QSARs for Predicting Cation Toxicity, Bioconcentration, Biosorption, and Binding - Fundamental QSARs for Metal Ions - page 191

predicted values in the prediction (MAPE) = 21.63. QSARs with the smallest AIC
are likely to have the most information regardless of the number of independent
variables.
Of the 27 QSARs that used ΔE 0 to predict cation toxicities, 13 used combina-
tions of ΔE 0 , AN, and ΔIP (labeled as 1 in Table 5.13 ) , 5 only used ΔE 0 (labeled
as 2 in Table 5.13), and 2 used combinations of ΔE 0 , AW, and ΔIP (labeled as 3
in Table 5.13). Of the 13 QSARs that used combinations of ΔE 0 , AN, and ΔI P,
Kaiser's (1980) QSAR for 12 cations had the highest r 2 ; adding In 3+ , Rh 3+ , and Pd 2+
to those 12 cations decreased r 2 (Table 5.13). Of the 5 QSARs that only used ΔE 0 ,
the Mendes et al. (2010) QSAR not only had the highest coefficient of determina-
tion, but the second-highest number of cations (18), only missing Cr 3+ and La 3+
from the McCloskey et al. (1996) QSAR for 20 cations (Table 5.13). Of the 2 that
used combinations of ΔE 0 , AW, and ΔIP, reducing the number of cations from 12
to 11 by eliminating Cu 2+ produced a more robust QSAR as explained previously
(Table 5.13).
5.2.4 E lEctronEgativity
Electronegativity is a chemical property that describes the power of an atom in a
molecule to attract electrons to itself (Pauling 1932). Eleven studies used electro-
negativity (X, X AR , or X m ) to predict cation toxicity ( Table 5.14 ). The studies of
Seifriz (1949), Danielli and Davies (1951), and Somers (1959) were discussed in
Walker et al. (2003). The Biesinger and Christensen (1972) test system is discussed
in Section 5.2.2. The following QSAR was extrapolated from Figure 1 of Biesinger
and Christensen (1972): pM = - 3 + 5 X AR , r 2 = 0.61, p < 0.001, n = 20. The test system
and cations used to develop the Biesinger and Christensen (1972) QSAR are listed
in Table 5.14.
Turner et al. (1985) and Khangarot and Ray (1989) did not report QSARs for
electronegativity. However, Khangarot and Ray did use linear regression analysis to
describe the relationship between 23 cation toxicities and standard 2-day Daphnia
magna EC 50 values based on immobilization: r 2 = 0.542 and p < 0.001. The Turner
et al. (1985) and Khangarot and Ray (1989) test systems and cations are listed in
Table 5.14.
Enache et al. (2000) developed 2 electronegativity QSARs. The test system for
both QSARs is listed in Table 5.14. Both QSARs relied on Allred-Rochow electro-
negativity (X AR ), standard reduction-oxidation potential (ΔE 0 ), and Pauling's elec-
tronegativity (X) ( Table 5.11 ). The first QSAR is for 12 cations listed in Table 5.11
and the second is for 11 cations listed in Table 5.11. The statistical analysis for the
12 cations produced a reliable QSAR, but improved results were obtained when the
copper ion was omitted from the analysis as explained in Section 5.2.3.
Enache et al. (2003) developed an electronegativity-based QSAR to predict the
toxicity of 11 cations (Table 5.11). The test system for the QSAR is listed in Table 5.14.
Workentine et al. (2008) used the test system listed in Table 5.14 and discussed
in Section 5.2.2. Linear regression analyses were conducted to determine the rela-
tionships between the measured toxicity values and X m : r 2 = 0.28, p = 0.006 (MIC 24 ),
r 2 = 0.04, p = 0.397 (MBC 100 ) and r 2 = 0.01 p = 0.725 (MBEC 100 ).