Dense, Longitudinal Phenotyping of Individuals Generates the Data to Transform Healthcare
TRISH Workshop 2018
Lee Hood
Senior VP and Chief Strategy OfficerInstitute for Systems Biology, Seattle
Senior Vice President and Chief Science OfficerProvidence St. Joseph Health, Seattle
July 31, 2018
The grand challenge for biology and medicine: Deciphering biological complexity—Caltech 1970
6
Scientific (quantitative)
Wellness: dense, longitudinal, personal phenotyping—2014Arivale 2015-today
7
Bringing P4 healthcare to
Providence St.Joseph Health
(and the US and other
Healthcare Systems)—2016-today
2016
I Participated in Seven Paradigm Changes in Biology Dealing with Complexity and Led to
Precision Medicine1
Brought engineering to biology—19701970
2The Human Genome Project—1990
3Cross-disciplinary biology—1992
4Systems biology—2000
Conceptualization of systems medicine and P4 healthcare 2000-2004
5
2014--The 108 Person Scientific (Quantitative) Wellness Project (Pioneers)Principal Investigators: Lee Hood and Nathan PriceUsing personal, dense, dynamic data clouds or dense phenotyping for wellness IRB approved study
Determinants of Health in U.S.
60%
30%
10%
Genetics
Behavior & environment
Health Care
Dense, dynamic, personal data clouds assess the integration of individual genetics and environment
Assays / Measurements for 108 Pioneers
Database of actionable
possibilities that will grow over
time
GENOME
Whole Genome Sequencing SNPs Millions
LABS
Detailed lab tests 3x(blood, urine, saliva)Clinical chem. 150
Metabolites 700Proteins 400Almost 1200
analytes
Continual self-tracking& lifestyle monitoring
MICROBIOME
Gut MicrobiomeRation of
microbial species 3x
Creating dense, dynamic, personal data clouds—dense phenotyping
SELF-TRACKING
Wellness Coaching for 108 ParticipantsA Critical Component of Scientific Wellness
Sandi Kaplan, MS, RD Craig Keebler, MD
Wellness Coach Study Physician
Education of pioneers about wellnessCompliance—70%Long term longitudinal recruitment of participants
Clinical Labs Discovery: Improvements in blood health with behavioral coaching
50%
55%
60%
65%
70%
75%
80%
85%
90%
95%
100%
Cardiovascular Diabetes Inflammation Nutrition
% ch
ange
in o
ut-o
f-ran
ge m
easu
rem
ents
Baseline 3 months 6 months
Improved by 33%Improved by 6% Improved by 12% Improved by 21%
Examples of Improved Wellness
• Arthritis and hemachromatosis• High cortisol and colon cancer• B12 deficiency and Bell’s palsy• Vitamin D and blocking genetic variants for
uptake• Increased WBC count and leukemia• LDL particle size abnormal with normal LDL
cholesterol—significant carotid arterial plaque• Ferritin trending down—anemia• Liver markers trending up—fatty acid liver
Creation of a Consumer-Based Scientific Wellness Company
2015 LAUNCH
Arivale Clients
• Currently about 4000 pioneers
• In all US states except New York. nationally distributed clinical trials
• To date, more than 60
• Wellness to disease transitions—for most major chronic diseases
• ISB agreement to analyze all data in collaboration with Arivale
The Hubble Telescope allows us to probe the dark matter of the universe just as dense and dynamic personal data clouds allow us to probe the dark matter of human biology and disease.
Personal Dense, Dynamic Data Clouds: Probing the DarkMatter of WellnessAnd Disease
Arivale/ISB Computation Platform to deal with personal data clouds
Statistical Correlations
Deriving Insights from Data: New Frontiers—3500 Statistical Correlations
A
B C
Price, Magis, Earls , Hood, et al, Nature Biotechnology (2017)
Identification of 70 multi-omic functional communities (modules) in the correlation
network.
Serotonin
Essential Fatty Acids
Bladder Cancer Risks
Microbiome diversity
Nutritional status
CardioMetabolic
• Cholesterol is positively associated with alpha-tocopherol (Vitamin E)
• Cholesterol is negatively associated with endogenous thyroxine
• A beneficial side effect of the drug thryroxine(Synthroid) is lowering LDL cholesterol
Total cholesterol community as one of70 communities
Price, Magis, Earls, Hood et al, Nature Biotechnology, 2017
Price, Magis, Earls…Hood, Nature Biotechnology, 2017
Dynamical Statistical Correlations between State 1 and State 2 Are Informative
Systems Medicine and Big Data
We can determine your polygenic risk for more than 100 diseases.
