Quantitative structure-property relationship (QSPR) for the adsorption of organic compounds onto activated carbon cloth: Comparison between multiple linear regression and neural network
C. Brasquet et al., Quantitative structure-property relationship (QSPR) for the adsorption of organic compounds onto activated carbon cloth: Comparison between multiple linear regression and neural network, ENV SCI TEC, 33(23), 1999, pp. 4226-4231
Citations number
43
Categorie Soggetti
Environment/Ecology,"Environmental Engineering & Energy
The adsorption of 55 organic compounds is carried out onto a recently disco
vered adsorbent, activated carbon cloth. Isotherms are modeled using the Fr
eundlich classical model, and the large database generated allows qualitati
ve assumptions about the adsorption mechanism. However, to confirm these as
sumptions, a quantitative structure-property relationship methodology is us
ed to assess the correlation between an adsorbability parameter (expressed
using the Freundlich parameter K) and topological indices related to the co
mpounds molecular structure (molecular connectivity indices, MCI). This cor
relation is set up by mean of two different statistical tools, multiple lin
ear regression (MLR) and neural network (NN). A principal component analysi
s is carried out to generate new and uncorrelated variables. It enables the
relations between the MCI to be analyzed, but the multiple linear regressi
on assessed using the principal components (PCs) has a poor statistical qua
lity and introduces high order PCs, too inaccurate for an explanation of th
e adsorption mechanism. The correlations are thus set up using the original
variables (MCI), and both statistical teals, multiple linear regression an
d neural network, are compared from a descriptive and predictive point of v
iew. To compare the predictive ability of both methods, a test database of
10 organic compounds is used. Results show the good descriptive ability of
NN compared with that of MLR, with more than 68% variance explained by NN,
whereas MLR allows only 44% variance explanation. However, the predictive a
bility of NN seems to be low, especially when the structure of the test com
pounds is not well described in the training database. The good descriptive
ability of NN is then exploited to carry out a variable analysis using the
Garson weight partitioning method and to give information about the adsorp
tion process. This study shows that fiat molecules seem to be better adsorb
ed onto activated carbon fibers than bulky molecules, because of an adsorpt
ion which is located between the micrographitic planes of fibers. The adsor
ption process occurs via an electron donor-acceptor interaction between the
surface of the activated carbon fiber(donor) and the solute (acceptor). Co
nsequently, the aromatic compounds with electron-withdrawing substituents s
eem to be favored. Furthermore, the lower the solute affinity for the aqueo
us media, the greater seems to be the adsorption.