The Human Phenotypic Disease Network

1
The Human PhenotypicDisease Network
Cesar A. Hidalgo Center for Complex Network
Research
2
Collaborators
Nicholas Christakis Laurie Meneades Dept. of
Health Policy, Harvard Medical School
Christine Teutsch Massachusetts General
Hospital Harvard Medical School
Laszlo Barabasi Nicholas Blumm Center for
Network Science, Northeastern University
3
Some Diseases are known to be related
4
There are no single causes for each disease
ENPP1
HFCS
SLS
Ectonuclide Pyrophosphate Phosphodiesterase 1
High Fructose Corn Syrup
Sedentary Life Style
5
Determining causes is labor and capital intensive
6
Phenotyping is an art
Courtesy of Zoghbi Lab
7
There are many possible disease pairs
ICD-9
15,000 possible diagnoses
8
Millions of doctors phenotype patients every day
9
Ocean of Data
10
Disease 2
Disease 1
11
Disease 1
Disease 2
Affect Same Individuals Significantly more than
Random
12
Methods
13
Study Population
13039018 patients 32341348 records
(hospitalizations)
Fields Available
Identifier, Time of Visit, State, Age, Gender,
Poverty (0-1),
Disease Fields
Up to 10 diagnosis (from 950 cat (3digit) /
14000 (5digit) )
14
Building a Net
N
P1
P2
15
Data
ICD-9 code Disease Name Prevalence
401.9 Unspecified essential hypertension 4163381
414.0 Coronary Atherosclerosis 2679957
428 Congestive Heart Failure 2343516
496 Chronic Airway Obstruction 1878976
427.31 Atrial Fibrillation 1806969
276.5 Volume Depletion Disorder 1533833
276.8 Hypokalemia 1448586
250 Diabetes Mellitus w/o mention of complication 1303342
486 Pneumonia, organism unspecified 1212287
411.1 Intermediate Coronary Syndrome 1137913
16
Word of Caution
17
Positive Associations
Diseases appear are more likely to be correlated
than anti-correlated
18
Building a Net
N
P1
P2
19
Word of Caution
Underestimates overlap of Common Phenotypes
1
N
Overestimate overlap of Rare Phenotypes
20
1 Harley JB. Nature Genetics 39(9) 1053 (2007) 2
Scott LJ et al, Science 316 1341 (2007)
21
Visualizing the Network
22
(No Transcript)
23
Results
24
Patient 1
25
Patient 2
26
Patient 3
27
Patient N
Visit 1
Visit 2
Visit 1
Visit 2
A
a
A
a
B
b
B
b
C
g
C
g
ltfgtr
ltfgtd
28
Patient N
Hidalgo et al Science, 2007
ltfgtr
ltfgtd
ltfgt
29
Patient N
ltfgt
30
(No Transcript)
31
Connectivity and Lethality
Connectivity S f
32
R0.5051
33
(No Transcript)
34
(No Transcript)
35
Prot- robustness
Yeast
62
24
18
Hubs are more essential (lethal)...
Jeong, Mason, Barabasi, and Oltvai, Nature 411,
41-42 (2001)
36
P53 P(k)
p53 network (mammals)
Goh et al, PNAS, 2007
37
Disease Genes
Lethal Genes
38
Somatic mutations
Mendelian disease
39
Online
40
http//hudine.neu.edu
41
http//hudine.neu.edu
42
http//hudine.neu.edu
43
Outlook
44
Search for Disease Causes
45
Search for drug targets/ indications
46
Hospital Design