Raport de cercetare - Lorentz JÄNTSCHI

y' = (y1', y2', …, ym') = Xc = Hy
4. Fix n, the number of leaving molecules.
5. Loop over all the molecular subsets M(n). For every subset:
5.1. Solve the linear system
5.2. Keep the predicted values.
In QSAR studies it is very common to search for an optimal set of representative descriptors. The
algorithm described in previous section can be conceived as to be a tool which gives a statistic, rcv
attached to a set of them, those defining the matrix X. It is obvious that the algorithm can be
repeated as many times as sets of descriptors are tested in a QSAR project. Hence, the whole
procedure furnishes a new method for selecting the best set of descriptors according to the criteria
of maximization of the value rcv. This will be presumably the most immediate utility of the
LnOT. In this case, in order to speed up the process, it is not only recommended to implement the
practical expressions outlined in previous section but also not to perform the gaussian fitting
described above. It is faster to obtain the correlation coefficients directly from the set of arithmetic
mean values attached to every series of molecular cross-validated values.
The leave-n-out theorem has been demonstrated.and as a consequence, general and explicit
expressions attached to full linear leave-many-out cross-validation processes have been given. The
formulation will allow to easily construct computer codes oriented to be applied to QSAR
problems and other fields.
÷ Lucrare: A partial least squares regression study with antioxidant flavonoid compounds
÷ Autori: Karen C. Weber, Kathia M. Honorio, Aline T. Bruni, Adriano D. Andricopulo, Alberico
B. F. da Silva
÷ Sursa: Structural Chemistry, 2006, Volum 17(3), p. 307-313
÷ Rezumat:
The quantitative structure-activity relationship of a set of 19 flavonoid compounds presenting
antioxidant activity was studied by means of PLS (Partial Least Squares) regression. The
optimization of the structures and calculation of electronic properties were done by using
thesemiempirical method AM1. A reliable model (r 2 =0.806 and q 2 =0.730) was obtained and
from this model it was possible to consider some aspects of the structure of the flavonoid
compounds studied that are related with their free radical scavenging ability. The quality of the
PLS model obtained in this work indicates that it can be used in order to design new flavonoid
compounds that present ability to scavenge free radicals.
÷ Detalii de interes:
From Table 5 we can see that the samples are very well fitted, as the residuals of prediction are
very small (0.02–0.04). This constitutes a strong indication that the model is reliable since it
consistently predicted the TEAC values of the test set.Through the validation tests applied in the
model we can suggest that it can be used for understanding and predicting the antioxidant activity
of new flavonoid compounds before their synthesis and tests against free radicals. Furthermore,
the use of a multivariate methodology to build a regression model allows us to deal with various
features of flavonoid compounds that can be related to their reactivity with free radicals,
especially the influence of substitutents that are not hydroxyls, what constitutes a more complete
description of the problem than taking into account only the number or the presence of hydroxyl
groups at certain positions.
36