Skip to content
DraconiticResearch
Search Sources API
226,742 studies31 sourcesUpdated 34 min ago
Peer-reviewedCC BY 4.0ACS Omega

Pre-Training Synthetic Accessibility Filtering in PyRMD under Transferred Percentile Cutoffs across 95 Target Data Sets

  1. 1Federal University of Amazonas , , , ,
  2. 2Center for Innovation in Artificial Intelligence for Health (CIIA─Saúde/UFMG) , , ,

Abstract

Synthetic accessibility filtering before cross-validation changes both the compounds available for learning and those used to evaluate ligand-based virtual screening models. We assessed SAscore filtering in PyRMD using SA-Filter Impact Benchmarking (SA-FIB), with 95 unfiltered target data sets and 438 filtered conditions across five thresholds. Historical residual-distance percentile cutoffs were transferred, while absolute epsilon values were estimated within each training subset. Mean paired F1 differences were −0.1078, −0.0361, −0.0147, −0.0074, and −0.0011 at SA ≤ 2.5, 3.0, 3.5, 4.0, and 5.0, respectively, relative to the corresponding eligible-cohort baselines. At SA ≤ 5.0, the F1 difference was not significant after Bonferroni correction, while ROC-AUC and PRC-AUC increased by 0.005085 and 0.003060. Combining duplicated processed pools into 92 units retained significance for ROC-AUC but not PRC-AUC. Reoptimizing percentile cutoffs within inner folds for the 92 eligible SA ≤ 3.5 data sets changed mean outer-fold F1 by +0.000294, with 30 improvements, 22 ties, and 40 decreases. Responses varied by target and metric. Historical parameter selection, reused outer folds, and differences between baseline and filtered evaluation populations limit interpretation to sensitivity within this PyRMD protocol. These findings support target-specific validation of synthetic accessibility filters.

Related

  1. Efficiency and Accessibility Are Not Enough: Chemical-Validity Filtering in High-Throughput Screening of Organic Photovoltaic Materials
    Patrick Sorrel Mvoto Kongo, Steve Cabrel Teguia Kouam, Jean-Pierre TCHAPET NJAFA +1
    ChemRxivPreprintCC BYGeneral chemistryChemRxiv
  2. A neural surrogate for the natural-gas compressibility factor: reference-equation accuracy from synthetic training data alone, benchmarked on 2426 independent laboratory measurements
    Emir Borovac
    ChemRxivPreprintLicence not statedGeneral chemistryChemRxiv
  3. Synthetic data training for machine-learning-assisted ARPES experiments
    Meng Xing Na, Chris Zhou, S. K. Y. Dufresne +2
    NewtonPeer-reviewedCC BY-NCFull text6 citedMaterials & polymersOpenAlex
  4. Beyond Moisture Content: Time-Domain NMR for Quantifying Molecular Mobility, Water Accessibility and Food Function
    Molecules (Basel, Switzerland)Peer-reviewedCC BYFull textGeneral chemistryEurope PMC
  5. From Coprecipitation to Nanostructure: Formation, Morphology, and Reactive-Site Accessibility in Fe-Containing Binary Layered Double Hydroxides
    Alua Alikeyeva, Nurbolat Kudaibergenov, Kairzhan Shalmagambetov +1
    Nanomaterials (Basel, Switzerland)Peer-reviewedCC BYFull textMaterials & polymersEurope PMC
BibTeX
@article{ferreira2026pretraining,
  title = {{Pre-Training Synthetic Accessibility Filtering in PyRMD under Transferred Percentile Cutoffs across 95 Target Data Sets}},
  author = {Fabio Jorge de Nazare Ferreira and Tallita Marques Machado and Fernanda Guilhon-Simplicio},
  journal = {ACS Omega},
  year = {2026},
  month = nov,
  doi = {10.1021/acsomega.6c10178},
  url = {https://pubs.acs.org/acsodf/article/doi/10.1021/acsomega.6c10178/5471111/Pre-Training-Synthetic-Accessibility-Filtering-in},
  note = {CC BY 4.0},
}
RIS
TY  - JOUR
TI  - Pre-Training Synthetic Accessibility Filtering in PyRMD under Transferred Percentile Cutoffs across 95 Target Data Sets
AU  - Fabio Jorge de Nazare Ferreira
AU  - Tallita Marques Machado
AU  - Fernanda Guilhon-Simplicio
PY  - 2026
DA  - 2026/11/16
JO  - ACS Omega
DO  - 10.1021/acsomega.6c10178
UR  - https://pubs.acs.org/acsodf/article/doi/10.1021/acsomega.6c10178/5471111/Pre-Training-Synthetic-Accessibility-Filtering-in
AB  - Synthetic accessibility filtering before cross-validation changes both the compounds available for learning and those used to evaluate ligand-based virtual screening models. We assessed SAscore filtering in PyRMD using SA-Filter Impact Benchmarking (SA-FIB), with 95 unfiltered target data sets and 438 filtered conditions across five thresholds. Historical residual-distance percentile cutoffs were transferred, while absolute epsilon values were estimated within each training subset. Mean paired F1 differences were −0.1078, −0.0361, −0.0147, −0.0074, and −0.0011 at SA ≤ 2.5, 3.0, 3.5, 4.0, and 5.0, respectively, relative to the corresponding eligible-cohort baselines. At SA ≤ 5.0, the F1 difference was not significant after Bonferroni correction, while ROC-AUC and PRC-AUC increased by 0.005085 and 0.003060. Combining duplicated processed pools into 92 units retained significance for ROC-AUC but not PRC-AUC. Reoptimizing percentile cutoffs within inner folds for the 92 eligible SA ≤ 3.5 data sets changed mean outer-fold F1 by +0.000294, with 30 improvements, 22 ties, and 40 decreases. Responses varied by target and metric. Historical parameter selection, reused outer folds, and differences between baseline and filtered evaluation populations limit interpretation to sensitivity within this PyRMD protocol. These findings support target-specific validation of synthetic accessibility filters.
SN  - 2470-1343
ER  -