Task

MassSpecGym Molecule retrieval

This task ranks candidate molecular structures for an observed MS/MS spectrum.

Sources (2)

pluskal-lab/MassSpecGym official source; pluskal-lab/MassSpecGym massspecgym/models/retrieval/base.py · Pinned README: three challenges; dataset and DataModule; evaluation base classes; pinned massspecgym/models/retrieval/base.py evaluation methods

10 evaluations · 40 metric rows

At a glance

Inputs, training, access and other details

Explanatory profile: source reviewed · Automated source review, 2026-09-16. Review applies to the cited claims; unresolved fields are listed below. Numerical results retain their own review status.

Data, procedure and scoring
Property	Description and evidence
Datasets	MS/MS spectrum plus a candidate set; rank the matching molecule. Sources (2) pluskal-lab/MassSpecGym official source; pluskal-lab/MassSpecGym massspecgym/models/retrieval/base.py · Pinned README: three challenges; dataset and DataModule; evaluation base classes; pinned massspecgym/models/retrieval/base.py evaluation methods
Splits	MCES molecular clusters are grouped into fixed training, validation and test folds, stratified by acquisition metadata. Cross-fold molecular bond-edit distance is at least 10; all spectra follow the assigned molecular fold. Sources massspecgym primary benchmark evidence · Section 3.4; Supplementary Information 2.5
Metrics	Mean hit rate at configured top-k cutoffs, with optional MCES distance for the top-ranked candidate. Sources (2) pluskal-lab/MassSpecGym official source; pluskal-lab/MassSpecGym massspecgym/models/retrieval/base.py · Pinned README: three challenges; dataset and DataModule; evaluation base classes; pinned massspecgym/models/retrieval/base.py evaluation methods
Baselines	The README illustrates a DeepSets-style spectrum-to-fingerprint retrieval baseline. Sources (2) pluskal-lab/MassSpecGym official source; pluskal-lab/MassSpecGym massspecgym/models/retrieval/base.py · Pinned README: three challenges; dataset and DataModule; evaluation base classes; pinned massspecgym/models/retrieval/base.py evaluation methods
Leakage controls	Maximum common edge subgraph (MCES) clustering keeps molecules connected by a bond-edit distance below 10 in the same fold. The split additionally balances instrument, collision-energy, adduct and molecule-frequency metadata; this is stronger than simply separating 2D InChIKeys. Sources massspecgym primary benchmark evidence · Section 3.4; Supplementary Information 2.5; Tables 2–4
Uncertainty	Tables 2–4 report 99.9% bootstrap confidence intervals using 20,000 resamples. These intervals summarize test-example sampling, not variation across independently retrained models. Sources massspecgym primary benchmark evidence · Section 3.4; Supplementary Information 2.5; Tables 2–4
Entity type	Constituent benchmark task: MassSpecGym Molecule retrieval Sources (2) pluskal-lab/MassSpecGym official source; pluskal-lab/MassSpecGym massspecgym/models/retrieval/base.py · Pinned README: three challenges; dataset and DataModule; evaluation base classes; pinned massspecgym/models/retrieval/base.py evaluation methods
Organisms	Molecule identity rather than organism classification defines these tasks. · Not applicable Sources (2) pluskal-lab/MassSpecGym official source; pluskal-lab/MassSpecGym massspecgym/models/retrieval/base.py · Pinned README: three challenges; dataset and DataModule; evaluation base classes; pinned massspecgym/models/retrieval/base.py evaluation methods
Assays	Tandem mass spectra paired with molecular structures. Sources (2) pluskal-lab/MassSpecGym official source; pluskal-lab/MassSpecGym massspecgym/models/retrieval/base.py · Pinned README: three challenges; dataset and DataModule; evaluation base classes; pinned massspecgym/models/retrieval/base.py evaluation methods
Allowed inputs	MS/MS spectrum plus the supplied molecule candidate set. Sources (2) pluskal-lab/MassSpecGym official source; pluskal-lab/MassSpecGym massspecgym/models/retrieval/base.py · Pinned README: three challenges; dataset and DataModule; evaluation base classes; pinned massspecgym/models/retrieval/base.py evaluation methods
Adaptation	Supervised train/validation/test learning; pretrained or new models use the task-specific interfaces. Sources (2) pluskal-lab/MassSpecGym official source; pluskal-lab/MassSpecGym massspecgym/models/retrieval/base.py · Pinned README: three challenges; dataset and DataModule; evaluation base classes; pinned massspecgym/models/retrieval/base.py evaluation methods

How it works

How it worksEvaluation procedure

Conceptual procedure. Task variants and protocol versions retain their separate scoring conditions.

Sources (3)

pluskal-lab/MassSpecGym official source; pluskal-lab/MassSpecGym massspecgym/models/retrieval/base.py; massspecgym primary benchmark evidence · Pinned README: three challenges; dataset and DataModule; evaluation base classes; pinned massspecgym/models/retrieval/base.py evaluation methods; Section 3.4; Supplementary Information 2.5

Evaluation methodology

MS/MS spectrum plus a candidate set; rank the matching molecule. MCES molecular clusters are grouped into fixed training, validation and test folds, stratified by acquisition metadata. Cross-fold molecular bond-edit distance is at least 10; all spectra follow the assigned molecular fold. Mean hit rate at configured top-k cutoffs, with optional MCES distance for the top-ranked candidate. Maximum common edge subgraph (MCES) clustering keeps molecules connected by a bond-edit distance below 10 in the same fold. The split additionally balances instrument, collision-energy, adduct and molecule-frequency metadata; this is stronger than simply separating 2D InChIKeys.

Sources (3)

Evaluation design

Benchmarks bring together tasks and protocols. A task describes the biological question; a protocol defines a particular test.

Benchmarks

MassSpecGym

Protocols

These source-backed links do not make different protocols or scores interchangeable.

Published comparisons

Explore the results reported under one evaluation protocol. Each figure keeps its source, dataset and metric together; it is not a ranking across studies.

Comparison and metric

MassSpecGym · main · Table 3

Hit rate @1 (percent) · Higher values are better for this metric.

