BioSM

BioSM identifies unknown endogenous mammalian biochemical structures in LC/MS-derived chemical structure spaces by matching compounds to a library of known mammalian biochemical scaffolds using graph matching methods.

Key Features:

• Scaffold-Based Identification: Uses a curated set of known mammalian biochemical compounds as scaffolds to reference and identify structurally related compounds.
• Graph Matching Methods: Applies graph matching techniques to compare structural similarity between unknown compounds and the scaffold library.
• High Accuracy and Robustness: Achieved 95% correct prediction for identifying Kyoto Encyclopedia of Genes and Genomes (KEGG) compounds as endogenous mammalian biochemicals in a leave-one-out cross-validation using an initial set of 1565 scaffolds.
• Comprehensive Data Set Analysis: Evaluated against Human Metabolome Database (HMDB) metabolites, KEGG plant secondary metabolites, DrugBank and USAN drugs, Chembridge and Chemsynthesis synthetic chemicals, and PubChem, yielding annotations for 89% of HMDB compounds, 72% of plant metabolites, 48% of drug structures, 29% of synthetic chemicals, and 34% of randomly selected PubChem compounds.
• Scalability and Model Robustness: Assessed an expanded scaffold list of 3927 biochemical compounds, which improved sensitivity and specificity while producing overall comparable dataset annotation results.
• Qualitative and Quantitative Annotation: Provides binary (yes/no) annotation and quantitative ranking of candidate endogenous mammalian biochemicals within chemical spaces.

Scientific Applications:

• Metabolite identification: Annotates and ranks candidate biochemical structures from complex biofluids analyzed by LC/MS to support metabolomics studies.
• Metabolic pathway elucidation and biomarker discovery: Aids elucidation of metabolic pathways and discovery of biomarkers by identifying endogenous mammalian biochemicals.
• Disease mechanism investigation: Facilitates investigation of disease mechanisms through comprehensive annotation of biochemical structure space.

Methodology:

Operates on LC/MS-derived chemical structures using a curated library of endogenous mammalian biochemical scaffolds, graph matching algorithms for structural comparison, leave-one-out cross-validation for performance assessment, and evaluation with expanded scaffold sets; offers both binary annotation and quantitative ranking.