You are on page 1of 32

Turning partial orders into

total orders, a step-by-step


approach
Guillermo Restrepoa,b, Eugenio J. Llanosc and Rainer
Brüggemannd

a
Laboratorio de Química Teórica, Universidad de Pamplona, Colombia

b
Department of Mathematics, University of Bayreuth, Germany

Corporación Scio, Bogotá, Colombia


c

d
Department of Ecohydrology, Leibniz-Institute of Freshwater Ecology
and Inland Fisheries, Berlin, Germany
What is a partial order?

A binary relation on P is called a


partial order if:
1. x ∈P ⇒ x ≤x,
2. x, y ∈P, x ≤y and y ≤x ⇒ x = y,
3. x, y, z ∈P, x ≤y and y ≤z ⇒ x ≤
z.

Some examples
Ø

a b c

ab ac bc When
ranking…

abc

Hasse diagrams
a

b
a c

b Ranking c

d e
d

e
Subjectivit
n-dimensional y one-
problem dimensional
problem
f
Rn R
W
hy?
Rankin Decision
g making
Selecting a good set of descriptors
a set of chemicals for biological
screening
a good
girlfriend/boyfriend
a good airline or bus! for coming to
Duluth
A B C D E F
d b d b d b
q1 q2 q3
a 0 8 2 b d b d b d b d

b 9 13 12
a e e e e e a a
c 2 1 5
c c a a e e
d 10 14 11 c

e 4 3 7 Hasse diagram
a a c c c c

Data matrix Total set of linear extensions

What is the reason of the


incomparabilities?
Conflictive descriptor
values
Is it possible to turn incomparabilities into comparabilities
considering as many subjectivity as possible?
Yes, it is! Aggregating
anticorrelated or not so
correlated descriptors
q and q are the most anticorrelated
2 3
properties in the example
 ( x)  g q2 ( x)  (1  g ) q3 ( x)
Indicates
 ( y )  g q2 ( y )  (1  g ) q3 ( y ) “priority”
x y

g' z g''
x gc y

y x~y x 1
gc 
q ( x)  q2 ( y )
z z z 1 2
q3 ( x)  q3 ( y )
 ( x) =  ( y )
g Still
1 incomparabilitie
d
a || e s!
b
S4 b<d
a e
a || c
c
0.562
b d d
a<e
b
b<d
q1, ϕ
S3
e
a e a || c
a c
0.556
a<e b d
c
S2 b || d e
q1, q2, q3
a || c a c
0.357
b d
a<e
e
S1 b || d
c
a<c
a
0
A C E
We need a new aggregation dd dd dd
bb bb bb

 ( x)  h  ( x)  (1  h) q1 ( x) e e
aa
aa

cc e

 ( x)  (1  h) q1 ( x)  hg q2 ( x)  h(1  g ) q3 ( x) aa


0.50
cc

0.92 1
c
h
ψ 0
S1(S4) S2(S4) S3(S4)
g
A C
1 dd dd
d
a || e bb bb
b
S4 b<d ee e
a e
a || c cc aa
c
0.562
d
ψ aa cc
b d a<e 0 0.54 1
h
b S1(S3) S2(S3)
b<d
q1, ϕ
S3
e
a e a || c
a c A B D
0.556 dd bb bb
a<e b d
c
S2 b || d e bb dd dd
q1, q2, q3
a || c a c ψ e e e
0.357
b d cc cc a
a<e
e a aa cc
b || d
S1
c 0 0.78 0.85 1 h
a<c S1(S2) S2(S2) S3(S2)
a
0
A B
ψ
Stability field d b
b d
Region of the priorities e e

space where the ranking c


a
c
a
does not change 0 0.61 1 h
S1(S1) S2(S1)
q1, q2, q3

S1 S2 S3 S4
g ∈ (0,0.357) g ∈ (0.357,0.556) g ∈ (0.556,0.562) g ∈ (0.562,1)

ψ ψ ψ ψ

S1(S1)
S2(S1)
h ∈ (0,0.61) S3(S2)
h ∈ (0.61,1)
h ∈ (0.85,1)

S1(S2) S1(S4)
h ∈ (0,0.78) h ∈ (0,0.50)
S1(S3)
S2(S2) h ∈ (0,0.54) S2(S4)

h (0.78,0.85) ∈
h (0.50,0.92)

