Metabolic Phenotypes Of Diabetic Kidney Disease - Ville-Petteri Mäkinen
-
Upload
australian-bioinformatics-network -
Category
Health & Medicine
-
view
304 -
download
3
description
Transcript of Metabolic Phenotypes Of Diabetic Kidney Disease - Ville-Petteri Mäkinen
METABOLIC PHENOTYPES OF DIABETIC KIDNEY DISEASE
Ville-Petteri Mäkinen University of California, Los Angeles, USA
Imperial College, London, UK South Australian Health and Medical Research Institute, Adelaide, AU
Figuring out biology from complex data:
NATIONAL HEALTH AND NUTRITION EXAMINATION SURVEY 2005-2010, AGE ≥ 20
United States Renal Data System 2012 Annual Data Report
MORTALITY, AGING AND KIDNEY DISEASE
Age
Moody WE, Edwards NC, Chue CD et al. Heart 2013:365–372Foley RN et al. Am J Kidney Dis 1998:S112–19
TYPE 1 DIABETESAutoimmune disease
No insulin secretion
High blood glucose
Complications
BLOOD GLUCOSE IN T1D
Meals, physical activity, sleep patterns etc. make it difficult to inject optimal amounts of insulin during the day.
• Insulin resistance (life style) ⇒ increased insulin demand
• Insufficient insulin response (genetics)
• Vicious cycle of glucose and lipid toxicity
• Chronic inflammation and end-organ damage
TYPE 1 DIABETESAutoimmune disease
No insulin
High blood glucose
Complications
TYPE 2 DIABETES*
*Any diabetes where an exact cause cannot be identified.
Thorn et al. (2005) Diabetes Care 28:2019–2024
TYPE 1 DIABETESAutoimmune disease
No insulin
High blood glucose
Life style and genetics
Insulin deficiency
Diabetes (type 2)
Complications
Vascular diseases
SYSTEMIC METABOLIC DISEASE
Obesity
Metabolic stress
COMPLEX DISEASE CHALLENGE• Combination of environment and genes
- network of inter-dependent causes. • Chronic diseases, gradual development
- ambiguous lines between health and disease. • Quantitative biology, qualitative end-points
- diagnostic definitions used as risk factors.
COMPLEX DISEASE CHALLENGE• Noisy diagnoses and stochastic end-points
- poor predictive performance. • Datasets and experiments are unique
- poor performance outside original study. • Individual baseline state is unique
- single time point is of limited value.
OBJECTIVES• Understand the statistical patterns within a
cohort of patients with type 1 diabetes. • Typical lipoprotein characteristics associated
with complications and mortality. • Conceptually simple framework that can
handle large number of variables, discrete and continuous traits, and missing data.
• Four traits: - hairiness - head size - eye size - mood.
Basic statistics - Easy interpretation - Pre-defined groups - Impractical for complex data
Full data - Easy interpretation - High resolution - Impractical for large datasets
Regression and classification - Interpreted through parameters - High descriptive power - Danger of over-fitting
ORGANIZE SAMPLES IN 2D BASED ON SIMILARITY
SUMMARIZE LOCAL SAMPLES
N samples
K model phenotypes, K ≪ N
SUMMARY OF PHENOTYPIC DIVERSITY
Bar height indicates statistically normalized deviation
VISUALIZATION OF TRAIT VALUES
Eye size Hairiness Mood
TRAIT ASSOCIATIONS
Eye size
Hai
rine
ss
Tukiainen et al. Biochem Biophys Res Commun, 375:356-361, 2008Cognitive decline and Alzheimer's
Kumpula et al. J Lipid Res 51:431-439, 2010Lipoprotein structure and composition
Intima media thicknessWurtz et al 2010
Mäkinen et al. J Proteome Res 11:1782-1790, 2012Kidney disease progression
Diabetic complicationsMäkinen et al. Mol Syst Biol 4:167, 2008
Lipoprotein subclasses in type 1 diabetesMäkinen et al. J Intern Med 273:383–395, 2013
STUDY DESIGN
The Finnish Diabetic Nephropathy Study Folkhälsan Research Center, Helsinki, Finland
Healthy C-peptide > 0.2 nmol/L
DIABETES TYPE
Target HbA1c < 7%
DIABETES MANAGEMENT
Healthy AER < 30 mg/24h
AER = albumin excretion rate
KIDNEY DISEASE
Glycated Hb (%)
Log
AER
RAW DATA POINTSr = 0.25 −log P > 17
AER
(m
g/24
h)
HbA1c (%)
300
30
10
8.0 9.0
AER = albumin excretion rate
AER
(m
g/24
h)
HbA1c (%)
300
30
10
8.0 9.0
AER = albumin excretion rate
AER
(m
g/24
h)
HbA1c (%)
300
30
10
8.0 9.0
AER = albumin excretion rate
MODEL PHENOTYPES
AER
(m
g/24
h)
HbA1c (%)
300
30
10
8.0 9.0
AER = albumin excretion rate
Biomarkers used in diagnoses were included in training set.
Sex differences were adjusted before analysis.
