Quantitative structure-activity relationships for 2-[(phenylmethyl)sulfonyl]pyridine 1-oxide herbicides

Abstract

Phenyl-substituted analogues of 2-[(phenylmethyl)sulfonyl]pyridine 1-oxide preemergent herbicides were examined in order to determine quantitative relationships between structure and activity against the following three weed species: switch grass (Panicum virgatum L.), barnyard grass (Echinochloa crusgalli L. Beauv.), and green foxtail (Setaria viridis L. Beauv.). Analogues were chosen to provide maximum parameter orthogonality. Regression analysis yielded structure-activity relationships wherein the most significant substituent parameters associated with herbicidal activity were found to be the partition coefficient (pi), the molar refractivity (MR), and two indicator variables, Z (denoting the presence of an alpha-methyl group) and H (denoting an ortho substituent capable of hydrogen bonding). For green foxtail, the structure-activity relationship was found to be: -log ED50 = 0.43 pi -0.052MR + 0.50H + 0.24Z + 0.61, where ED50 is expressed in moles per acre. The regression equations were found to explain 79-93% of the bioactivity for the three weed species studied. It was further shown that these equations represent the best possible correlations within the limitations of the biological data.

13C NMR–Distance Matrix Descriptors: Optimal Abstract 3D Space Granularity for Predicting Estrogen Binding

“…Hence, it would be desirable to set an upper limit for R 2 on the basis of the standard deviation (σ) of the original measurements. Applying the methodology described by Doweyko et al, 43 100 simulated models using data with σ equal to that of the EDKB data (IC 50 σ = 2.73 × 10 −5 ) were generated. The average R 2 was 0.89 (ranging from 0.55 to 0.98) with a standard deviation of 0.073.…”

Complementary PLS and KNN algorithms for improved 3D-QSDAR consensus modeling of AhR binding

“…In our earlier work [8] an automated partial least squares (PLS) algorithm was used to process data from regularly tessellated 3D-SDAR fingerprints and to derive averaged (composite model) predictions from 100 randomized training/hold-out test set pairs. A technique [10] based on the standard deviation of the experimental data was employed to determine a “realistic” upper bound for coefficient of determination. A Y-scrambling procedure [11,12] assessed the probability of generating seemingly “good” models by chance.…”

“…Under the assumption that for similar compounds and experimental settings the average relative experimental error would vary insignificantly, based on the above report at least ~17% error in the EC 50 data should be expected. However, since Mekenyan et al [22] compiled their dataset from various sources, a negative impact on the accuracy of data which would further lower the “realistic” R 2 [10] should be anticipated.…”

Alignment-independent technique for 3D QSAR analysis

“…2b R Test 2 RMSD Consensus R Test 2 (a) (b) (c) (d) Global minimum energy (Model 1) 4 LV; 16 ppm; 1.0 Å 0.92 0.06 0.60 0.77 0.62 +3.3 % 0.65 +10 % Global minimum energy (Model 2) 3 LV; 8 ppm; 1.0 Å 0.92 0.07 0.60 0.77 0.58 −0.9 % 0.64 +15 % Alignment, 50:50 electronic:steric 2 LV; 6 ppm; 1.0 Å 0.85 0.06 0.57 0.80 Alignment, best-of-each 2 LV; 6 ppm; 1.0 Å 0.84 0.06 0.56 0.80 2D > 3D conversion 3 LVs; 8 ppm; 1.5 Å 0.91 0.05 0.61 0.75 Parameters were conformation basis, number of Latent Variables (LVs) in the PLS model, 3D-SDAR fingerprint granularity (chemical shifts in ppm; interatomic distances in Å) including predictive accuracy (R Test 2 ) based on consensus predictions from composite models of differing granularity and conformation basis. (All R 2 values in non -bold fonts are from composites based on averages from 100 random training/test set partitions) a Replicate data for a similar estrogen receptor binding bioassay from the same data base, allowed calculation of an upper bound for modeling accuracy by the method of Doweyko et al [ 30 ]. The calculation yie...…”