S2(S3) S3(S4)

h (0.54,1) ∈
h (0.92,1)

bb dd bb dd dd

dd bb dd bb bb

ee ee ee ee aa

cc cc aa aa ee

aa aa cc cc cc
B A D C E
h
1
E
0.92 D
0.85
A B C D E F
B B d b d b d b
0.78 C
C b d b d b d
0.61
0.54 e e e e a a
0.5
c c a a e e
A
A a a c c c c
A A

0 g
0 S1 0.357 S2 0.556 S3 0.562 S4 1

The area indicates the probability of a linear order


1st
example
0 problematic substances in the aquatic environment

2 1,2,3-Trichlorobenzene C: 90 percentile of
the observed
4 1,2-Dichlorobenzene
concentration in EU
7 1,3-Dichlorobenzene waters
12 Acenaphthene
19 Azinphos-methyl
PNEC: Predicted no
35 Desisopropylatrazine effect
38 Dichlorvos concentration
65 Metazachlor
70 Parathion-methyl
EFSi: Indirect
74 Propazine
aquatic effect score
7 4
256 linear
extensions!

2 12 35 74
Least
C and EFSi
correlated
65 properties

19 38

ϕ ( x) = g ⋅ C ( x) + (1 − g ) EFSi ( x)

70

ψ ( x) = h ⋅ ϕ ( x) + (1 − h) ⋅ PNEC ( x)

For various g values, one can group similar ones


1
0.936

h 0.5

0.290

0.197

0.078

0
0 0.5 1
g 0.984
1
0.936 7 4 1,2-DCB

4 7 1,3-DCB

2 2

74 74

12 12
h 0.5
35 35

65 65

38 19
0.290

19 70
0.197
70 38
0.078

0
0 0.5 1 % 84 15
g 0.984
2nd Refrigerants
example

Destruction of
CFCs (chlorofluorocarbons)
Refrigerants stratospheric
HCFCs (hydrochlorofluorocarbons) ozone

ALT (atmospheric lifetime)


CFCs HCFCs ODP (ozone depletion potential)
GWP (global warming potential)

Possible
solutions
CFCs HCFCs Production Alternative ALT, ODP, GWP
Consumption refrigerants
Refrigerants
must be
assessed
“Good” “Bad”

refrigerants

Assessmen
t

Ranking!
Refrigerants

Octafluorocyclobuta
Chlorofluorocarbons Hydrofluorocarbons
1 CCl3F ne
11 CHF3 27 (CF2)4
2 CCl2F2
12 CH2F2
3 CCl2F-CClF2 Alkyl-fluoroalkylethers
13 CHF2-CF3
4 CClF2-CClF2 28 CF3-O-CH3
14 CH2F-CF3
29 CF3-CF2-O-CH3
15 CH3-CF3
30 CF3-(CF2)2-O-CH3
16 CH3-CHF2
31 CF3-(CF2)3-O-CH3
17 CHF2-CH2-CF3
32 CF3-(CF2)3-O-C2H5
18 CF3-CH2-CF3
33 CF3-(CF2)6-O-C2H5
19 CF3-CHF-CF3
Hydrochlorofluorocarb Di(fluoroalkyl)ethers
Hydrocarbons
ons 20 CH3-CH2-CH3 34 CF3-O-CHF2
50 CHClF
60 CHCl2-CF3 21 CH3-(CH2)2-CH3 35 CHF2-O-CHF2
70 CHClF-CF3 22 C(CH3)3
Chloromethanes
80 CCl2F-CH3 23 CH3-(CH2)3-CH3 36 CH2Cl2
90 CClF2-CH3 24 CH3-CH=CH2 37 CH3Cl
10 CHCl2F
25 CO2 38 CF3I

