京都大学 京都大学 大学院医学研究科 人間健康科学系専攻 ビッグデータ医科学分野
← Back to Bibliography

HELM-BERT: Topology-Aware Representations for Chemically Modified Peptides

2026.08.03 ORAL
第18回 CBI若手の会講演会 (Online, Student Talk 5)

Seungeon Lee, Takuto Koyama, Itsuki Maeda, Shigeyuki Matsumoto, Yasushi Okuno.

Overview

Student talk at the 18th CBI Wakate-no-Kai Seminar (第18回 CBI若手の会講演会), an online meeting organized by the CBI Society’s early-career researchers’ association under the theme 「越境するキャリアと研究がひらく、創薬の次世代」 (Cross-border careers and research opening the next generation of drug discovery), co-sponsored by Shiga University.

The talk presented the content of the HELM-BERT paper published in Journal of Chemical Information and Modeling: pretraining an encoder-only transformer directly on HELM notation so that monomer identity and covalent topology are represented natively, and the resulting behaviour on cyclic peptide membrane permeability and peptide–protein interaction prediction. The session was followed by discussion with the other participants on representation choices for chemically modified peptides and on where notation-level pretraining is worth applying.

This presentation received the Excellent Lecture Award (優秀講演賞).

Session: 学生発表5 (14:55–15:20)

Affiliation: Kyoto University Graduate School of Medicine, Department of Biomedical Data Intelligence

expand_less