Retrieve the correct molecule from up to 256 candidates using an input spectrum; compare separately within the main challenge and molecular-formula challenge. MCES single-linkage molecular clustering at threshold 10; fixed held-out test split

Evaluation protocol · MassSpecGym · main

Random (main) · Configuration · Author-reported evaluation0.37 (0.24-0.54)
DeepSets (main) · Configuration · Author-reported evaluation1.47 (1.18-1.77)
Fingerprint FFN (main) · Configuration · Author-reported evaluation2.54 (2.17-2.99)
DeepSets + Fourier features (main) · Configuration · Author-reported evaluation5.24 (4.71-5.83)
MIST (main) · Configuration · Author-reported evaluation14.64 (13.82-15.54)

Source order is preserved. Plotted marks show point estimates; uncertainty, where reported, is retained in the printed values and table. Differences do not establish statistical significance.

MassSpecGym: A benchmark for the discovery and identification of molecules · Table 3: Hit rate @1, main

Values, uncertainty and evidence

Hit rate @1: original source values
Tested entity	Printed value	Uncertainty	Evidence
Random (main) · Configuration	0.37 (0.24-0.54) percent	type: bootstrap confidence interval; level: 0.999; resamples: 20000; lower: 0.24; upper: 0.54	Author-reported evaluation · source checkedMassSpecGym: A benchmark for the discovery and identification of molecules · Table 3, main challenge, Random, Hit rate @1
DeepSets (main) · Configuration	1.47 (1.18-1.77) percent	type: bootstrap confidence interval; level: 0.999; resamples: 20000; lower: 1.18; upper: 1.77	Author-reported evaluation · source checkedMassSpecGym: A benchmark for the discovery and identification of molecules · Table 3, main challenge, DeepSets, Hit rate @1
Fingerprint FFN (main) · Configuration	2.54 (2.17-2.99) percent	type: bootstrap confidence interval; level: 0.999; resamples: 20000; lower: 2.17; upper: 2.99	Author-reported evaluation · source checkedMassSpecGym: A benchmark for the discovery and identification of molecules · Table 3, main challenge, Fingerprint FFN, Hit rate @1
DeepSets + Fourier features (main) · Configuration	5.24 (4.71-5.83) percent	type: bootstrap confidence interval; level: 0.999; resamples: 20000; lower: 4.71; upper: 5.83	Author-reported evaluation · source checkedMassSpecGym: A benchmark for the discovery and identification of molecules · Table 3, main challenge, DeepSets + Fourier features, Hit rate @1
MIST (main) · Configuration	14.64 (13.82-15.54) percent	type: bootstrap confidence interval; level: 0.999; resamples: 20000; lower: 13.82; upper: 15.54	Author-reported evaluation · source checkedMassSpecGym: A benchmark for the discovery and identification of molecules · Table 3, main challenge, MIST, Hit rate @1

Scope and limitations

Source metric equations define hit rates as fractions, while table values use percentage scale; recorded values are not rescaled.
Zero generation accuracy is a reported result, not missingness.
No checkpoint revision inferred from method name.
Bootstrap CIs reflect resampling of this test set; not independently reproduced and not seed standard deviations.
Full table retained, including weaker/random methods and unavailable formula-simulation similarity cells.

Source transcription and grouping reviewed by automated source review on 2026-09-17. These experiments were not independently reproduced by rewire.

Tested entities and results

Release 2026-09-17-d277315f7d76 · 10 evaluations · 40 metric rows. Different protocols are not a single leaderboard.

