Decision trees versus support vector machine for classification of androgen receptor ligands

Area of Science:

• Computational toxicology
• Environmental chemistry
• Endocrinology

Background:

• Increasing need for in silico models to predict chemical toxicity due to limitations in experimental assays.
• Endocrine disruptors are environmental and industrial chemicals that can interfere with natural hormone functions.
• Structure-activity relationship (SAR) models are crucial for predicting endocrine disruption potential.

Purpose of the Study:

• To develop and evaluate in silico models for categorizing the relative binding affinity (RBA) of diverse chemicals to the androgen receptor.
• To identify key chemical descriptors influencing androgen receptor binding.
• To compare the performance of recursive partitioning trees with support vector machines for this classification task.

Main Methods:

• Utilized a dataset of approximately 200 chemicals across various structural classes.
• Employed recursive partitioning trees for classification based on descriptors from CODESSA software.
• Calculated topological, geometrical, and quantum chemical properties as descriptors, including hydrophobicity (log P) and charge distribution parameters.

Main Results:

• Successfully classified chemicals into four activity groups based on log RBA values.
• Hydrophobicity, Balaban index, and charge distribution descriptors were identified as key factors in chemical partitioning.
• Achieved straightforward separation for strongly active and approximately 90% of inactive compounds.
• More complex trees were required for differentiating moderate and weak binders.

Conclusions:

• In silico models, particularly recursive partitioning trees, demonstrate significant potential for predicting androgen receptor binding affinity.
• Key chemical descriptors related to hydrophobicity and charge distribution are critical for accurate toxicity predictions.
• The developed models offer a viable alternative to experimental assays for screening endocrine-disrupting chemicals.