405
12 Consensus Drug Design Using IT Microcosm
The substances were studied in vitro on a model of ascorbate-induced lipid peroxidation (LP) in the homogenate of rat liver [53]. The indices of the antioxidant
activity were as follows: EC 50 –compound concentration (mol/L) inhibiting LP by
50 %; Δ (10
−6
)—percentage of LP inhibition at a concentration 1·10
−6
M of the studied substance; Ind 10 —the module of the exponent of the measure of the substance
concentration inducing LP inhibition by 10 % (using the value of 2 to 7 points). Trolox C (CAS 53188-07-1) was studied as the comparison drug, and it yielded EC 50
values of 2.76·10
−6
M, Δ (10
−6
) = 36.9 %, and Ind 10 = 7.
Based on the results of cluster analysis in combination with an expert assessment, the following activity classes were distinguished:
• high—EC 50 < 5.0 · 10
−6
M or Δ (10
−6
) > 20.0 % (78 compounds), or Ind 10 = 6, 7 (86
compounds);
• moderate—5.0 · 10
−6
≤ EC 50 < 1.0 · 10
−4
M or 10.0 < Δ (10
−6
) ≤ 20.0 % (69 compounds), or Ind 10 = 5 (59 compounds);
• high or moderate—EC 50 < 1.0 · 10
−4
M or Δ (10
−6
) > 10.0 % (147 compounds), or
Ind 10 = 5, 6, 7 (145 compounds); and
• low—EC 50 ≥ 1.0 · 10
−4
M or Δ (10
−6
) ≤ 10.0 % (178 compounds), or Ind 10 = 2, 3, 4
(165 compounds).
The results of computational accuracy testing for the prediction of the extent of the
antioxidant activity of the condensed azole derivatives are shown in Table 12.12.
All of the obtained decision rules proved inadequate for the moderate activity;
therefore, they were excluded from further prediction.
Substances with a high antioxidant activity were sought among 721 novel, untested condensed azole derivatives of 15 chemical classes (see Sect. 3.2) using several consensus approaches to the selection of the most promising compounds, including a method of using three strategies in combination, and testing the spectrum
of the predicted estimates for noncontradiction.
The key criteria for selecting experimental-study candidates were as follows
(“A” stands for a positive prediction estimate of a compound activity as “high”):
• 2A or 3A in all six decision rules, with a conformity coefficient of the predictionestimate spectrum (21) K High > 0.5;
• 3A for the index EC 50 and 3A for Ind 10 simultaneously;
• 3A for EC 50 or 3A for Ind 10 , while K High = 1.0; and
• 3A for EC 50 and K High > 0.8.
Altogether, 41 substances were selected according to these criteria and studied experimentally.
Twenty-eight compounds (68.3 %) with an expressed antioxidant activity Δ
(10
−6
) > 10.0 % were found; 17 of them (41.5 %) were compounds showing a high
activity Δ (10
−6
) > 20.0 %. Of 17 highly active compounds, 13 compounds (76.5 %)
are comparable to the reference drug Trolox C in antioxidant activity, and four compounds (23.5 %) exceed its activity.
For the Δ (10
−6
) index, the training set contains 45.2 % compounds with an expressed
activity, including 24.0 % compounds with a high activity. The primary screening did
not employ in silico methods, so the percentage of the obtained active substances can be
regarded as an indication of the accuracy of the intuitive human prediction performed
12 Consensus Drug Design Using IT Microcosm
The substances were studied in vitro on a model of ascorbate-induced lipid peroxidation (LP) in the homogenate of rat liver [53]. The indices of the antioxidant
activity were as follows: EC 50 –compound concentration (mol/L) inhibiting LP by
50 %; Δ (10
−6
)—percentage of LP inhibition at a concentration 1·10
−6
M of the studied substance; Ind 10 —the module of the exponent of the measure of the substance
concentration inducing LP inhibition by 10 % (using the value of 2 to 7 points). Trolox C (CAS 53188-07-1) was studied as the comparison drug, and it yielded EC 50
values of 2.76·10
−6
M, Δ (10
−6
) = 36.9 %, and Ind 10 = 7.
Based on the results of cluster analysis in combination with an expert assessment, the following activity classes were distinguished:
• high—EC 50 < 5.0 · 10
−6
M or Δ (10
−6
) > 20.0 % (78 compounds), or Ind 10 = 6, 7 (86
compounds);
• moderate—5.0 · 10
−6
≤ EC 50 < 1.0 · 10
−4
M or 10.0 < Δ (10
−6
) ≤ 20.0 % (69 compounds), or Ind 10 = 5 (59 compounds);
• high or moderate—EC 50 < 1.0 · 10
−4
M or Δ (10
−6
) > 10.0 % (147 compounds), or
Ind 10 = 5, 6, 7 (145 compounds); and
• low—EC 50 ≥ 1.0 · 10
−4
M or Δ (10
−6
) ≤ 10.0 % (178 compounds), or Ind 10 = 2, 3, 4
(165 compounds).
The results of computational accuracy testing for the prediction of the extent of the
antioxidant activity of the condensed azole derivatives are shown in Table 12.12.
All of the obtained decision rules proved inadequate for the moderate activity;
therefore, they were excluded from further prediction.
Substances with a high antioxidant activity were sought among 721 novel, untested condensed azole derivatives of 15 chemical classes (see Sect. 3.2) using several consensus approaches to the selection of the most promising compounds, including a method of using three strategies in combination, and testing the spectrum
of the predicted estimates for noncontradiction.
The key criteria for selecting experimental-study candidates were as follows
(“A” stands for a positive prediction estimate of a compound activity as “high”):
• 2A or 3A in all six decision rules, with a conformity coefficient of the predictionestimate spectrum (21) K High > 0.5;
• 3A for the index EC 50 and 3A for Ind 10 simultaneously;
• 3A for EC 50 or 3A for Ind 10 , while K High = 1.0; and
• 3A for EC 50 and K High > 0.8.
Altogether, 41 substances were selected according to these criteria and studied experimentally.
Twenty-eight compounds (68.3 %) with an expressed antioxidant activity Δ
(10
−6
) > 10.0 % were found; 17 of them (41.5 %) were compounds showing a high
activity Δ (10
−6
) > 20.0 %. Of 17 highly active compounds, 13 compounds (76.5 %)
are comparable to the reference drug Trolox C in antioxidant activity, and four compounds (23.5 %) exceed its activity.
For the Δ (10
−6
) index, the training set contains 45.2 % compounds with an expressed
activity, including 24.0 % compounds with a high activity. The primary screening did
not employ in silico methods, so the percentage of the obtained active substances can be
regarded as an indication of the accuracy of the intuitive human prediction performed