Evaluation setupMetricEvidence origin

Results grouped by the exact reported evaluation
Metric and finding	Coverage and uncertainty	Evidence
Random (main): MassSpecGym · main Configuration: Random (main)Protocol: MassSpecGym · main (MassSpecGym Molecule retrieval)Dataset: MassSpecGym · main Retrieve the correct molecule from up to 256 candidates using an input spectrum; compare separately within the main challenge and molecular-formula challenge. MCES single-linkage molecular clustering at threshold 10; fixed held-out test split Author-reported evaluation · Evaluation metadata: needs review
30.81 (30.40-31.21) MCES @1 Unit: edge-edit distance · Direction: lower	Uncertainty: type: bootstrap confidence interval; level: 0.999; resamples: 20000; lower: 30.40; upper: 31.21 Scored: Not reported · Eligible: Not reported	source checkedMassSpecGym: A benchmark for the discovery and identification of molecules · Table 3, main challenge, Random, MCES @1 Source checking is not independent reproduction.
2.01 (1.68-2.39)% Hit rate @5 Unit: percent · Direction: higher	Uncertainty: type: bootstrap confidence interval; level: 0.999; resamples: 20000; lower: 1.68; upper: 2.39 Scored: Not reported · Eligible: Not reported	source checkedMassSpecGym: A benchmark for the discovery and identification of molecules · Table 3, main challenge, Random, Hit rate @5 Source checking is not independent reproduction.
DeepSets + Fourier features (main): MassSpecGym · main Configuration: DeepSets + Fourier features (main)Protocol: MassSpecGym · main (MassSpecGym Molecule retrieval)Dataset: MassSpecGym · main Retrieve the correct molecule from up to 256 candidates using an input spectrum; compare separately within the main challenge and molecular-formula challenge. MCES single-linkage molecular clustering at threshold 10; fixed held-out test split Author-reported evaluation · Evaluation metadata: needs review
22.13 (21.85-22.43) MCES @1 Unit: edge-edit distance · Direction: lower	Uncertainty: type: bootstrap confidence interval; level: 0.999; resamples: 20000; lower: 21.85; upper: 22.43 Scored: Not reported · Eligible: Not reported	source checkedMassSpecGym: A benchmark for the discovery and identification of molecules · Table 3, main challenge, DeepSets + Fourier features, MCES @1 Source checking is not independent reproduction.
12.58 (11.80-13.42)% Hit rate @5 Unit: percent · Direction: higher	Uncertainty: type: bootstrap confidence interval; level: 0.999; resamples: 20000; lower: 11.80; upper: 13.42 Scored: Not reported · Eligible: Not reported	source checkedMassSpecGym: A benchmark for the discovery and identification of molecules · Table 3, main challenge, DeepSets + Fourier features, Hit rate @5 Source checking is not independent reproduction.
Random (formula): MassSpecGym · formula Configuration: Random (formula)Protocol: MassSpecGym · formula (MassSpecGym Molecule retrieval)Dataset: MassSpecGym · formula Retrieve the correct molecule from up to 256 candidates using an input spectrum; compare separately within the main challenge and molecular-formula challenge. MCES single-linkage molecular clustering at threshold 10; fixed held-out test split Author-reported evaluation · Evaluation metadata: needs review
27.74 (26.52-28.84)% Hit rate @20 Unit: percent · Direction: higher	Uncertainty: type: bootstrap confidence interval; level: 0.999; resamples: 20000; lower: 26.52; upper: 28.84 Scored: Not reported · Eligible: Not reported	source checkedMassSpecGym: A benchmark for the discovery and identification of molecules · Table 3, bonus formula challenge, Random, Hit rate @20 Source checking is not independent reproduction.
11.35 (10.60-12.12)% Hit rate @5 Unit: percent · Direction: higher	Uncertainty: type: bootstrap confidence interval; level: 0.999; resamples: 20000; lower: 10.60; upper: 12.12 Scored: Not reported · Eligible: Not reported	source checkedMassSpecGym: A benchmark for the discovery and identification of molecules · Table 3, bonus formula challenge, Random, Hit rate @5 Source checking is not independent reproduction.
13.87 (13.70-14.03) MCES @1 Unit: edge-edit distance · Direction: lower	Uncertainty: type: bootstrap confidence interval; level: 0.999; resamples: 20000; lower: 13.70; upper: 14.03 Scored: Not reported · Eligible: Not reported	source checkedMassSpecGym: A benchmark for the discovery and identification of molecules · Table 3, bonus formula challenge, Random, MCES @1 Source checking is not independent reproduction.
DeepSets (formula): MassSpecGym · formula Configuration: DeepSets (formula)Protocol: MassSpecGym · formula (MassSpecGym Molecule retrieval)Dataset: MassSpecGym · formula Retrieve the correct molecule from up to 256 candidates using an input spectrum; compare separately within the main challenge and molecular-formula challenge. MCES single-linkage molecular clustering at threshold 10; fixed held-out test split Author-reported evaluation · Evaluation metadata: needs review
4.42 (3.92-4.97)% Hit rate @1 Unit: percent · Direction: higher	Uncertainty: type: bootstrap confidence interval; level: 0.999; resamples: 20000; lower: 3.92; upper: 4.97 Scored: Not reported · Eligible: Not reported	source checkedMassSpecGym: A benchmark for the discovery and identification of molecules · Table 3, bonus formula challenge, DeepSets, Hit rate @1 Source checking is not independent reproduction.
14.46 (13.58-15.36)% Hit rate @5 Unit: percent · Direction: higher	Uncertainty: type: bootstrap confidence interval; level: 0.999; resamples: 20000; lower: 13.58; upper: 15.36 Scored: Not reported · Eligible: Not reported	source checkedMassSpecGym: A benchmark for the discovery and identification of molecules · Table 3, bonus formula challenge, DeepSets, Hit rate @5 Source checking is not independent reproduction.
15.04 (14.89-15.19) MCES @1 Unit: edge-edit distance · Direction: lower	Uncertainty: type: bootstrap confidence interval; level: 0.999; resamples: 20000; lower: 14.89; upper: 15.19 Scored: Not reported · Eligible: Not reported	source checkedMassSpecGym: A benchmark for the discovery and identification of molecules · Table 3, bonus formula challenge, DeepSets, MCES @1 Source checking is not independent reproduction.
Fingerprint FFN (main): MassSpecGym · main Configuration: Fingerprint FFN (main)Protocol: MassSpecGym · main (MassSpecGym Molecule retrieval)Dataset: MassSpecGym · main Retrieve the correct molecule from up to 256 candidates using an input spectrum; compare separately within the main challenge and molecular-formula challenge. MCES single-linkage molecular clustering at threshold 10; fixed held-out test split Author-reported evaluation · Evaluation metadata: needs review
20.00 (19.01-20.98)% Hit rate @20 Unit: percent · Direction: higher	Uncertainty: type: bootstrap confidence interval; level: 0.999; resamples: 20000; lower: 19.01; upper: 20.98 Scored: Not reported · Eligible: Not reported	source checkedMassSpecGym: A benchmark for the discovery and identification of molecules · Table 3, main challenge, Fingerprint FFN, Hit rate @20 Source checking is not independent reproduction.
7.59 (6.96-8.28)% Hit rate @5 Unit: percent · Direction: higher	Uncertainty: type: bootstrap confidence interval; level: 0.999; resamples: 20000; lower: 6.96; upper: 8.28 Scored: Not reported · Eligible: Not reported	source checkedMassSpecGym: A benchmark for the discovery and identification of molecules · Table 3, main challenge, Fingerprint FFN, Hit rate @5 Source checking is not independent reproduction.
