2DS01lec5.ppt 396KB Jun 23 2011 08:36:52 PM
2DS01
Statistics 2 for Chemical
Engineering
lecture 5
2
Contents
• high-throughput screening
• combinatorial chemistry
• overview of previous lectures
3
Breakthrough in experimentation
•
•
•
•
robotic sample preparation
miniaturization of reactors
high-level automatization of sensors
pharmaceutical industry:
– routine creation and testing of 1000 to
1000000 distinct compounds (libraries)
• techniques are now also being applied in
material development
• new companies:
– Symyx (www.symyx.com)
– Avantium (www.avantium.com)
4
High-throughput screening
• typical cycle of experimentation:
–
–
–
–
–
thousands of reactions in few hours
few hours of statistical analyses
thousands of reactions in few hours
few hours of statistical analyses
---
• new chemical may be developed in 3 weeks rather
than 3 years
• Which statistical techniques are important?
• How do the classical techniques of the previous lectures fit
in?
• Which new techniques are necessary?
5
combinatorial synthesis approach
6
Multireactor vessels
7
Overview of experimental strategies
combinatorial
methods
factor/level
determinatio
n
chemica
l
intuition
screening
designs
polynomia
l models
response
surface
methods
semiempirical
models
optimal
designs for
non-linear
models
firstprinciples
equations
chemical/physic
al
knowledge
scientific understanding
8
Multistage screening
9
Multireactor vessels
10
Combinatorial explosion: example 2nd year
project
R1-BY2 + R2 –X
R1-R2
• 4 different catalysts, 10 continuous change
equivalence: 0.01-0.10, with mixtures:
• 4 different bases, 10 continuous change
equivalence: 0.01-0.10, with mixtures
• 3 solvents
• temperature: 50C-120 C, steps of 10 C
• 3 choices for both X and R1
• 6 choices for R2
• Total number of possibilities: 3.3 * 107
11
Experimental strategies: combinatorial organic
synthesis
• structural descriptors are calculated for each
compound
• similarity coefficients are calculated between
compound pairs
• compounds are selected using multivariate methods
(based on clustering, dissimilarities, etc.)
Possible because target is single compound
12
Experimental strategies: materials
development
• currently descriptors less well developed
(complex interactions / processing)
• need for other strategies
Common approaches:
1. High-speed array strategies
2. True combinatorial design strategies
13
High-speed array strategies
1. gradient arrays
2. quaternary mask arrays
3. high-speed versions of conventional
experimental designs
14
Gradient arrays
100% A
• continuous spread
• point techniques
100% B
• uniform spacing?
• data analysis?
100% C
15
Quaternary mask arrays
4^5 = 1024 possibilities in
20 sputter operations!!
16
Detail quaternary masks
17
High-speed versions of conventional
designs
• cost of experimentation is low
• high resolution designs are possible
– full factorials
– central composite designs
– special cubic mixture designs
• 3rd and higher order interactions are
important !
• use in second stage of screening (after “hit”
has been found)
• complicated experiments may require extra
statistical features (nesting, random effects)
18
True combinatorial design strategies
•
•
•
•
split-and-pool / split-and-combine
representational strategy
index library strategy
all 2-way combinations strategy
19
20
Representational strategy
• similar to
one-factor-at-a-time
strategy
• will not identify
interactions
21
Index library strategy
is limited strategy
like representational
strategy
22
All 2-way combinations strategy
• 19*18/2 = 171
• for all 3-way combinations: (19*18*17)/(1*2*3) =
969 runs
23
N-way combinations
• gain possible by noting that 1 2 3 4 5 contains
– 10 2-way combinations
– 10 3-way combinations
– 5 4-way combinations
• orthogonal arrays
• Latin squares
24
Some WWW sites on combinatorial
chemistry
• http://www.combichem.net/
• Homepage of Furka:
http://szerves.chem.elte.hu/Furka/
• http://www.aae.enscm.fr/anciens/94mc/combchem.htm
• http://www.combinatorial.com/
• Molecular diversity page:
http://www.5z.com/divinfo/
• Links to several papers:
http://chemengineer.miningco.com/cs/combinatori
alchem/index.htm
25
Literature
• J.N. Cawse, Experimental Strategies for Combinatorial
and High-Throughput Materials Development, Acc.
Chem. Res. 34 (2001), 213-221
• R. Hoogenboom et al., Combinatorial Methods,
Automated Synthesis and High-Throughput Screening
in Polymer Research: Past and Present, Macromol.
Rapid Commun. 24 (2003), 15-32
• G-J.M. Gruter et al., R&D Intensification in Polymer
Catalyst and Product Development by Using HighThroughput Experimentation and Simulation,
Macromol. Rapid Commun. 24 (2003), 73-80.
• W.A. Warr, Combinatorial Chemistry and Molecular
Diversity. An Overview, J. Chem. Inf. Comput. Sci. (37)
1997, 134-140.
