What Random Splits Measure in Polymer Glass-Transition Prediction: Chemistry Holdouts, Applicability Domains, and Prior Integration
4 affiliations
- 1College of Transportation, Ludong University, Yantai 264025, China
- 2Analysis and Testing Center, Ludong University, Yantai 264025, China
- 3Graduate School of Convergence Science and Technology, Seoul National University, Seoul 08826, Republic of Korea
- 4National Engineering Laboratory for Advanced Tire Equipment and Key Materials, Qingdao University of Science and Technology, Qingdao 266061, China
Abstract
Predicting the glass-transition temperature (Tg) helps select polymers for a service-temperature window, but accuracy on familiar structures may not transfer to a new family. We compare validation protocols on 7367 curated homopolymers. Histogram gradient boosting gives R2 = 0.899 and mean absolute error (MAE) = 23.9 K under random cross-validation, 0.857 and 28.5 K under scaffold holdout, a class-macro mean of 0.355 and 43.3 K under curator-class holdout, and −3.79 and 79.5 K under high-Tg extrapolation. Similarity explains only 6–11% of within-protocol absolute-error variance in the first three settings. A separate density functional theory (DFT) band-gap dataset shows a moderate fingerprint-cluster holdout penalty and poor property-range extrapolation for the tested baselines. Random-forest split-conformal empirical coverage falls from 0.899 under random splitting to 0.683 and 0.214 under class and property holdout; the tested prospective similarity gate provides little protection. A fold-fitted group-contribution prior is more sensitive to uniform additive bias in a squared-loss penalty than as a neural-network input, but this comparison is architecture-dependent. Class-holdout intervals and tests are descriptive or nominal because training sets overlap. The 279 multiply reported polymers support only the between-source variability analysis. Validation should match the intended polymer family and property range.
Related
- Applicability-Domain-Aware Evaluation of Machine-Learning NMR Chemical-Shift Prediction in Small drug Molecules: GT-NMR for 13C and CASCADE for 1H
- Spatial Applicability of Fictitious Point Feedback for Frequency Response Prediction in Unidirectional Carbon Fiber-Reinforced Polymer Plates
- Defining Applicability Domains for Metal–Ligand Selectivity Prediction: Predictive Uncertainty vs. Molecular Similarity
- Beyond the Ligand Applicability Domain: Generalization and Local Resolution in Structure-Aware Foundation Models for Affinity Prediction
- Robust generative transition-state models for unseen chemistry
BibTeX
@article{han2026what,
title = {{What Random Splits Measure in Polymer Glass-Transition Prediction: Chemistry Holdouts, Applicability Domains, and Prior Integration}},
author = {Deao Han and Yangkun Shao and Xueru Zhong and Zhanzhao Song and Quanzhong Zhang and Chuansheng Wang and Deshang Han and Lin Wang},
journal = {Polymers},
year = {2026},
month = oct,
doi = {10.3390/polym18192445},
url = {https://europepmc.org/article/PMC/PMC13667063},
note = {CC BY},
}
RIS
TY - JOUR
TI - What Random Splits Measure in Polymer Glass-Transition Prediction: Chemistry Holdouts, Applicability Domains, and Prior Integration
AU - Deao Han
AU - Yangkun Shao
AU - Xueru Zhong
AU - Zhanzhao Song
AU - Quanzhong Zhang
AU - Chuansheng Wang
AU - Deshang Han
AU - Lin Wang
PY - 2026
DA - 2026/10/01
JO - Polymers
DO - 10.3390/polym18192445
UR - https://europepmc.org/article/PMC/PMC13667063
AB - Predicting the glass-transition temperature (Tg) helps select polymers for a service-temperature window, but accuracy on familiar structures may not transfer to a new family. We compare validation protocols on 7367 curated homopolymers. Histogram gradient boosting gives R2 = 0.899 and mean absolute error (MAE) = 23.9 K under random cross-validation, 0.857 and 28.5 K under scaffold holdout, a class-macro mean of 0.355 and 43.3 K under curator-class holdout, and −3.79 and 79.5 K under high-Tg extrapolation. Similarity explains only 6–11% of within-protocol absolute-error variance in the first three settings. A separate density functional theory (DFT) band-gap dataset shows a moderate fingerprint-cluster holdout penalty and poor property-range extrapolation for the tested baselines. Random-forest split-conformal empirical coverage falls from 0.899 under random splitting to 0.683 and 0.214 under class and property holdout; the tested prospective similarity gate provides little protection. A fold-fitted group-contribution prior is more sensitive to uniform additive bias in a squared-loss penalty than as a neural-network input, but this comparison is architecture-dependent. Class-holdout intervals and tests are descriptive or nominal because training sets overlap. The 279 multiply reported polymers support only the between-source variability analysis. Validation should match the intended polymer family and property range.
SN - 2073-4360
ER -