Scientific Publications by FDA Staff

Environ Toxicol Chem 2014 Jun;33(6):1271-82

Partial least square and k-nearest neighbor algorithms for improved 3D quantitative spectral data-activity relationship consensus modeling of acute toxicity.

Stoyanova-Slavova IB, Slavov SH, Pearce B, Buzatu DA, Beger RD, Wilkes JG

A diverse set of 154 chemicals that included FDA-regulated compounds tested for their aquatic toxicity in Daphnia magna were modeled by a three-dimensional quantitative spectral data-activity relationship (3D-QSDAR). Two distinct algorithms, namely partial least squares (PLS) and Tanimoto similarity based k-nearest neighbors (KNN) were used to process bin occupancy descriptor matrices obtained after tessellation of the 3D-QSDAR space into regularly sized bins. The performance of models utilizing bins ranging in size from 2 ppm x 2 ppm x 0.5 A to 20 ppm x 20 ppm x 2.5 A was explored. Rigorous quality control (QC) criteria were imposed: i) one hundred randomized 20% hold-out test sets were generated and the average R2 test of the respective models was used as a measure of their performance and ii) a Y-scrambling procedure was used to identify chance correlations. A consensus between the best performing composite PLS model using 0.5 A x 14 ppm x 14 ppm bins and 10 latent variables (average R2 test = 0.770) and the best composite KNN model using 0.5 A x 8 ppm x 8 ppm and 2 neighbors (average R2 test = 0.801) offered an improvement of about 7.5% (R2 test consensus = 0.845). Projection of the most frequently occurring bins on the standard coordinate space indicated that the presence of primary or secondary amino group - substituted aromatic systems - would result in an increased toxic effect in Daphnia. The presence of a second aromatic ring with highly electronegative substituents 5 to 7 A apart from the first ring would lead to a further increase in toxicity.