24.66 (24.38-24.94) MCES @1 Unit: edge-edit distance · Direction: lower	Uncertainty: type: bootstrap confidence interval; level: 0.999; resamples: 20000; lower: 24.38; upper: 24.94 Scored: Not reported · Eligible: Not reported	source checkedMassSpecGym: A benchmark for the discovery and identification of molecules · Table 3, main challenge, Fingerprint FFN, MCES @1 Source checking is not independent reproduction.
DeepSets (main): MassSpecGym · main Configuration: DeepSets (main)Protocol: MassSpecGym · main (MassSpecGym Molecule retrieval)Dataset: MassSpecGym · main Retrieve the correct molecule from up to 256 candidates using an input spectrum; compare separately within the main challenge and molecular-formula challenge. MCES single-linkage molecular clustering at threshold 10; fixed held-out test split Author-reported evaluation · Evaluation metadata: needs review
6.21 (5.64-6.82)% Hit rate @5 Unit: percent · Direction: higher	Uncertainty: type: bootstrap confidence interval; level: 0.999; resamples: 20000; lower: 5.64; upper: 6.82 Scored: Not reported · Eligible: Not reported	source checkedMassSpecGym: A benchmark for the discovery and identification of molecules · Table 3, main challenge, DeepSets, Hit rate @5 Source checking is not independent reproduction.
25.11 (24.84-25.39) MCES @1 Unit: edge-edit distance · Direction: lower	Uncertainty: type: bootstrap confidence interval; level: 0.999; resamples: 20000; lower: 24.84; upper: 25.39 Scored: Not reported · Eligible: Not reported	source checkedMassSpecGym: A benchmark for the discovery and identification of molecules · Table 3, main challenge, DeepSets, MCES @1 Source checking is not independent reproduction.
MIST (formula): MassSpecGym · formula Configuration: MIST (formula)Protocol: MassSpecGym · formula (MassSpecGym Molecule retrieval)Dataset: MassSpecGym · formula Retrieve the correct molecule from up to 256 candidates using an input spectrum; compare separately within the main challenge and molecular-formula challenge. MCES single-linkage molecular clustering at threshold 10; fixed held-out test split Author-reported evaluation · Evaluation metadata: needs review
41.12 (39.98-42.34)% Hit rate @20 Unit: percent · Direction: higher	Uncertainty: type: bootstrap confidence interval; level: 0.999; resamples: 20000; lower: 39.98; upper: 42.34 Scored: Not reported · Eligible: Not reported	source checkedMassSpecGym: A benchmark for the discovery and identification of molecules · Table 3, bonus formula challenge, MIST, Hit rate @20 Source checking is not independent reproduction.
9.57 (8.88-10.30)% Hit rate @1 Unit: percent · Direction: higher	Uncertainty: type: bootstrap confidence interval; level: 0.999; resamples: 20000; lower: 8.88; upper: 10.30 Scored: Not reported · Eligible: Not reported	source checkedMassSpecGym: A benchmark for the discovery and identification of molecules · Table 3, bonus formula challenge, MIST, Hit rate @1 Source checking is not independent reproduction.
22.11 (21.10-23.13)% Hit rate @5 Unit: percent · Direction: higher	Uncertainty: type: bootstrap confidence interval; level: 0.999; resamples: 20000; lower: 21.10; upper: 23.13 Scored: Not reported · Eligible: Not reported	source checkedMassSpecGym: A benchmark for the discovery and identification of molecules · Table 3, bonus formula challenge, MIST, Hit rate @5 Source checking is not independent reproduction.
DeepSets + Fourier features (formula): MassSpecGym · formula Configuration: DeepSets + Fourier features (formula)Protocol: MassSpecGym · formula (MassSpecGym Molecule retrieval)Dataset: MassSpecGym · formula Retrieve the correct molecule from up to 256 candidates using an input spectrum; compare separately within the main challenge and molecular-formula challenge. MCES single-linkage molecular clustering at threshold 10; fixed held-out test split Author-reported evaluation · Evaluation metadata: needs review
33.46 (32.39-34.59)% Hit rate @20 Unit: percent · Direction: higher	Uncertainty: type: bootstrap confidence interval; level: 0.999; resamples: 20000; lower: 32.39; upper: 34.59 Scored: Not reported · Eligible: Not reported	source checkedMassSpecGym: A benchmark for the discovery and identification of molecules · Table 3, bonus formula challenge, DeepSets + Fourier features, Hit rate @20 Source checking is not independent reproduction.
6.56 (5.95-7.16)% Hit rate @1 Unit: percent · Direction: higher	Uncertainty: type: bootstrap confidence interval; level: 0.999; resamples: 20000; lower: 5.95; upper: 7.16 Scored: Not reported · Eligible: Not reported	source checkedMassSpecGym: A benchmark for the discovery and identification of molecules · Table 3, bonus formula challenge, DeepSets + Fourier features, Hit rate @1 Source checking is not independent reproduction.
MIST (main): MassSpecGym · main Configuration: MIST (main)Protocol: MassSpecGym · main (MassSpecGym Molecule retrieval)Dataset: MassSpecGym · main Retrieve the correct molecule from up to 256 candidates using an input spectrum; compare separately within the main challenge and molecular-formula challenge. MCES single-linkage molecular clustering at threshold 10; fixed held-out test split Author-reported evaluation · Evaluation metadata: needs review
59.15 (57.89-60.39)% Hit rate @20 Unit: percent · Direction: higher	Uncertainty: type: bootstrap confidence interval; level: 0.999; resamples: 20000; lower: 57.89; upper: 60.39 Scored: Not reported · Eligible: Not reported	source checkedMassSpecGym: A benchmark for the discovery and identification of molecules · Table 3, main challenge, MIST, Hit rate @20 Source checking is not independent reproduction.
15.37 (15.12-15.62) MCES @1 Unit: edge-edit distance · Direction: lower	Uncertainty: type: bootstrap confidence interval; level: 0.999; resamples: 20000; lower: 15.12; upper: 15.62 Scored: Not reported · Eligible: Not reported	source checkedMassSpecGym: A benchmark for the discovery and identification of molecules · Table 3, main challenge, MIST, MCES @1 Source checking is not independent reproduction.
Fingerprint FFN (formula): MassSpecGym · formula Configuration: Fingerprint FFN (formula)Protocol: MassSpecGym · formula (MassSpecGym Molecule retrieval)Dataset: MassSpecGym · formula Retrieve the correct molecule from up to 256 candidates using an input spectrum; compare separately within the main challenge and molecular-formula challenge. MCES single-linkage molecular clustering at threshold 10; fixed held-out test split Author-reported evaluation · Evaluation metadata: needs review
5.09 (4.57-5.66)% Hit rate @1 Unit: percent · Direction: higher	Uncertainty: type: bootstrap confidence interval; level: 0.999; resamples: 20000; lower: 4.57; upper: 5.66 Scored: Not reported · Eligible: Not reported	source checkedMassSpecGym: A benchmark for the discovery and identification of molecules · Table 3, bonus formula challenge, Fingerprint FFN, Hit rate @1 Source checking is not independent reproduction.
14.69 (13.83-15.56)% Hit rate @5 Unit: percent · Direction: higher	Uncertainty: type: bootstrap confidence interval; level: 0.999; resamples: 20000; lower: 13.83; upper: 15.56 Scored: Not reported · Eligible: Not reported	source checkedMassSpecGym: A benchmark for the discovery and identification of molecules · Table 3, bonus formula challenge, Fingerprint FFN, Hit rate @5 Source checking is not independent reproduction.
14.94 (14.79-15.09) MCES @1 Unit: edge-edit distance · Direction: lower	Uncertainty: type: bootstrap confidence interval; level: 0.999; resamples: 20000; lower: 14.79; upper: 15.09 Scored: Not reported · Eligible: Not reported	source checkedMassSpecGym: A benchmark for the discovery and identification of molecules · Table 3, bonus formula challenge, Fingerprint FFN, MCES @1 Source checking is not independent reproduction.

