This document summarizes research on using machine learning approaches to predict the binding pose of protein-ligand complexes for drug discovery applications. Specifically, it describes building statistical scoring functions (SFs) using features extracted from existing SFs and training machine learning models like random forests and boosted regression trees to directly predict the root-mean-square deviation (RMSD) between predicted and native poses. Evaluation on several benchmark datasets shows the machine-learning SFs outperform conventional empirical and knowledge-based SFs in identifying near-native poses within 1-3 angstroms of the crystal structure.