ZODIAC

ZODIAC reranks and refines SIRIUS-generated molecular formula candidates from mass spectrometry-based metabolomics data to enable de novo estimation of molecular formulas for large (>500 Da) and novel compounds not present in structure databases.

Key Features:

• Network-based reranking: Implements a network-based algorithm that reranks molecular formula candidates generated by SIRIUS.
• Fragmentation tree computation: Performs fragmentation tree computation to support formula evaluation.
• Bayesian inference with Gibbs sampling: Applies Bayesian statistics implemented via Gibbs sampling to refine formula probabilities.
• Integration with SIRIUS: Enhances SIRIUS molecular formula outputs by re-evaluating and reranking candidate formulas.
• Focus on large and novel formulas: Targets de novo estimation for compounds exceeding 500 Daltons and molecular formulas absent from existing databases.
• Computational efficiency: Provides swift and efficient processing that reduces incorrect annotations.
• Performance improvement: Achieved a 16.2-fold decrease in annotation errors on plant extract datasets with compounds predominantly above 700 Daltons.
• Cross-dataset robustness: Demonstrated consistent annotation improvements across four additional diverse datasets.
• Novel formula discovery: Enabled identification of novel molecular formulas such as C24H47BrNO8P not recorded in public molecular structure databases.

Scientific Applications:

• De novo molecular formula estimation: Estimating molecular formulas for large and novel compounds in mass spectrometry-based metabolomics.
• Molecular formula candidate prioritization: Improving the ranking of formula candidates produced by SIRIUS to support downstream structure elucidation.
• Annotation error reduction: Reducing incorrect annotations in complex datasets, including plant extracts with high-mass compounds.
• Discovery of novel chemistry: Identifying molecular formulas absent from public molecular structure databases to expand known chemical space.

Methodology:

Performs fragmentation tree computation and applies a network-based reranking algorithm combined with Bayesian statistics using Gibbs sampling to refine SIRIUS-generated molecular formula candidates.