Papers and result coverage

Last literature check: 2026-09-17. Primary-source discovery and table/protocol screening; source checked is not independently reproduced. Raw acquisitions not automatically numerical publication approval.

Paper or primary resource	Version	Reference
MassSpecGym: A benchmark for the discovery and identification of molecules	2410.23326v1	Read source

What is still missing

Independent batch review before import; preserve existing observation identities.

Search and extraction details

complete comparison tables extracted pending publication review

Searches

MassSpecGym 2410.23326 benchmark results Table 1 Table 2

Evidence locations

Tables2–4; task definitions and supplementary split table

Strengths and limitations

Strengths supported by sources

One release supplies explicit task interfaces and predefined splits for three complementary predictions.
Sources (2)
pluskal-lab/MassSpecGym official source; pluskal-lab/MassSpecGym massspecgym/models/retrieval/base.py · Pinned README: three challenges; dataset and DataModule; evaluation base classes; pinned massspecgym/models/retrieval/base.py evaluation methods

Limitations and conditions

Chemical-formula-assisted tasks provide extra input information and must remain separate from unassisted tasks. The MCES split constrains structural similarity but cannot establish independence from every external pretraining corpus.
Sources
massspecgym primary benchmark evidence · Section 3.4; Supplementary Information 2.5; Tables 2–4

Profile review details

Primary paper and/or task implementation reviewed for the explicitly cited methodology claims. Scope-limited absence is recorded only after the documented source search; no model runs or independent reproduction.

Stable record: discovery-benchmark-massspecgym-molecule-retrieval

Applicable tests and references

Applicability is distinct from a completed evaluation.

Mass spectral cosine matching · proposed

Evidence table

Inspect claims, sources and review details

Trace each statement to its source and review. A context-only reference supports the record generally; it does not verify an individual field. Source checking does not reproduce an experiment.

One row per statement and cited source. Multiple citations are not independent evaluations. Shared locators are labelled explicitly.

