Han S., Lee B.G., Lim D.-W. et al. · Chemistry of Materials · 2024 · 11280-11287
Open a family to keep every result attached to its sample, method and conditions.
Paraphrased for this database from the authors’ stated interpretations — never quoted verbatim — and kept separate from reported measurements.
No first-hand MOF synthesis route is reported in the supplied main text or SI; the work curates literature structures/data and performs ML modelling.
Caveat: Original synthesis recipes may exist in the 172 source papers but are not extracted here because they are not first-hand rows in this paper and the raw source set is not supplied.
11281 · Data Extraction and Curation · Figure 1 · Linked to 2 structured results
Randomly splitting entries can give artificially high performance because the same MOF may appear in training and validation/test under different conditions.
Caveat: Main reported evaluation therefore splits by MOF structure.
11283 · Performance Comparison · Figure S3 · Linked to 3 structured results
The model could potentially improve if quantitative guest or water adsorption/coordinated-water information were available.
Caveat: The authors note that solvent removal from experimental structures can obscure coordinated water.
11284 · Performance Comparison · Figure 5g,h · Linked to 2 structured results
The transformer-based transfer-learning freeze model gave the best proton-conductivity prediction performance among the evaluated models, with test MAE 0.91.
Caveat: Prediction error is still approximately one order of magnitude in log conductivity.
11283 · Performance Comparison · Table 1 · Linked to 5 structured results
Guest descriptors and differences in linker connections were identified as important factors for predicting proton conductivity.
Caveat: Importance rankings varied across cross-validation folds and raw feature-importance values are mainly shown graphically.
11284 · Performance Comparison · Figure S6 · Linked to 3 structured results
Temperature and relative humidity are fundamental predictors of proton conductivity, while guest and MOF information improve accuracy.
Caveat: The claim is model-derived and based on a literature data set with heterogeneous experimental protocols.
11284 · Performance Comparison · Figure 5 · Linked to 5 structured results
Names and aliases are kept exactly within the paper’s own identity model.
| Material | Composition | Structure context | Source |
|---|---|---|---|
| IKOTUZ | Not specified | unknown · Model SystemExample MOF/CSD identifier shown in the database workflow and attention-score analysis. | S-12 · Supporting Information · Figure S7 |
| KUXREC | Not specified | unknown · Model SystemExample MOF/CSD identifier included in SI Table S1 and used in attention-score analysis with imidazole guest molecules. | S-2 · Supporting Information · Table S1 |
| proton-conductive MOF database | Not specifiedvarious metals across literature MOFs · various organic linkers across literature MOFs | unknown · Model SystemAggregate set of 248 MOF structural files curated from CSD/reference-code CIFs and literature proton-conductivity reports. | 11282 · Data Extraction and Curation · Figure 2 |
| SI Table S1 named CIF subset | Not specifiedvarious · various | unknown · Model SystemNamed CIF/reference-code subset listed as included in the Supporting Information or previous research. | S-2 · Supporting Information · Table S1 |
Sample form, processing state and composition status define the context for measurements.
| Sample | Form and role | Processing and geometry | Source |
|---|---|---|---|
| IKOTUZ without guest moleculesresearch_0725__mat__mat_ikotuz | Model · Model System · Pristine Framework | Attention-score example for transfer-learning analysis without guest molecules. | S-12 · Supporting Information · Figure S7 |
| KUXREC with imidazole guestresearch_0725__mat__mat_kuxrec | Model · Model System · Guest Loaded | Attention-score example for transfer-learning analysis with imidazole guest molecules. | S-12 · Supporting Information · Figure S7 |
| literature proton-conductivity entriesresearch_0725__mat__mat_proton_mof_database | Unknown · Paper Level Unspecified · Unknown | Conductivity values digitised or extracted from prior experimental MOF papers under varied temperature, relative humidity, and guest conditions. | 11281 · Data Extraction and Curation · Figure 1 |
| curated 248-CIF proton-conductive MOF model/data setresearch_0725__mat__mat_proton_mof_database | Model · Model System · Model | Literature proton-conductivity data linked to curated CIF structures, temperatures, RH values, and guest molecules. | 11281 · Results and Discussion · Figure 1 |