comb_HTS_10-10

ACS Combinatorial Science REVIEW
Figure 8. Loadings plot of the developed PCA model derived from a

Figure 7. D-Glyph plot of a catalyst library data. Four elements (Cr, Mn, combinatorial library of multicomponent oxides (Mo, Ni, Mn, Cr, and
Co, and Te) are plotted on a 3D quaternary mixtures plot, while three Co) and the selectivity of chemical products (acetone, propionic
activities are mapped to three graphical attributes of a sphere glyph: color is adelhyde, acreolein, and 1,5-hexadiene).18,97,103
mapped to activities for acrolein, size is mapped to activities for 1,
5-hexadiene, and intensity is mapped to activities for acetone. The intensity
of the glyphs is adjusted on the left plot to accentuate the levels of acetone.13 selectivity of metal ions on catalytic products but also can help to
assess the relative impact of these elements.
Importantly, while PCA and other pattern recognition tools
behavior can be modified by the effects of metal/metal or metal/ are helpful in assessing the relative impact of multiple parameters
oxygen/metal interactions. In this respect, it is important to on materials properties, they are not predictive tools. For
know how to choose the “right” combination of metals. Just to predicting of materials properties, other methods should be
consider which combination is appropriate and why is in itself a applied.97 Just as informatics can aid in searching and discovering
challenge, and hence when one considers mixed metal catalysts critical data from combinatorial experiments,71 76 it can also
with 4, 5, and more metals, we clearly have, what appears to be a serve as a predictive modeling tool based on CHT experimenta-
prohibitive problem on exploring in a reasonably timely manner tion. For instance, Li et al. have shown,104 a new approach
the catalytic behavior of such a vast array of chemistries. involving a hybrid data mining to determine which aspects of a
The groups of Maier and Rajan13,102,103 addressed this issue polymeric molecular structure have the most impact on the
through a study of 103 combinatorial catalyst library. The release kinetics behavior for drug delivery (see Figure 9). They
chemistries sampled the complete composition spread of a five used these molecular descriptors to model the release profile of
dimensional search space containing the elements Cr, Co, Mn, proteins from a combinatorial library of biodegradable polymers.
Mo, and Ni prepared by a sol gel recipe and tested by the The mathematical approach for modeling the release profile was
high-throughput screening reactor. To unravel the complexity of based on a combination of genetic algorithms (GA) and support
these materials compositions, a visualization approach shown in vector regression methods. The polymers explored from the
Figure 7 has been utilized to track the raw data from com- combinatorial libraries included linearly varying molar composi-
binatorial experiments. This visualization approach facilitated tions of binary formulations of copolymers, with the protein
the establishment of the framework on how data mining tech- release rate found to increase with decreasing polymer hydro-
niques can be applied to unravel the complexity of combinatorial phobicity. From the new attribute selection approach, it was
experiments. determined that number of nonconjugated backbone COO
The further application of data mining techniques addressed bonds, number of aromatic rings, and number of CH2 bonds
two questions, such as (1) the selectivity of a given metal to were most important for accurately modeling the release kinetics
catalytic activity (i.e., a reaction product) and (2) the relative behavior. This work introduces a new approach for understanding
influence between metal chemistry and catalytic activity. The which factors control drug release and has important implications
loading plot of the developed PCA model provided the correla- for the rational design of new polymeric carriers for drug delivery
tions between the presence of a specific metal species and the In addition to materials composition, materials microstructure
selectivity of a given reaction product (see Figure 8). The relative and processing parameters play equally important effects. Unravel-
correlations of different chemistries and their influence on ing the interactions between these effects is critical. Thus, infor-
the selectivity of chemical reactions can be visualized through the matics techniques is applied to high-throughput screening data
screening of the geometrical relationships between variables from combinatorial synthesis methods. In one representative
represented in this loadings plot. In this case Mo or the pair example, the ability of a polymer system to both modulate its
(Mo, Co) has positive correlations with acetone formation, while effect on the immune response and to deliver a drug or antigen in a
Ni and Mn are tend to inhibit the formation of all the reaction controlled fashion provides an ideal platform for both vaccine
products, especially acetone. This example has shown that the adjuvants and multicomponent implants. Many factors influence
data mining techniques can not only help to identify the relative the innate and adaptive immune responses, including the activation
588 dx.doi.org/10.1021/co200007w |ACS Comb. Sci. 2011, 13, 579–633

comb_HTS_10-10

Uploaded by

Copyright:

Available Formats

comb_HTS_10-10

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

comb_HTS_10-10

Uploaded by

Copyright:

Available Formats

ACS Combinatorial Science REVIEW

Figure 8. Loadings plot of the developed PCA model derived from a

You might also like