GWAS Variants Have Been Identified for Over 127 Diseases and Disease Traits
ADHD COPD MyopiaAlzheimer's disease Crohn's disease Obesity
Anorexia Esophageal cancer OsteoarthritisAsthma Gout Osteoporosis
Atrial fibrillation Grave's disease Ovarian cancerBreast cancer Hematocrit Pancreatic cancer
Bipolar disorder Hypertension Parkinson's diseaseBlood pressure Hypothyroidism Primary biliary cirrhosis
Bone mineral density Inflammatory bowel disease Prostate cancerInflammation Iron levels Psoriasis
Calcium Lung Cancer Rheumatoid arthritisCardiovascular disease Lupus Schizophrenia
Celiac disease Macular degeneration StrokeCholesterol levels Magnesium levels Type 1 Diabetes
Chronic kidney disease Metabolic syndrome Type 2 DiabetesColorectal cancer Migraine Ulcerative colitis
Coronary heart disease Multiple sclerosis Urate levels
LDL cholesterol in Participants Shows Monotonic Relationship with ‘Genetic Risk’
75
85
95
105
115
125
135
0
5
10
15
20
25
30
35
40
Very Low Low Medium High Very High
LDL C
hole
stre
ol (m
g/dL
)
Num
ber o
f par
ticip
ants
in th
is ris
k ra
nge
Genetic risk vs. Disease state
LDL Cholesterol Levels vs Genetic Risk (59 variants)Genetic Risk Baseline (No Meds)
Example: Cystine Is Negatively Correlated with Increasing Genetic Risk for Inflammatory Bowel Disease
cystine (plasma) glutathione (intracellular)cysteine (intracellular)
Cystine and cysteine are limiting precursors of glutathione synthesis.
reactive oxygen species (ROS)glutathione (intracellular)
O2-
H2O2
O2-OH
OHOxidative stress
Glutathione consumes ROS associated with chronic inflammation and colonic damage.
Only one individual in our study had been diagnosed with IBD.Independent case-control study identified abnormally low cystine in IBD patients1.
1. Sido, B. et al. (1998). Impairment of intestinal glutathione synthesis in patients with inflammatory bowel disease. Gut 42, 485–492.
Increasing genetic risk
State Transition: Metabolites, Clinical Chemistries, and Protein Expression Levels Diverge with Age and Provide Metrics for Wellness
Incr
easin
g Va
rianc
e of
Exp
ress
ion
Observe Increasing Outliers with Age:Metabolomics
Increasing decades of life
IS BIOLOGICAL AGE RELATED TO HEALTH?
John Earls
Type 2 diabetes INCREASED biological age across all analyte classes tested
Coronary Artery Disease INCREASED biological age across (almost) all analyte classes tested
Activity DECREASED biological age across all analyte classes tested
Health Aging
• Assess how Arivale program decreases ones biological age
• Do N=1 experiments to decrease one’s biological age
Assess Resilience
• Determine how resistant biological age changes are to relevant stressors—either short term or long term.
State Transitions: Wellness to Disease
Aug Dec Apr SepBlood &
MicrobiomeBlood Blood &
MicrobiomeBlood
Example – identifying outlier forpioneer with pancreatic cancer
Levels of protein of interest
N=1990All draws/All pioneers
Dec2015
Apr2016
Sep2016
Aug2015
Over the course of the four blood draws, this pioneer has consistently been the largest outlier for the protein of interest.
DiagnosedJanuary 2017
2015 2016 2017
1. Yachida, S et al. (2010). Distant metastasis occurs late during the genetic evolution of pancreatic cancer. Nature, 467(7319), 1114–1117.
A 2010 study by Bert Vogelstein estimates at least 5 years between parental, non-metastatic founder cell and metastatic ability in pancreatic cancer1.
Identifying disease-perturbed networks pre-diagnosis: Early example for pancreatic cancer
0 1 2 3 4 5 6 7 8
VISIT 1
VISIT 2VISIT 3
VISIT 4
Distribution of # of significantly low p-values
Eve
948 clients/how many of 9 potentialcancer-related networks perturbed
Number of disease-perturbednetworks with >3 outlier proteins
Building computational systems for truly personalized disease diagnostics
• Define wellness state(s) from cohort
• Identify divergent values for each individual—analytes, networks and correlations—at each blood draw– Informed by dynamics– How unlikely?
• Functional analysis of individual divergences at each blood draw (signal to noise issue)—many divergences prior to diagnosis– Deep context learning to interpret
meaning– AI/Machine Learning/Systems
Biology• Every individual will know all
their disease-related outliers which eventually will trigger early reversal of most chronic diseases
• Scientific wellness will be the front-end of healthcare system
Early Reversal of Chronic Diseases: Preventive Medicine of the 21st Century
• In following about 4,000 or more patients over an extended time period, we have started to see wellness to 60 earliest disease transitions for all common diseases (as measured by blood analytes).
• Use data clouds to develop blood biomarkers for the earliest transitions for each disease and disease-perturbed network biology analyses to identify drug candidates/life style changes for therapies to reverse each disease at its earliest transition.