P = 0.66 −log P > 13 −log P > 20
−log P > 15 −log P > 12 −log P > 8
Height (cm)
Trig
lyce
ride
s (m
mol
/L)
RAW DATA POINTS
r = −0.10 −log P > 8
MODEL PHENOTYPES
Height (cm)
Trig
lyce
ride
s (m
mol
/L)
Wadén et al. (2005) Diabetes 28:2019–2024
Laser-treated for eye disease n = 1181 Large vessel diseases
n = 343n = 268
Laser-treated for eye disease n = 1181 Large vessel diseases
n = 343n = 268
−log P > 30 −log P > 22
Prevalence of retinopathy
Prevalence of vascular disease
MODEL PHENOTYPES
FULL STUDY DESIGN
The Finnish Diabetic Nephropathy Study Folkhälsan Research Center, Helsinki, Finland
−log P > 30 −log P > 22
−log P > 6 −log P > 17 −log P > 28
Urinary AER (mg/24h)
All-
caus
e m
orta
lity
MODEL PHENOTYPES
STANDARDIZED MORTALITY RATE
AER < 30 30 < AER < 300 AER > 300 End-stage renal disease
0.8
2.7
9.2
18.3
Groop et al. (2009) Diabetes 58:1651-1658
−log P > 6 −log P > 17 −log P > 28
P = 0.086 P = 0.036 P < 0.001
TAKE-HOME MESSAGE?• Carefully investigate observed lipoprotein
characteristics. • Determine overall measures that are
- applicable to all lipoprotein subclasses - easy to relate to existing medical literature.
• Test if these phenotypes are related to clinical data.
LIPOPROTEIN LIPIDSTriglycerides
Cholesterol
SIMPLIFIED LIPOPROTEIN LIPIDS
Phenotype I Lowest cholesterol
Phenotype II Lowest TG:C ratio
Phenotype III Highest TG:C ratio
Normal AER at baseline 93% 77% 8%Microalbuminuria at baseline 5% 12% 7%Macroalbuminuria at baseline 2% 5% 57%End-stage renal disease at baseline <1% 4% 27%
Cholesterol (mmol/L) 3.9 5.0 5.3Triglycerides (mmol/L) 0.76 0.81 2.11HDL cholesterol (mmol/L) 1.22 1.82 1.06
Recommended cholesterol < 5.0 mmol/L
Recommended triglycerides < 1.7 mmol/L
Recommended HDL cholesterol > 1.1 mmol/L (men) Recommended HDL cholesterol > 1.3 mmol/L (women)
Phenotype I Lowest cholesterol
Phenotype II Lowest TG:C ratio
Phenotype III Highest TG:C ratio
Type 1 diabetes duration (years) 12 (11 - 13) 19 (21 - 23) 30 (28 - 31)
Body-mass index (kg/m2) 23.4 (23.1 - 23.8) 23.8 (23.6 - 24.2) 25.6 (25.0 - 26.4)Insulin dose (IU/kg) 0.67 (0.63 - 0.71) 0.66 (0.64 - 0.68) 0.69 (0.66 - 0.72)
Systolic blood pressure (mmHg) 124 (122 - 126) 131 (129 - 133) 145 (141 - 149)Diastolic blood pressure (mmHg) 76 (75 - 77) 80 (79 - 81) 84 (82 - 85)
Hemoglobin A1c (%) 7.5 8.1 9.1
Urinary albumin excretion (mg/24h) 8.6 13 596Estimated GFR (mL/min per 1.73m2) 105 94 48
C-reactive protein (mg/L) 1.15 1.44 3.86
Serum adiponectin (mg/L) 9.3 16 18
Phenotype I Lowest cholesterol
Phenotype II Lowest TG:C ratio
Phenotype III Highest TG:C ratio
Progression from normal AER 4% (2% - 6%) 2% (1% - 4%) 10% (4% - 20%)
Progression from microalbuminuria <1% 15% (6% - 29%) 38% (18% - 57%)
Progression from macroalbuminuria 1% (0 - 4%) 13% (3% - 26%) 42% (31% - 51%)
Deceased at follow-up <3% 6% (4% - 9%) 40% (32% - 47%)
Excess enrichment of triglycerides across every lipoprotein subclass is a part of a metabolic phenotype with high vascular risk in type 1 diabetes
Life style and genetics
Insulin deficiency
Diabetes Vascular diseases
SYSTEMIC METABOLIC DISEASE
Obesity
Metabolic stress
SCORING SYSTEM FOR METABOLIC SYNDROME
• Obesity • High blood glucose
(or diabetes) • Abnormal blood lipids
(TG and HDLC) • High blood pressure
Unhealthy life
Systemic metabolic stress
Insulin resistance
Type 2 diabetes
Heart attack Stroke
Obesity
Unhealthy life
Systemic metabolic stress
Insulin resistance
Kidney disease
Heart attack Stroke
Obesity
Type 1 diabetes
Aging
Basic statistics - Easy interpretation - Pre-defined groups - Impractical for complex data
Full data - Easy interpretation - High resolution - Impractical for large datasets
Regression and classification - Interpreted through parameters - High descriptive power - Danger for over-fitting
ACKNOWLEDGMENTSFolkhälsan Research Center The FinnDiane Group Prof Per-Henrik Groop Carol Forsblom Markku Lehto Lena M Thorn Valma Harjutsalo University of Oulu &
University of Eastern Finland Prof Mika Ala-Korpela Pasi Soininen Tuulia Tynkkynen Antti Kangas
Aalto University School of Science and Tech. Prof Kimmo Kaski Tomi Peltola