Topological Features in Profiling the Antimalarial Activity Landscape of Anilinoquinolines: A Multipronged QSAR Study

Table 6

Architecture and goodness of fit of BP-ANN models for antimalarial activity of anilinoquinolines (Table 2) from selected feature sets of CP-MLR, GA, and their common five descriptors.

BP-ANN architecture and parameters
Layer			Nodes			Training parameters
Input			5 + 1 (bias)			Learning rate (𝜇)						0.57–0.66
Hidden			6			Momentum (𝛼)						0.55–0.77
Output			1			Transfer function						Sigmoid
						Optimization algorithm						Levenberg-Marquardt
						Iterations (𝜆)						17–40
Model statistics
Feature set^a	𝜇	𝛼	𝑟^2b			RMSEP			RSEP (%)			MAE (%)
			Train	Valid	Test	Train	Valid	Test	Train	Valid	Test	Train	Valid	Test
CP-MLR	0.601	0.558	0.813	0.814	0.818	0.289	0.256	0.267	3.914	3.404	3.564	6.457	9.859	9.879
GA	0.577	0.657	0.883	0.886	0.878	0.229	0.200	0.219	3.099	2.662	2.925	6.019	8.349	9.287
Common	0.66	0.77	0.823	0.810	0.757	0.284	0.279	0.333	3.852	3.710	4.448	6.819	10.832	11.810

^aThe ANN input features of CP-MLR and GA sets, respectively, correspond to (1) and (6) given in Table 3. The ANN input features common sets are features common to CP-MLR and GA.
^b𝑟²: squared correlation coefficient; RMSEP: root-mean-square error of prediction; RSEP: relative standard error of prediction; MAE: mean absolute error.