• Thus individuals will have diseases reversed before the diseases manifest themselves as a chronic disease phenotype—an approach to eventually eliminating chronic diseases--preventive medicine of the 21st century
Cognitive Learning –clinically proven—cell
phone deliveryMike Merzenich UCSFSynaptic Health with
Multimodal, IndividuallyAdaptive therapy (36
elements)
Dense Phenotyping Hood and Price ISB
PET Metabolic Scanning Mike Phelps UCLA
PSJH Strategic Partnership for Alzheimer’sHoag, Swedish, St. John’s, ISB and others
Multimodal Therapies
DiagnosticsBlood biomarkers for
1) earliest wellnessto disease
transition and 2) distinguish 3 AD subtypes
Rod Shankle HoagCocoa AD Trial
4-10 years before clinical diagnosis
Alzheimer’s ClinicsN=1 trials with ISB
Swedish, and possibly St. John’s
N=1 Experiments on Cancer
M.M., 69 yo female - Diagnosis: Invasive Bladder Carcinoma (pT3a; High grade)
Transurethral Bladder Biopsy
(TURBT)
PIONEER 100 PROGRAMPersonal GenomeRegular Blood Draws
Blood works MetabolomicsProteomicsMicrobiome
ACTION: “Anti-inflammatory Diet”à Very low w6/w3
Fall 2017~ 2015
Blood in urine
à Her2Copy number gainà CDKN2A p16 focal deletionà Cyclin D copy number gain
TUMOR DNA 262-sitesTargeted Exome sequencing
BLUE: MEASUREMENTGREEN = INTERVENTIONRED = specific patient findings
TRADITIONAL “PRECISION MEDICINE”
SURGERY: CYSTECTOMY
CHEMO-THERAPY
NEW: N=1 in P4 MEDICINEPERSONAL DENSE DYNAMIC DATA CLOUDS
TARGETED THERAPYü Trastuzumabü palbociclib
TRANSCRIPTOMEà Her2 increasedà Cyclin D higher (?)
PROTEOME (Somalogic 5000 proteins)
3D Tumor Organoidà drug susceptibiliy
Multi-plex IHC
Spring 2018
TUMOR WGSTumor MutationalBurden (TBD)
• High Tumor Lymphocyte Score• High IFNg signature score• Low TGF• Low Inflammation scoreconf
irmsconfirm
s
confirms
Indication for futureCHECKPOINT INHIBITORAs next line therapy
confirms//
Applications of Dense Phenotyping
Personal, Dense, Longitudinal Data Clouds on Human Individuals Will Allow Us To:
• Optimize human wellness (with coaching)• Follow disease progression, response to therapy
and return to health—understand disease mechanisms
• Identify earliest wellness to disease transitions and reverse early—preventive medicine of the 21st
century (follow those at high risk for a disease to find earliest transition)
• N=1 experiments, then stratification, are the key to deconvoluting some biological complexities, e.g., nutrition, resilience and aging
• 2-step clinical trails—First, assess with 50 patients. Second, then 50 patients all responders.
• Genentech took Herceptin to FDA for successful approval with only 46 patients.
Alaska, California, Oregon, Montana, Washington, New Mexico, Texas
ISB & Providence St. Joseph HealthAffiliation
States served 7
Hospitals 50
Physicians 7500
RNs 36,000
Unique patients served each year
5 million
Total Assets $23 billion
Third largest not-for-profit healthcare system in the USIntegrated Medical Electronic Health Records for 30 million patients
TRANSLATIONAL PILLAR:
ScientificWellness
Scientific Wellness: Opportunities for Astronauts• Optimize wellness: before, during and after space• Assess genetic risks for relevant diseases• Follow high risk individuals for earliest relevant
disease transitions. Look for analytes to reverse.• N=1 experiments to determine what the stress of
space does to 10s of individuals• Determine dynamical biological ages to assess
wellness and follow it as an astronaut goes though 3 state changes: before, during and after space.
• Employ more accurate assessments of bone density, immune competence, and other relevant physiological features
• Digital devices could be powerful—e.g. heart rate variability
• Assess resilience during training• Statistical correlations between earth, space, and
earth data will be informative.
Reduce Cost of Scientific Wellness
• Reduce the costs of the assays more than 10-fold in 5-10 years. Driven by scale of wellness studies. Genome. Metabolites. Immunity.
• Reduce dimensionality and optimize analyteanalyses. Tricorder with 5000 measurements in 10 years at home
• Use avatars to amplify enormous what coaches can do and so scale coaching.
• Insurance will pay for scientific wellness in 5 years. Wellness clinical trial
• Improve wellness for the individual• Reverse diseases at their earliest
transitions--eliminate many chronic diseases
• Reverse ever escalating healthcare costs
• Scientific wellness will be the front-end of healthcare systems—scientific wellness will be the key to dealing with individual health and chronic diseases
What Will 21st
Scientific Wellness Achieve?