37 evidence rows matching the loaded filters

Claims, original sources and review scope · Release 2026-09-17-d277315f7d76
Property and statement	Original source and location	Review and provenance
Diagram caption Conceptual procedure. Task variants and protocol versions retain their separate scoring conditions. Individual claims	pluskal-lab/MassSpecGym massspecgym/models/retrieval/base.py Original source ↗ Pinned README: three challenges; dataset and DataModule; evaluation base classes; pinned massspecgym/models/retrieval/base.py evaluation methods; Section 3.4; Supplementary Information 2.5 Shared locator for this statement’s cited sources; not a separate locator for each citation. Version: f259fe3780d5bd227fc6ece36ce6f397c2eef716 Retrieved: 2026-09-16T20:42:47.889112+00:00	source checked automated source review · 2026-09-16 Audit details Primary paper and/or task implementation reviewed for the explicitly cited methodology claims. Scope-limited absence is recorded only after the documented source search; no model runs or independent reproduction. Field: `attributes.profile.diagram.caption` Source artifact SHA-256: `2ae08f58bd6db0e00430b54c078b45ae91b9252aac02ee358fef9d8ba1b3f81f` Hash scope: Hash scope not separately documented; inspect source record Inspected artifact
Diagram caption Conceptual procedure. Task variants and protocol versions retain their separate scoring conditions. Individual claims	massspecgym primary benchmark evidence Original source ↗ Pinned README: three challenges; dataset and DataModule; evaluation base classes; pinned massspecgym/models/retrieval/base.py evaluation methods; Section 3.4; Supplementary Information 2.5 Shared locator for this statement’s cited sources; not a separate locator for each citation. Version: 2410.23326v1 Retrieved: 2026-09-16T21:04:58.775040+00:00	source checked automated source review · 2026-09-16 Audit details Primary paper and/or task implementation reviewed for the explicitly cited methodology claims. Scope-limited absence is recorded only after the documented source search; no model runs or independent reproduction. Field: `attributes.profile.diagram.caption` Source artifact SHA-256: `82176d50e8947c8b9baa2a0d91493f5680c0e4c7e25a2266ff7879f48a58c40c` Hash scope: Hash scope not separately documented; inspect source record Inspected artifact
Diagram caption Conceptual procedure. Task variants and protocol versions retain their separate scoring conditions. Individual claims	pluskal-lab/MassSpecGym official source Original source ↗ Pinned README: three challenges; dataset and DataModule; evaluation base classes; pinned massspecgym/models/retrieval/base.py evaluation methods; Section 3.4; Supplementary Information 2.5 Shared locator for this statement’s cited sources; not a separate locator for each citation. Version: f259fe3780d5bd227fc6ece36ce6f397c2eef716 Retrieved: 2026-09-16T10:30:23.954980+00:00	source checked automated source review · 2026-09-16 Audit details Primary paper and/or task implementation reviewed for the explicitly cited methodology claims. Scope-limited absence is recorded only after the documented source search; no model runs or independent reproduction. Field: `attributes.profile.diagram.caption` Source artifact SHA-256: `08bf3607e6e2e5462b81eac85d0e71d9d23ce1c9bf1a370c9d1079ecd60ee2d8` Hash scope: Hash scope not separately documented; inspect source record
Diagram steps ["Allowed inputs: MS/MS spectrum plus the supplied molecule candidate set.","Splits: MCES molecular clusters are grouped into fixed training, validation and test folds, stratified by acquisition metadata. Cross-fold molecular bond-edit distance is at least 10; all spectra follow the assigned molecular fold.","Metrics: Mean hit rate at configured top-k cutoffs, with optional MCES distance for the top-ranked candidate."] Individual claims	pluskal-lab/MassSpecGym massspecgym/models/retrieval/base.py Original source ↗ Pinned README: three challenges; dataset and DataModule; evaluation base classes; pinned massspecgym/models/retrieval/base.py evaluation methods; Section 3.4; Supplementary Information 2.5 Shared locator for this statement’s cited sources; not a separate locator for each citation. Version: f259fe3780d5bd227fc6ece36ce6f397c2eef716 Retrieved: 2026-09-16T20:42:47.889112+00:00	source checked automated source review · 2026-09-16 Audit details Primary paper and/or task implementation reviewed for the explicitly cited methodology claims. Scope-limited absence is recorded only after the documented source search; no model runs or independent reproduction. Field: `attributes.profile.diagram.steps` Source artifact SHA-256: `2ae08f58bd6db0e00430b54c078b45ae91b9252aac02ee358fef9d8ba1b3f81f` Hash scope: Hash scope not separately documented; inspect source record Inspected artifact
Diagram steps ["Allowed inputs: MS/MS spectrum plus the supplied molecule candidate set.","Splits: MCES molecular clusters are grouped into fixed training, validation and test folds, stratified by acquisition metadata. Cross-fold molecular bond-edit distance is at least 10; all spectra follow the assigned molecular fold.","Metrics: Mean hit rate at configured top-k cutoffs, with optional MCES distance for the top-ranked candidate."] Individual claims	massspecgym primary benchmark evidence Original source ↗ Pinned README: three challenges; dataset and DataModule; evaluation base classes; pinned massspecgym/models/retrieval/base.py evaluation methods; Section 3.4; Supplementary Information 2.5 Shared locator for this statement’s cited sources; not a separate locator for each citation. Version: 2410.23326v1 Retrieved: 2026-09-16T21:04:58.775040+00:00	source checked automated source review · 2026-09-16 Audit details Primary paper and/or task implementation reviewed for the explicitly cited methodology claims. Scope-limited absence is recorded only after the documented source search; no model runs or independent reproduction. Field: `attributes.profile.diagram.steps` Source artifact SHA-256: `82176d50e8947c8b9baa2a0d91493f5680c0e4c7e25a2266ff7879f48a58c40c` Hash scope: Hash scope not separately documented; inspect source record Inspected artifact
Diagram steps ["Allowed inputs: MS/MS spectrum plus the supplied molecule candidate set.","Splits: MCES molecular clusters are grouped into fixed training, validation and test folds, stratified by acquisition metadata. Cross-fold molecular bond-edit distance is at least 10; all spectra follow the assigned molecular fold.","Metrics: Mean hit rate at configured top-k cutoffs, with optional MCES distance for the top-ranked candidate."] Individual claims	pluskal-lab/MassSpecGym official source Original source ↗ Pinned README: three challenges; dataset and DataModule; evaluation base classes; pinned massspecgym/models/retrieval/base.py evaluation methods; Section 3.4; Supplementary Information 2.5 Shared locator for this statement’s cited sources; not a separate locator for each citation. Version: f259fe3780d5bd227fc6ece36ce6f397c2eef716 Retrieved: 2026-09-16T10:30:23.954980+00:00	source checked automated source review · 2026-09-16 Audit details Primary paper and/or task implementation reviewed for the explicitly cited methodology claims. Scope-limited absence is recorded only after the documented source search; no model runs or independent reproduction. Field: `attributes.profile.diagram.steps` Source artifact SHA-256: `08bf3607e6e2e5462b81eac85d0e71d9d23ce1c9bf1a370c9d1079ecd60ee2d8` Hash scope: Hash scope not separately documented; inspect source record
Diagram title Evaluation procedure Individual claims	pluskal-lab/MassSpecGym massspecgym/models/retrieval/base.py Original source ↗ Pinned README: three challenges; dataset and DataModule; evaluation base classes; pinned massspecgym/models/retrieval/base.py evaluation methods; Section 3.4; Supplementary Information 2.5 Shared locator for this statement’s cited sources; not a separate locator for each citation. Version: f259fe3780d5bd227fc6ece36ce6f397c2eef716 Retrieved: 2026-09-16T20:42:47.889112+00:00	source checked automated source review · 2026-09-16 Audit details Primary paper and/or task implementation reviewed for the explicitly cited methodology claims. Scope-limited absence is recorded only after the documented source search; no model runs or independent reproduction. Field: `attributes.profile.diagram.title` Source artifact SHA-256: `2ae08f58bd6db0e00430b54c078b45ae91b9252aac02ee358fef9d8ba1b3f81f` Hash scope: Hash scope not separately documented; inspect source record Inspected artifact
Diagram title Evaluation procedure Individual claims	massspecgym primary benchmark evidence Original source ↗ Pinned README: three challenges; dataset and DataModule; evaluation base classes; pinned massspecgym/models/retrieval/base.py evaluation methods; Section 3.4; Supplementary Information 2.5 Shared locator for this statement’s cited sources; not a separate locator for each citation. Version: 2410.23326v1 Retrieved: 2026-09-16T21:04:58.775040+00:00	source checked automated source review · 2026-09-16 Audit details Primary paper and/or task implementation reviewed for the explicitly cited methodology claims. Scope-limited absence is recorded only after the documented source search; no model runs or independent reproduction. Field: `attributes.profile.diagram.title` Source artifact SHA-256: `82176d50e8947c8b9baa2a0d91493f5680c0e4c7e25a2266ff7879f48a58c40c` Hash scope: Hash scope not separately documented; inspect source record Inspected artifact
Diagram title Evaluation procedure Individual claims	pluskal-lab/MassSpecGym official source Original source ↗ Pinned README: three challenges; dataset and DataModule; evaluation base classes; pinned massspecgym/models/retrieval/base.py evaluation methods; Section 3.4; Supplementary Information 2.5 Shared locator for this statement’s cited sources; not a separate locator for each citation. Version: f259fe3780d5bd227fc6ece36ce6f397c2eef716 Retrieved: 2026-09-16T10:30:23.954980+00:00	source checked automated source review · 2026-09-16 Audit details Primary paper and/or task implementation reviewed for the explicitly cited methodology claims. Scope-limited absence is recorded only after the documented source search; no model runs or independent reproduction. Field: `attributes.profile.diagram.title` Source artifact SHA-256: `08bf3607e6e2e5462b81eac85d0e71d9d23ce1c9bf1a370c9d1079ecd60ee2d8` Hash scope: Hash scope not separately documented; inspect source record
Datasets MS/MS spectrum plus a candidate set; rank the matching molecule. Individual claims	pluskal-lab/MassSpecGym massspecgym/models/retrieval/base.py Original source ↗ Pinned README: three challenges; dataset and DataModule; evaluation base classes; pinned massspecgym/models/retrieval/base.py evaluation methods Shared locator for this statement’s cited sources; not a separate locator for each citation. Version: f259fe3780d5bd227fc6ece36ce6f397c2eef716 Retrieved: 2026-09-16T20:42:47.889112+00:00	source checked automated source review · 2026-09-16 Audit details Primary paper and/or task implementation reviewed for the explicitly cited methodology claims. Scope-limited absence is recorded only after the documented source search; no model runs or independent reproduction. Field: `attributes.profile.facts.0.value` Source artifact SHA-256: `2ae08f58bd6db0e00430b54c078b45ae91b9252aac02ee358fef9d8ba1b3f81f` Hash scope: Hash scope not separately documented; inspect source record Inspected artifact

