The thesis "Relating Protein Pharmacology by Ligand Chemistry" by Michael James Keiser explores a novel approach to understanding protein function based on the chemical similarity of ligands. The study begins with a collection of 65,000 ligands annotated for hundreds of drug targets. Using ligand topology, similarity scores between each set of ligands are calculated, and a statistical model is developed to rank the significance of these scores. The resulting similarity scores are expressed as networks, which reveal biologically meaningful clusters of protein targets despite being connected solely by chemical similarity. The "Similarity Ensemble Approach" (SEA) is used to compare drugs to target sets, leading to unexpected links. For example, methadone, emetine, and loperamide were found to bind to muscarinic M3, α2 adrenergic, and neurokinin NK2 receptors, respectively. These predictions were experimentally confirmed, highlighting the potential of SEA in predicting drug promiscuity and side effects. The thesis also explores the mapping of drug space from the perspective of small molecule metabolism, revealing novel territory for metabolic drug discovery. The chemical similarity between drugs and metabolites can suggest drug toxicity, routes of metabolism, and polypharmacology. Overall, the thesis demonstrates that the SEA approach is systematic and comprehensive, offering insights into side effects and new indications for many drugs. The work combines the strengths of both sequence and structure-based approaches, leveraging the extensive data on known drug-like molecules to identify and formalize patterns among ligands that reflect functional relationships among proteins.The thesis "Relating Protein Pharmacology by Ligand Chemistry" by Michael James Keiser explores a novel approach to understanding protein function based on the chemical similarity of ligands. The study begins with a collection of 65,000 ligands annotated for hundreds of drug targets. Using ligand topology, similarity scores between each set of ligands are calculated, and a statistical model is developed to rank the significance of these scores. The resulting similarity scores are expressed as networks, which reveal biologically meaningful clusters of protein targets despite being connected solely by chemical similarity. The "Similarity Ensemble Approach" (SEA) is used to compare drugs to target sets, leading to unexpected links. For example, methadone, emetine, and loperamide were found to bind to muscarinic M3, α2 adrenergic, and neurokinin NK2 receptors, respectively. These predictions were experimentally confirmed, highlighting the potential of SEA in predicting drug promiscuity and side effects. The thesis also explores the mapping of drug space from the perspective of small molecule metabolism, revealing novel territory for metabolic drug discovery. The chemical similarity between drugs and metabolites can suggest drug toxicity, routes of metabolism, and polypharmacology. Overall, the thesis demonstrates that the SEA approach is systematic and comprehensive, offering insights into side effects and new indications for many drugs. The work combines the strengths of both sequence and structure-based approaches, leveraging the extensive data on known drug-like molecules to identify and formalize patterns among ligands that reflect functional relationships among proteins.