LipidOracle: Ambiguity-aware lipid annotation across fragmentation methods
Abstract
Tandem mass spectra (MS²) rarely determine a complete lipid structure, yet annotation software reports a best-scoring name and leaves users to judge whether the spectrum distinguishes it from other isomers. LipidOracle enumerates every isomer compatible with a precursor, converts the scores into probabilities with a softmax calibrated on standards, and reports the structural detail on which the credible candidates agree. An independent audit tests whether any observed fragment separates the reported structure from its closest rival. Annotation depth, from sum composition to double-bond position, is thus set by the evidence in each spectrum. The same inference handled collision-, electron- and ultraviolet photodissociation spectra, with only the fragment model exchanged. On standards, it located double bonds among thousands of candidate isomers with small credible sets, and in human plasma and mouse liver it named chains and double-bond positions only where the spectra supported them. The approach extends to any activation method whose fragments can be predicted.
Related
- LLM-based cell type annotation harmonization across single-cell studies using GCTHarmony.
- Annotation-aware multidomain validation of NucS and MutS-MutL across 23,435 prokaryotic reference genomes
- GPCR-PTM: an interpretable, topology-aware rule-based framework for residue-level post-translational modification annotation and prediction across the human GPCRome
- Meta-analysis reveals microbiome signatures for colorectal cancer that are universal across age groups and sequencing methods
- Convergence is not correctness: context-dependent performance of enhanced-sampling methods across biological complexity
BibTeX
@misc{wu2026lipidoracle,
title = {{LipidOracle: Ambiguity-aware lipid annotation across fragmentation methods}},
author = {Vincen Wu and Abraham Moyal and Adriano Rutz and Nicola Zamboni},
howpublished = {ChemRxiv},
year = {2026},
month = oct,
doi = {10.26434/chemrxiv.15009711/v1},
url = {https://chemrxiv.org/doi/full/10.26434/chemrxiv.15009711/v1},
note = {CC BY},
}
RIS
TY - UNPB
TI - LipidOracle: Ambiguity-aware lipid annotation across fragmentation methods
AU - Vincen Wu
AU - Abraham Moyal
AU - Adriano Rutz
AU - Nicola Zamboni
PY - 2026
DA - 2026/10/01
PB - ChemRxiv
DO - 10.26434/chemrxiv.15009711/v1
UR - https://chemrxiv.org/doi/full/10.26434/chemrxiv.15009711/v1
AB - Tandem mass spectra (MS²) rarely determine a complete lipid structure, yet annotation software reports a best-scoring name and leaves users to judge whether the spectrum distinguishes it from other isomers. LipidOracle enumerates every isomer compatible with a precursor, converts the scores into probabilities with a softmax calibrated on standards, and reports the structural detail on which the credible candidates agree. An independent audit tests whether any observed fragment separates the reported structure from its closest rival. Annotation depth, from sum composition to double-bond position, is thus set by the evidence in each spectrum. The same inference handled collision-, electron- and ultraviolet photodissociation spectra, with only the fragment model exchanged. On standards, it located double bonds among thousands of candidate isomers with small credible sets, and in human plasma and mouse liver it named chains and double-bond positions only where the spectra supported them. The approach extends to any activation method whose fragments can be predicted.
ER -