Statistics 2 for Chemical
Engineering
lecture 5
2
Contents
• high-throughput screening
• combinatorial chemistry
• overview of previous lectures
3
Breakthrough in experimentation
•
•
•
•
robotic sample preparation
miniaturization of reactors
high-level automatization of sensors
pharmaceutical industry:
– routine creation and testing of 1000 to
1000000 distinct compounds (libraries)
• techniques are now also being applied in
material development
• new companies:
– Symyx (www.symyx.com)
– Avantium (www.avantium.com)
4
High-throughput screening
• typical cycle of experimentation:
–
–
–
–
–
thousands of reactions in few hours
few hours of statistical analyses
thousands of reactions in few hours
few hours of statistical analyses
---
• new chemical may be developed in 3 weeks rather
than 3 years
• Which statistical techniques are important?
• How do the classical techniques of the previous lectures fit
in?
• Which new techniques are necessary?
5
combinatorial synthesis approach
6
Multireactor vessels
7
Overview of experimental strategies
combinatorial
methods
factor/level
determinatio
n
chemica
l
intuition
screening
designs
polynomia
l models
response
surface
methods
semiempirical
models
optimal
designs for
non-linear
models
firstprinciples
equations
chemical/physic
al
knowledge
scientific understanding
8
Multistage screening
9
Multireactor vessels
10
Combinatorial explosion: example 2nd year
project
R1-BY2 + R2 –X
R1-R2
• 4 different catalysts, 10 continuous change
equivalence: 0.01-0.10, with mixtures:
• 4 different bases, 10 continuous change
equivalence: 0.01-0.10, with mixtures
• 3 solvents
• temperature: 50C-120 C, steps of 10 C
• 3 choices for both X and R1
• 6 choices for R2
• Total number of possibilities: 3.3 * 107
11
Experimental strategies: combinatorial organic
synthesis
• structural descriptors are calculated for each
compound
• similarity coefficients are calculated between
compound pairs
• compounds are selected using multivariate methods
(based on clustering, dissimilarities, etc.)
Possible because target is single compound
12
Experimental strategies: materials
development
• currently descriptors less well developed
(complex interactions / processing)
• need for other strategies
Common approaches:
1. High-speed array strategies
2. True combinatorial design strategies
13
High-speed array strategies
1. gradient arrays
2. quaternary mask arrays
3. high-speed versions of conventional
experimental designs
14
Gradient arrays
100% A
• continuous spread
• point techniques
100% B
• uniform spacing?
• data analysis?
100% C
15
Quaternary mask arrays
4^5 = 1024 possibilities in
20 sputter operations!!
16
Detail quaternary masks
17
High-speed versions of conventional
designs
• cost of experimentation is low
• high resolution designs are possible
– full factorials
– central composite designs
– special cubic mixture designs
• 3rd and higher order interactions are
important !
• use in second stage of screening (after “hit”
has been found)
• complicated experiments may require extra
statistical features (nesting, random effects)
18
True combinatorial design strategies
•
•
•
•
split-and-pool / split-and-combine
representational strategy
index library strategy
all 2-way combinations strategy
19
20
Representational strategy
• similar to
one-factor-at-a-time
strategy
• will not identify
interactions
21
Index library strategy
is limited strategy
like representational
strategy
22
All 2-way combinations strategy
• 19*18/2 = 171
• for all 3-way combinations: (19*18*17)/(1*2*3) =
969 runs
23
N-way combinations
• gain possible by noting that 1 2 3 4 5 contains
– 10 2-way combinations
– 10 3-way combinations
– 5 4-way combinations
• orthogonal arrays
• Latin squares
24
Some WWW sites on combinatorial
chemistry
• http://www.combichem.net/
• Homepage of Furka:
http://szerves.chem.elte.hu/Furka/
• http://www.aae.enscm.fr/anciens/94mc/combchem.htm
• http://www.combinatorial.com/
• Molecular diversity page:
http://www.5z.com/divinfo/
• Links to several papers:
http://chemengineer.miningco.com/cs/combinatori
alchem/index.htm
25
Literature
• J.N. Cawse, Experimental Strategies for Combinatorial
and High-Throughput Materials Development, Acc.
Chem. Res. 34 (2001), 213-221
• R. Hoogenboom et al., Combinatorial Methods,
Automated Synthesis and High-Throughput Screening
in Polymer Research: Past and Present, Macromol.
Rapid Commun. 24 (2003), 15-32
• G-J.M. Gruter et al., R&D Intensification in Polymer
Catalyst and Product Development by Using HighThroughput Experimentation and Simulation,
Macromol. Rapid Commun. 24 (2003), 73-80.
• W.A. Warr, Combinatorial Chemistry and Molecular
Diversity. An Overview, J. Chem. Inf. Comput. Sci. (37)
1997, 134-140.