All evidence (CSV)All evidence (JSONL)How to read the evidence

Sources and history

Release 2026-09-17-d277315f7d76 · Record review: discovered

4 source records and release history

pluskal-lab/MassSpecGym official source · Original source · f259fe3780d5bd227fc6ece36ce6f397c2eef716
MassSpecGym: A benchmark for the discovery and identification of molecules · Original source · 2410.23326v1
pluskal-lab/MassSpecGym massspecgym/models/retrieval/base.py · Original source · f259fe3780d5bd227fc6ece36ce6f397c2eef716
massspecgym primary benchmark evidence · Original source · 2410.23326v1

Download this release

Technical metadata and extraction receipts

Stable ID: discovery-benchmark-massspecgym-molecule-retrieval

areas: metabolomics
entity level: task
scope note: Specialist molecular or omics evaluation; protocol details require review before numerical comparison.
task: Rank candidate structures from a tandem mass spectrum
version: Not reported
comparison panels: id: part2-massspecgym-table3-65ea68de77; title: MassSpecGym · main · Table 3; protocol id: paper-protocol-a18b4bc79049513fb7; dataset id: paper-dataset-43f783b5d7a7151aa9; metric: Hit rate @1; unit: percent; direction: higher; result ids: paper-result-9b0cc277a3cbc6c13d; paper-result-acf85ffa57a1906421; paper-result-a0d66c2c9d6bc8466d; paper-result-d73962270e10586574; paper-result-d5320e4453da6b09dd; source ids: part2-massspecgym-arxiv-v1; source locator: Table 3: Hit rate @1, main; context: Retrieve the correct molecule from up to 256 candidates using an input spectrum; compare separately within the main challenge and molecular-formula challenge. MCES single-linkage molecular clustering at threshold 10; fixed held-out test split; caveats: Source metric equations define hit rates as fractions, while table values use percentage scale; recorded values are not rescaled.; Zero generation accuracy is a reported result, not missingness.; No checkpoint revision inferred from method name.; Bootstrap CIs reflect resampling of this test set; not independently reproduced and not seed standard deviations.; Full table retained, including weaker/random methods and unavailable formula-simulation similarity cells.; review: method: automated_source_review; date: 2026-09-17; id: part2-massspecgym-table3-dffdc0b0ef; title: MassSpecGym · main · Table 3; protocol id: paper-protocol-a18b4bc79049513fb7; dataset id: paper-dataset-43f783b5d7a7151aa9; metric: Hit rate @5; unit: percent; direction: higher; result ids: paper-result-35b54face42cd4df98; paper-result-3588b0e1eb7078ee40; paper-result-3bda21118a05d4e278; paper-result-46e2e44bcf768e94b2; paper-result-ed71048c3ebde971d5; source ids: part2-massspecgym-arxiv-v1; source locator: Table 3: Hit rate @5, main; context: Retrieve the correct molecule from up to 256 candidates using an input spectrum; compare separately within the main challenge and molecular-formula challenge. MCES single-linkage molecular clustering at threshold 10; fixed held-out test split; caveats: Source metric equations define hit rates as fractions, while table values use percentage scale; recorded values are not rescaled.; Zero generation accuracy is a reported result, not missingness.; No checkpoint revision inferred from method name.; Bootstrap CIs reflect resampling of this test set; not independently reproduced and not seed standard deviations.; Full table retained, including weaker/random methods and unavailable formula-simulation similarity cells.; review: method: automated_source_review; date: 2026-09-17; id: part2-massspecgym-table3-629ab3b450; title: MassSpecGym · main · Table 3; protocol id: paper-protocol-a18b4bc79049513fb7; dataset id: paper-dataset-43f783b5d7a7151aa9; metric: Hit rate @20; unit: percent; direction: higher; result ids: paper-result-f871d7d35c275e848e; paper-result-e3dc19ea2e9e59e293; paper-result-29a4b30ef091c9e0e8; paper-result-cde595ad3523bf9f27; paper-result-6b9b0e247bfdc9c274; source ids: part2-massspecgym-arxiv-v1; source locator: Table 3: Hit rate @20, main; context: Retrieve the correct molecule from up to 256 candidates using an input spectrum; compare separately within the main challenge and molecular-formula challenge. MCES single-linkage molecular clustering at threshold 10; fixed held-out test split; caveats: Source metric equations define hit rates as fractions, while table values use percentage scale; recorded values are not rescaled.; Zero generation accuracy is a reported result, not missingness.; No checkpoint revision inferred from method name.; Bootstrap CIs reflect resampling of this test set; not independently reproduced and not seed standard deviations.; Full table retained, including weaker/random methods and unavailable formula-simulation similarity cells.; review: method: automated_source_review; date: 2026-09-17; id: part2-massspecgym-table3-b934e3725f; title: MassSpecGym · main · Table 3; protocol id: paper-protocol-a18b4bc79049513fb7; dataset id: paper-dataset-43f783b5d7a7151aa9; metric: MCES @1; unit: edge-edit distance; direction: lower; result ids: paper-result-002aa84a8224be7916; paper-result-3ff182c671377da6da; paper-result-57fd7d0cd8b73c56e2; paper-result-0dcae24f1e2c35b445; paper-result-80aa6ca411586dd8e5; source ids: part2-massspecgym-arxiv-v1; source locator: Table 3: MCES @1, main; context: Retrieve the correct molecule from up to 256 candidates using an input spectrum; compare separately within the main challenge and molecular-formula challenge. MCES single-linkage molecular clustering at threshold 10; fixed held-out test split; caveats: Source metric equations define hit rates as fractions, while table values use percentage scale; recorded values are not rescaled.; Zero generation accuracy is a reported result, not missingness.; No checkpoint revision inferred from method name.; Bootstrap CIs reflect resampling of this test set; not independently reproduced and not seed standard deviations.; Full table retained, including weaker/random methods and unavailable formula-simulation similarity cells.; review: method: automated_source_review; date: 2026-09-17; id: part2-massspecgym-table3-ac4b0b4611; title: MassSpecGym · formula · Table 3; protocol id: paper-protocol-1e7bb17e6ce294d906; dataset id: paper-dataset-e3dee7ae24615eecfb; metric: Hit rate @1; unit: percent; direction: higher; result ids: paper-result-96eb3d9e4116e85dc3; paper-result-19979776fcf90cd085; paper-result-6d5f81ef670db6f81b; paper-result-7e55a965688b10e2cc; paper-result-761f932c6a35c888bb; source ids: part2-massspecgym-arxiv-v1; source locator: Table 3: Hit rate @1, formula; context: Retrieve the correct molecule from up to 256 candidates using an input spectrum; compare separately within the main challenge and molecular-formula challenge. MCES single-linkage molecular clustering at threshold 10; fixed held-out test split; caveats: Source metric equations define hit rates as fractions, while table values use percentage scale; recorded values are not rescaled.; Zero generation accuracy is a reported result, not missingness.; No checkpoint revision inferred from method name.; Bootstrap CIs reflect resampling of this test set; not independently reproduced and not seed standard deviations.; Full table retained, including weaker/random methods and unavailable formula-simulation similarity cells.; review: method: automated_source_review; date: 2026-09-17; id: part2-massspecgym-table3-a71e8b64c9; title: MassSpecGym · formula · Table 3; protocol id: paper-protocol-1e7bb17e6ce294d906; dataset id: paper-dataset-e3dee7ae24615eecfb; metric: Hit rate @5; unit: percent; direction: higher; result ids: paper-result-6bf388327e502fbdd6; paper-result-1b7eee640fec71e99b; paper-result-773438d255878a0817; paper-result-a2021eb25d2e915417; paper-result-7a7bc4ebdca6051342; source ids: part2-massspecgym-arxiv-v1; source locator: Table 3: Hit rate @5, formula; context: Retrieve the correct molecule from up to 256 candidates using an input spectrum; compare separately within the main challenge and molecular-formula challenge. MCES single-linkage molecular clustering at threshold 10; fixed held-out test split; caveats: Source metric equations define hit rates as fractions, while table values use percentage scale; recorded values are not rescaled.; Zero generation accuracy is a reported result, not missingness.; No checkpoint revision inferred from method name.; Bootstrap CIs reflect resampling of this test set; not independently reproduced and not seed standard deviations.; Full table retained, including weaker/random methods and unavailable formula-simulation similarity cells.; review: method: automated_source_review; date: 2026-09-17; id: part2-massspecgym-table3-8dc4756cf6; title: MassSpecGym · formula · Table 3; protocol id: paper-protocol-1e7bb17e6ce294d906; dataset id: paper-dataset-e3dee7ae24615eecfb; metric: Hit rate @20; unit: percent; direction: higher; result ids: paper-result-1816c840f0de8bc10b; paper-result-c905d384ee9e3b5f91; paper-result-e1302d68232a47d02d; paper-result-46121102e555a35255; paper-result-3b6baab5e2664e30e6; source ids: part2-massspecgym-arxiv-v1; source locator: Table 3: Hit rate @20, formula; context: Retrieve the correct molecule from up to 256 candidates using an input spectrum; compare separately within the main challenge and molecular-formula challenge. MCES single-linkage molecular clustering at threshold 10; fixed held-out test split; caveats: Source metric equations define hit rates as fractions, while table values use percentage scale; recorded values are not rescaled.; Zero generation accuracy is a reported result, not missingness.; No checkpoint revision inferred from method name.; Bootstrap CIs reflect resampling of this test set; not independently reproduced and not seed standard deviations.; Full table retained, including weaker/random methods and unavailable formula-simulation similarity cells.; review: method: automated_source_review; date: 2026-09-17; id: part2-massspecgym-table3-ebdb34929b; title: MassSpecGym · formula · Table 3; protocol id: paper-protocol-1e7bb17e6ce294d906; dataset id: paper-dataset-e3dee7ae24615eecfb; metric: MCES @1; unit: edge-edit distance; direction: lower; result ids: paper-result-8b2608a402d6ab54e6; paper-result-47e90701446fc3f6b0; paper-result-88df2cfb82ff24f522; paper-result-8d04b85ca056e77ff0; paper-result-e5a6f34b3603af99f9; source ids: part2-massspecgym-arxiv-v1; source locator: Table 3: MCES @1, formula; context: Retrieve the correct molecule from up to 256 candidates using an input spectrum; compare separately within the main challenge and molecular-formula challenge. MCES single-linkage molecular clustering at threshold 10; fixed held-out test split; caveats: Source metric equations define hit rates as fractions, while table values use percentage scale; recorded values are not rescaled.; Zero generation accuracy is a reported result, not missingness.; No checkpoint revision inferred from method name.; Bootstrap CIs reflect resampling of this test set; not independently reproduced and not seed standard deviations.; Full table retained, including weaker/random methods and unavailable formula-simulation similarity cells.; review: method: automated_source_review; date: 2026-09-17
benchmark research: review date: 2026-09-17; status: complete_comparison_tables_extracted_pending_publication_review; primary sources: part2-massspecgym-arxiv-v1; inspected locators: Tables2–4; task definitions and supplementary split table; searched queries: MassSpecGym 2410.23326 benchmark results Table 1 Table 2; gaps: Independent batch review before import; preserve existing observation identities.; claim scope: Primary-source discovery and table/protocol screening; source checked is not independently reproduced. Raw acquisitions not automatically numerical publication approval.
historical missing metadata: dataset release: unextracted; metric implementation: unextracted; split manifest: unextracted; version: unextracted
metadata review scope: historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.
legacy kinds: benchmark
entity classification: review date: 2026-09-17; rationale: This record identifies the biological prediction question or a suite-specific task, rather than a uniquely fixed evaluated procedure. Preserve its task identity and leave split, model adaptation and scoring details on linked protocols/evaluations.; source ids: src-discovery-pluskal-lab-massspecgym; evidence-benchmark-massspecgym-retrieval-base-py; source locator: Pinned README: three challenges; dataset and DataModule; evaluation base classes; pinned massspecgym/models/retrieval/base.py evaluation methods; ambiguities: A paper- or suite-specific task may constrain some inputs or metrics; that alone does not make it interchangeable with a complete versioned protocol. No protocol equivalence is inferred.

Related records

parent: MassSpecGym
part of: MassSpecGym
applicable to: Mass spectral cosine matching
evaluates task: MassSpecGym · formula (MassSpecGym Molecule retrieval)
evaluates task: MassSpecGym · main (MassSpecGym Molecule retrieval)
subject: MassSpecGym Molecule retrieval: part of MassSpecGym

Suggest a correction