39 CH3-O-
CH3
26 CClBrF2 40
DFAE CFC BCF HFC PFC

29 1 2 33 35 22 8 23

15
30
7
12
10
3
11
24
AFAE 6
14
39 5
26 9
34 HCFC
25
40
13
28 4
DFAE:
27 CO2 Di(fluoroalkyl)ethers
HC CFC:
Chlorofluorocarbons
21 32 BCF:
16 Bromochlorofluorocarb
CM HFC:
31
on
18 OFB:
Hydrofluorocarbons
DME 36 AFAE:
17 Octafluorocyclobutane
HCFC
37 FIM Alkyl-fluoroalkylethers
38 :
Hydrochlorofluorocarbo
CM:
NH3 ns
20 19 DME:
Chloromethanes
FIM:
Dimethyl ether
HC:
DFAE CFC BCF HFC PFC

29 1 2 33 35 22 8 23

15
30
7
12
10
3
11
24
AFAE 6
14
39 5
26 9
34 HCFC
25
40
13
28 4
DFAE:
27 CO2 Di(fluoroalkyl)ethers
HC CFC:
Chlorofluorocarbons
21 32 BCF:
16 Bromochlorofluorocarb
CM HFC:
31
on
18 OFB:
Hydrofluorocarbons
DME 36 AFAE:
17 Octafluorocyclobutane
HCFC
37 FIM Alkyl-fluoroalkylethers
38 :
Hydrochlorofluorocarbo
CM:
NH3 ns
20 19 DME:
Chloromethanes
FIM:
Dimethyl ether
HC:
22 2 1 33 35 8 23 29

6
7

32

40 21
39
38

36

16

37

ALT priority
 ( x)  g ALT ( x)  (1  g ) ODP ( x)
ODP priority
S1 S2 S3 S4 S5 S6 S7 S8 S9 S10 S11 S12
H1 H2 H3 H4 H5 H6 H7 H8 H9 H10 H11

g
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
S10
S12
S11 Problemati
ALT 8 8 29 23 23 29 ALT
Many S9 c
8 29 23
problemati 2 35 substances
22 82 35
c S3 S6
29
S8
S4 23
8 (HFC)
ODP S52 35 ODP
substances 8
8 88
2 2229S1
22 S2
23 23 22 22
33
229 29 29 23 23 29 (DFAE)
8 2 S7
33
29 23 22
8
2 22 22
81 8 1 2 2 233 33 29 29 29 35
3335
35 222322
23 (PFC)
1
1 35 35
1 23 1 22 35 133 35
35
21 7 35
11 22 7 33
6 33 33
7 33
23
21 6 6
7 22 3371
1 40
6 32 6 7 7 33 7
32 640 396
21 1
21 6 6 6 7 739 7
6
21 324021
21 2132
40
6 40 32 40 39 39
40 39
21 21 32 40 1639
21 32 40
40 39 739
403936 39 39
Not so problematic
16
40 16 39substances
32 32 36 36
21
38 32 3716 1616
16
32
38 36 3616 3637 38 Ammonia
36 16 16
16
36 38 36 36
38 36 37
16
37 Dimethyl ether
37
38 38 38 36 3737 37
38 38 37 Invariant to
38 38 37 37 37
38 37
 ( x)  h  ( x)  (1  h) GWP( x)
 ( x)  hg ALT ( x)  h(1  g ) ODP( x)  (1  h) GWP( x)

h
1
E
0.92 D
0.85
B B
0.78 C
C

0.61
0.54
0.5

A
A
A
A

0 g
0 S1 0.357 S2 0.556 S3 0.562 S4 1
1
Landscape
of weights
(priorities)

h 0.5

0
0 0.5 1
g
109 stability fields!

How about their similarities?

Cluster
analysis

70 different stability
fields
h
1

0.5

0 g
0 0.5 1
S1 S2 S3 S4 S5 S6 S7 S8 S9 S10 S11 S12
29 29 29 22 23 22 23 23 23 23 22 22 22 22 22 23 23 23 23 23 23 23 22 23 23 22 22

