P‐43: Multi‐Modal Pre‐training Framework for Molecular Property Prediction
SID Symposium Digest of Technical Papers · 2026-06 · 57 (1) · 1838-1840
DOI: 10.1002/sdtp.70001_499 ↗Kurz gefasst
Fachartikel in der Zeitschrift SID Symposium Digest of Technical Papers, online erschienen am 11.09.2026. Verfasst von Min Zeng und 4 weiteren Autor:innen. Die Arbeit nennt 14 Literaturangaben.
Abstract (Originalsprache)
Traditional AI‐driven molecular property prediction methods often overlook the 3D stereochemical nature of molecules, leading to limited accuracy of molecular property prediction Here, we propose a multi‐modal pre‐training framework that integrates 3D molecular structure with 2D graphs and 1D SMILES strings to capture comprehensive structure‐property correlations. During fine‐tuning only SMILES are used to ensure computational efficiency. Evaluated on a curated dataset, our model achieves remarkable precision in predicting HOMO, LUMO, S₁ and T₁. This framework effectively balances accuracy and efficiency, offering a generalizable paradigm for high‐throughput molecular screening and property prediction
Zitieren
APA
Zeng, M., Liu, T., Huai, L., Liu, Y., & Jiang, X. (2026). P‐43: Multi‐Modal Pre‐training Framework for Molecular Property Prediction. SID Symposium Digest of Technical Papers, 57(1), 1838-1840. https://doi.org/10.1002/sdtp.70001_499
Vancouver
Zeng M, Liu T, Huai L, Liu Y, Jiang X. P‐43: Multi‐Modal Pre‐training Framework for Molecular Property Prediction. SID Symposium Digest of Technical Papers. 2026;57(1):1838-1840. doi:10.1002/sdtp.70001_499
BibTeX
@article{Zeng2026,
author = {Zeng, Min and Liu, Tong and Huai, Lian and Liu, Yuyu and Jiang, Xingqun},
title = {P‐43: Multi‐Modal Pre‐training Framework for Molecular Property Prediction},
journal = {SID Symposium Digest of Technical Papers},
year = {2026},
volume = {57},
number = {1},
pages = {1838--1840},
doi = {10.1002/sdtp.70001_499}
}
EUVIMED