GraphMZ-benchmark

GraphMZ SRFA cross-tool benchmark Paper-supporting reproducibility dataset for GraphMZ — a Neo4j graph-database application written in Python for high-resolution mass-spectrometry analysis (organic-contaminant detection and natural-organic-matter characterisation).This record provides a frozen snapshot of the SRFA cross-tool consensus analysis used to validate GraphMZ's formula-assignment performance. The companion GraphMZ-demo record (concept DOI: `10.5281/zenodo.20351944`) holds tutorial-only inputs and uses an always-latest concept DOI.All raw acquisitions are 21T FT-ICR negative-ESI on Suwannee River Fulvic Acid (SRFA — IHSS reference NOM), at the National High Magnetic Field Laboratory (MagLab, Tallahassee FL, USA). Cross-tool formula assignments were produced externally by CoreMS and MFAssignR. Files are uploaded flat; the filename prefix identifies the logical group. Cross-tool formula assignments — CoreMS, MFAssignR, Hawkes Per-tool, per-acquisition formula tables that feed the two-tier consensus pipeline. CoreMS and MFAssignR each produced four assignment tables: one for SRFA_1 and three replicates for SRFA_2. Filename Source CoreMS_MagLab_NegESI_SRFA_1.csv CoreMS, SRFA_1 acquisition CoreMS_MagLab_NegESI_SRFA_2_rep1.csv CoreMS, SRFA_2 replicate 1 CoreMS_MagLab_NegESI_SRFA_2_rep2.csv CoreMS, SRFA_2 replicate 2 CoreMS_MagLab_NegESI_SRFA_2_rep3.csv CoreMS, SRFA_2 replicate 3 MFAssignR_MagLab_NegESI_SRFA_1.csv MFAssignR, SRFA_1 acquisition MFAssignR_MagLab_NegESI_SRFA_2_rep1.csv MFAssignR, SRFA_2 replicate 1 MFAssignR_MagLab_NegESI_SRFA_2_rep2.csv MFAssignR, SRFA_2 replicate 2 MFAssignR_MagLab_NegESI_SRFA_2_rep3.csv MFAssignR, SRFA_2 replicate 3 Hawkes_common_SRFA_neg.csv Hawkes 2020 inter-lab SRFA consensus (1,125 formulas) — historical recall benchmark Raw `.pks` acquisitions Upstream of the formula-assignment tables — the original peak-list exports that CoreMS and MFAssignR ran on. Filename Sample MagLab_NegESI_SRFA_1.pks SRFA_1 (byte-identical to the file of the same name in the demo record, where it's used by NOM_demo1) MagLab_NegESI_SRFA_2_rep1.pks SRFA_2 replicate 1 MagLab_NegESI_SRFA_2_rep2.pks SRFA_2 replicate 2 MagLab_NegESI_SRFA_2_rep3.pks SRFA_2 replicate 3 Consensus ground truth — pipeline output Frozen output of the two-tier consensus pipeline (`graphmz.validation.consensus_ground_truth`), applied to the formula-assignment tables above. Shipped as a snapshot so paper reviewers can verify GraphMZ's claims against the canonical reference without re-running the pipeline. Filename Content Consensus_GT.csv Headline ground truth: Tier-1 ∩ Tier-2 Tier1_GT.csv SRFA_2 triplicate-per-tool ≥2-of-3 intersect, then cross-tool intersect Tier2_GT_SRFA1.csv Tier-2: cross-tool intersect on SRFA_1 Tier1_classification.csv Per-formula classification trace from Tier-1 Tier2_classification.csv Per-formula classification trace from Tier-2 hawkes_recall.csv Recall of the Hawkes 2020 formula set against each tier and the consensus subset Methods documentation Filename Content coverage_gap_analysis.md Coverage-gap analysis of GraphMZ vs. the consensus ground truth cross_tool_spec.md 16-column unified schema spec for the cross-tool ingestion How to fetch programmatically The files above are exposed via the GraphMZ data loader (`graphmz.data.load_benchmark`). The loader downloads from this record on first call, caches locally, and reconstructs the project's expected layout from filename prefixes (e.g. `CoreMS_*` / `MFAssignR_*` / `Hawkes_*` under `SRFA_benchmark/formula_assignments/`, `*.pks` under `SRFA_benchmark/pks_lists/`, `*_GT.csv` / `*_classification.csv` / `hawkes_recall.csv` under `SRFA_benchmark/consensus_gt/`). SHA256 checksums are verified against values baked into the GraphMZ package. See the GraphMZ project README and `benchmark_consensus_gt.ipynb` (or `benchmark_nom_ground_truth.ipynb`) for the analysis walkthrough.