by massive amounts of data (which are often more or less noisy), and a growing trend is the use of AI techniques to deconvolute raw data, identify patterns, and develop multi-analyte panels.


What We Do

As illustrated below, today's healthcare landscape consists of largely disjoint efforts by care providers, the patient themselves, and scientific researchers. We would like to build tools and services to afford the patient a more holistic and quantitative approach to monitoring and managing their health. We subscribe to the P4 medicine vision where healthcare should be personalized, participatory, preventive, and predictive.

Disparate and disjoint healthcare landscape

Our mission is to advance personalized precision health. We have two opinions: First, having more data is better than having less data. The term "overdiagnosis" is an oxymoron (think COVID-19 testing); the real issue is a need for better diagnosis methods and better follow-up action plans. Second, lab results are currently interpreted based on decades-old reference ranges and physicians' experience in combination with other observations; we believe there are better ways. For example, a growing trend in omics research is to use machine learning to train high-accuracy classifiers on top of multi-analyte panels (that outperform any individual analyte). We adhere to a data-driven approach where (multi-omic) molecular data and (multi-modal) clinical data are integrated using AI to eventually develop high-sensitivity high-specificity models for the screening, diagnosis, stratification, prognosis, and therapeutics of various complex diseases.