8 8 23 29 22 23 8 8 8 8 2 2 35 1 23 22 22 22 8 22 22 35 23 22 35 1 1

2 2 8 8 29 2 29 29 35 35 35 35 2 35 35 35 35 35 22 8 35 8 1 35 8 33 33

23 35 2 2 8 35 35 35 29 29 29 33 1 2 2 2 2 8 29 29 8 22 35 1 29 35 35

35 23 35 35 2 29 2 2 2 2 8 1 33 33 33 33 8 2 35 35 2 29 33 33 21 2 2

33 33 33 23 35 8 33 33 22 33 33 29 23 29 1 1 29 29 2 2 33 2 2 2 2 6 23

1 1 22 33 33 33 22 22 33 1 23 8 29 8 8 8 33 33 33 33 29 33 8 8 33 7 6

7 22 1 1 1 1 1 1 1 21 1 23 8 23 29 29 1 1 1 1 1 21 6 29 1 32 7

22 7 7 7 7 7 7 7 21 7 7 7 7 6 7 21 7 21 7 7 21 1 29 21 7 8 8

6 6 6 6 6 6 6 21 7 22 6 6 6 7 6 7 6 7 6 6 7 7 7 6 22 23 32

40 40 40 40 40 40 40 6 6 6 40 40 40 32 21 6 21 6 40 21 6 6 21 7 6 29 29

39 39 39 39 39 39 39 40 39 39 39 39 39 40 40 39 40 39 39 40 39 39 32 32 39 21 21

16 32 21 32 21 21 21 39 40 40 32 32 32 39 39 40 39 40 21 39 40 40 39 39 40 39 39

21 16 32 21 32 32 32 32 32 32 21 21 21 16 32 32 32 32 32 32 32 32 40 40 32 40 40

32 21 16 16 16 16 16 16 16 16 16 16 16 21 16 16 16 16 16 16 38 38 38 38 38 38 38

36 36 36 36 36 36 36 38 38 38 36 36 36 36 38 38 38 38 36 38 36 36 36 36 36 16 36

37 37 37 37 37 37 38 36 36 36 37 37 37 37 36 36 36 36 38 36 16 16 16 16 16 36 16

38 38 38 38 38 38 37 37 37 37 38 38 38 38 37 37 37 37 37 37 37 37 37 37 37 37 37

A B C D E F G H I J K L M N Ñ O P Q R S T U V W X Y Z
% 15 3 50 1 5 2 2 5 <1 <1 1 <1 <1 <1 3 1 4 2 2 2 <1 <1 <1 1 <1 <1 <1
29 29 DFAE
1
8 23 PFC
2 8 HFC
23 2 CFC
35 35 CFC
33 33 CFC
1 22 BCF
7 1 CFC
h 0.5 22 7 HCFC
6 6 HCFC
40 40 AFAE
39 39 AFAE
16 21 CO2
21 32 CM
32 16 HC
36 36 FIM
37 37 DME

0 38 38 NH3
0 0.5 1
g
% 15 50
Example of application
We want to see how the ranking
looks like if we give the following
priorities: 10 % ALT
40 % ODP
50 % GWP

 ( x)  hg ALT ( x)  h(1  g ) ODP( x)  (1  h) GWP( x)

hg = 0.1 h(1 − g ) = 0.4 1 − h = 0.5

g = 0.2
h = 0.5
1 23 PFC

22 BCF

29 DFAE

8 HFC

2 CFC

35 CFC

33 CFC

1 CFC

h 0.5 7 HCFC

6 HCFC

40 AFAE

39 AFAE

21 CO2

32 CM

16 HC

36 FIM

37 DME

38 NH3

0
0 0.2 0.5 1 % 5
g
Conclusions
The method here developed is a tool for answering the
following questions:

• How does the ranking look like if one prioritises


certain object descriptors?

• Can one trace back the priorities needed to have


a particular ranking?

• What is the probability of obtaining a certain


ranking by prioritisation?
References
Brüggemann, R.; Voigt, K.; Restrepo, G.; Simon, U. The
concept of stability fields and hot spots in ranking of
environmental chemicals. Environ. Modell. Softw. 2008, 23,
1000-1012.

Restrepo, G.; Weckert, M.; Brüggemann, R.; Gerstmann, S.;


Frank, H. Ranking of refrigerants, Environ. Sci. Technol.
2008, 42, 2925-2930.

Restrepo, G.; Brüggemann, R.; Weckert, M.; Gerstmann, S.;


Frank, H. Ranking patterns, an application to refrigerants.
MATCH Commun. Math. Comput. Chem. 2008, 59, 555-
584.

Restrepo, G.; Brüggemann, R. Partially ordered sets in the


analysis of alkanes fate in rivers. Croat. Chem. Acta 2007,
80, 261-270.
Thank
you!
Questions
?

You might also like