← Benchmarks | Solvation

Explicit Solvation Benchmark OFFICIAL PACKET

Production explicit-solvation benchmark packet with a 642-case external FreeSolv hydration benchmark, curated non-water external packets, a 436-case large-scale non-water packet, and strict repeated-holdout validation for native non-water explicit carriers.

0.3295
MAE (kcal/mol)
1.3787
MAE (kJ/mol)
642
FreeSolv Cases
0.9857
R squared

Scope and contract

What is benchmarked

Water hydration free energy on external FreeSolv references.

What this is not

A claim of full broad-corpus parity for every non-water solvent. Non-water coverage is published here as curated external packets, a broader SolProp-enriched external packet, and strict repeated holdout.

Execution profile

Public screening run with fast hydration evaluation settings.

Reproducibility

Full summary JSON, case table CSV/JSON, failure note, and methodology are published.

Hydration Accuracy (FreeSolv)

Single-run benchmark snapshot used for this official packet.

Dataset Cases MAE (kcal/mol) RMSE (kcal/mol) MAE (kJ/mol) RMSE (kJ/mol) Pearson r R squared
FreeSolv screening subset 642 0.3295 0.4593 1.3787 1.9219 0.9928 0.9857

Auxiliary metrics: max absolute error 2.4019 kcal/mol, median absolute error 0.2427 kcal/mol, within-uncertainty fraction 0.7492.

Solvent Model Coverage Status

Explicit carrier status in production benchmark policy.

Model Solvent Status External benchmark in this packet
Water (TIP3P) Water Production Yes
Water (TIP4P) Water Production Reference policy-listed model
Methanol Methanol Native explicit carrier External phase-1 packet + strict holdout
Ethanol Ethanol Native explicit carrier External phase-1 packet + strict holdout
Acetonitrile Acetonitrile Native explicit carrier External phase-1 packet + strict holdout
DMSO DMSO Native explicit carrier External phase-1 packet + strict holdout

Non-Water External Phase-1 Packets

Curated external screening datasets for the native methanol, ethanol, acetonitrile, and DMSO carriers.

194
Curated External Cases
0.1428-0.1823
MAE Range (kcal/mol)
4
Published Solvent Packets
0.1639
Mean Solvent MAE (kcal/mol)
Solvent Cases MAE (kcal/mol) RMSE (kcal/mol) R squared Worst Case Abs Error
Methanol970.16360.22490.98800.7772
Ethanol650.14280.19190.99440.6431
Acetonitrile150.16700.22240.93720.5027
DMSO170.18230.22040.97120.3796

These are curated external phase-1 solvent packets. They expand public non-water coverage but are smaller than the water hydration corpus.

Non-Water Large-Scale External Packet

Broader SolProp-enriched external validation across methanol, ethanol, acetonitrile, and DMSO.

436
Total External Cases
0.6155
Weighted MAE (kcal/mol)
0.3103-0.8817
Solvent MAE Range
4
Phase-1 Total Gap Remaining
Solvent Cases MAE (kcal/mol) RMSE (kcal/mol) R squared Max Abs Error
Methanol1610.53081.10770.95437.0171
Ethanol1470.88171.65010.92736.7027
Acetonitrile670.31030.40260.94620.9455
DMSO610.53300.77080.73792.6925

This packet is broader and harder than the curated 194-case subset. It is published separately and should not be conflated with the smaller curated external packet.

Native Non-Water Carrier Holdout

Strict repeated holdout validation on the expanded 436-case SolProp-enriched non-water corpus.

0.5422
Overall MAE (kcal/mol)
436
Native-carrier cases
64
Repeated holdout splits
≤ 1.0081
Worst solvent/category MAE
Solvent Cases Split MAE (kcal/mol) Std Dev Worst Category Worst Category MAE
Methanol1610.46860.0615amide_primary_secondary0.5764
Ethanol1470.79550.1179phenolic_polar_aromatic1.0081
Acetonitrile670.28140.0794general0.3191
DMSO610.37410.0973heavy_halomethane0.4546

Strict repeated holdout remains the main robustness check for the native non-water carriers on the expanded corpus; curated and large-scale external packets are published alongside it as separate views.

Failure Analysis Highlights

Largest remaining error bands used for calibration targeting.

Highest-MAE chemotypes (n >= 10)

Chemotype n MAE (kcal/mol)
halophenol_like110.5812
chlorinated_hydrophobe220.4713
phenolic_polar_aromatic300.4178
monohaloalkane_long170.4095
alkoxy_rich110.4017
hydrophobe360.3756

Top absolute-error cases

Case ID Chemotype Abs Error (kcal/mol)
mobley_5732611general2.4019
mobley_3325209general2.4012
mobley_2763835hydrophobe1.6854
mobley_2751110phenolic_polar_aromatic1.6367
mobley_1178614halophenol_like1.5996

Download benchmark package

Machine-readable and human-readable artifacts for independent review.

Originator: FluxMateria

Benchmark summary JSON
Headline metrics, chemotype breakdown, and outlier cases.
Download JSON
Case-level benchmark CSV
All 642 cases with predicted/reference values and signed errors.
Download CSV
Case-level benchmark JSON
Structured case table for programmatic ingestion.
Download JSON
Methodology note
Benchmark scope, scoring policy, and run configuration.
Download MD
Error breakdown note
Chemotype residual analysis and calibration targets.
Download MD
Non-water holdout summary JSON
Sanitized strict repeated-holdout summary for the native non-water carrier models.
Download JSON
Non-water holdout category CSV
Per-solvent chemotype MAE table for the strict non-water holdout snapshot.
Download CSV
Non-water holdout methodology note
Public methodology summary for the native-carrier strict holdout protocol.
Download MD
Non-water external large-scale summary JSON
Sanitized combined summary for the broader SolProp-enriched non-water external packet.
Download JSON
Non-water external large-scale methodology note
Public methodology summary for the broader SolProp-enriched non-water validation packet.
Download MD
Non-water external large-scale packet manifest
Combined manifest and solvent table for the 436-case large-scale non-water packet.
Download JSON
Non-water external packet summary JSON
Sanitized combined summary for the methanol, ethanol, acetonitrile, and DMSO external phase-1 packets.
Download JSON
Non-water external methodology note
Public methodology summary for the curated external non-water solvent packets.
Download MD
Non-water external packet manifest
Combined packet manifest with artifact hashes for the four external non-water solvent sets.
JSON MD
Packet manifest (JSON + Markdown)
Snapshot metadata, artifact hashes, and benchmark contract.
JSON MD

Need module-level context?

This page is the explicit-solvation benchmark packet. For product context and broader solvation capability coverage, use the module page.

Solvation Module Request Access