Automated Prediction of Protein Pair Distances Based on Deep Learning: A Novel Approach to Protein Structure Prediction in SCOP

Duo Feng; Shuaicheng Li; Yue Zhang

doi:10.1007/978-981-95-0695-8_33

Automated Prediction of Protein Pair Distances Based on Deep Learning: A Novel Approach to Protein Structure Prediction in SCOP

Duo Feng
, Shuaicheng Li
, Yue Zhang^*

^*Corresponding author for this work

City University of Hong Kong

Research output: Chapters, Conference Papers, Creative and Literary Works › RGC 32 - Refereed conference paper (with host publication) › peer-review

Abstract

This study introduces the innovative Protein Distance Model (PDM) for predicting structural distances between protein pairs in the Structural Classification of Proteins (SCOP) database. SCOP’s hierarchical classification presents challenges for direct prediction, so PDM predicts distances between unknown proteins and those with known classifications to infer structural categories. Incorporating NLP’s Attention mechanism and a dual-tower network, PDM extracts high-level features from two protein sequences to predict their distance. Transitioning from regression to a multiclass task improved performance, achieving over 90% AUC and Accuracy in binary classification and over 80% F1-Score and Accuracy in multiclass scenarios. PDM shows strong generalization, laying a foundation for future research in protein structural classification. The code is publicly available on GitHub. © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2026.

Original language	English
Title of host publication	Bioinformatics Research and Applications
Subtitle of host publication	21st International Symposium, ISBRA 2025, Helsinki, Finland, August 3–5, 2025, Proceedings, Part II
Editors	Jing Tang, Xin Lai, Zhipeng Cai, Wei Peng, Yanjie Wei
Place of Publication	Singapore
Publisher	Springer
Pages	411-426
Number of pages	16
ISBN (Electronic)	978-981-95-0695-8
ISBN (Print)	978-981-95-0694-1
DOIs	https://doi.org/10.1007/978-981-95-0695-8_33
Publication status	Published - 2026
Event	21st International Symposium on Bioinformatics Research and Applications (ISBRA 2025) - University of Helsinki, Helsinki, Finland Duration: 3 Aug 2025 → 5 Aug 2025 https://www.helsinki.fi/en/conferences/isbra2025

Publication series

Name	Lecture Notes in Computer Science (including subseries Lecture Notes in Bioinformatics)
Volume	15757
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Conference

Conference	21st International Symposium on Bioinformatics Research and Applications (ISBRA 2025)
Abbreviated title	ISBRA2025
Place	Finland
City	Helsinki
Period	3/08/25 → 5/08/25
Internet address	https://www.helsinki.fi/en/conferences/isbra2025

Research Keywords

Attention mechanism (NLP)
Deep Learning
Dual-tower network
Hierarchical classification
Protein Structure
SCOP

Access Output

10.1007/978-981-95-0695-8_33

Other Access Options

Check CityUHK Library

Permanent Link

Right click to copy link

Cite this

Feng, D., Li, S., & Zhang, Y. (2026). Automated Prediction of Protein Pair Distances Based on Deep Learning: A Novel Approach to Protein Structure Prediction in SCOP. In J. Tang, X. Lai, Z. Cai, W. Peng, & Y. Wei (Eds.), Bioinformatics Research and Applications: 21st International Symposium, ISBRA 2025, Helsinki, Finland, August 3–5, 2025, Proceedings, Part II (pp. 411-426). (Lecture Notes in Computer Science (including subseries Lecture Notes in Bioinformatics); Vol. 15757). Springer . https://doi.org/10.1007/978-981-95-0695-8_33

Feng, Duo ; Li, Shuaicheng ; Zhang, Yue. / Automated Prediction of Protein Pair Distances Based on Deep Learning : A Novel Approach to Protein Structure Prediction in SCOP. Bioinformatics Research and Applications: 21st International Symposium, ISBRA 2025, Helsinki, Finland, August 3–5, 2025, Proceedings, Part II. editor / Jing Tang ; Xin Lai ; Zhipeng Cai ; Wei Peng ; Yanjie Wei. Singapore : Springer , 2026. pp. 411-426 (Lecture Notes in Computer Science (including subseries Lecture Notes in Bioinformatics)).

@inproceedings{e0c85e7592ef45928f28dcebbd620afe,

title = "Automated Prediction of Protein Pair Distances Based on Deep Learning: A Novel Approach to Protein Structure Prediction in SCOP",

abstract = "This study introduces the innovative Protein Distance Model (PDM) for predicting structural distances between protein pairs in the Structural Classification of Proteins (SCOP) database. SCOP{\textquoteright}s hierarchical classification presents challenges for direct prediction, so PDM predicts distances between unknown proteins and those with known classifications to infer structural categories. Incorporating NLP{\textquoteright}s Attention mechanism and a dual-tower network, PDM extracts high-level features from two protein sequences to predict their distance. Transitioning from regression to a multiclass task improved performance, achieving over 90\% AUC and Accuracy in binary classification and over 80\% F1-Score and Accuracy in multiclass scenarios. PDM shows strong generalization, laying a foundation for future research in protein structural classification. The code is publicly available on GitHub. {\textcopyright} The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2026.",

keywords = "Attention mechanism (NLP), Deep Learning, Dual-tower network, Hierarchical classification, Protein Structure, SCOP",

author = "Duo Feng and Shuaicheng Li and Yue Zhang",

year = "2026",

doi = "10.1007/978-981-95-0695-8\_33",

language = "English",

isbn = "978-981-95-0694-1",

series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Bioinformatics)",

publisher = "Springer ",

pages = "411--426",

editor = "Jing Tang and Xin Lai and Zhipeng Cai and Wei Peng and Yanjie Wei",

booktitle = "Bioinformatics Research and Applications",

note = "21st International Symposium on Bioinformatics Research and Applications (ISBRA 2025), ISBRA2025 ; Conference date: 03-08-2025 Through 05-08-2025",

url = "https://www.helsinki.fi/en/conferences/isbra2025",

}

Feng, D, Li, S & Zhang, Y 2026, Automated Prediction of Protein Pair Distances Based on Deep Learning: A Novel Approach to Protein Structure Prediction in SCOP. in J Tang, X Lai, Z Cai, W Peng & Y Wei (eds), Bioinformatics Research and Applications: 21st International Symposium, ISBRA 2025, Helsinki, Finland, August 3–5, 2025, Proceedings, Part II. Lecture Notes in Computer Science (including subseries Lecture Notes in Bioinformatics), vol. 15757, Springer , Singapore, pp. 411-426, 21st International Symposium on Bioinformatics Research and Applications (ISBRA 2025), Helsinki, Finland, 3/08/25. https://doi.org/10.1007/978-981-95-0695-8_33

Automated Prediction of Protein Pair Distances Based on Deep Learning: A Novel Approach to Protein Structure Prediction in SCOP. / Feng, Duo; Li, Shuaicheng; Zhang, Yue.
Bioinformatics Research and Applications: 21st International Symposium, ISBRA 2025, Helsinki, Finland, August 3–5, 2025, Proceedings, Part II. ed. / Jing Tang; Xin Lai; Zhipeng Cai; Wei Peng; Yanjie Wei. Singapore: Springer , 2026. p. 411-426 (Lecture Notes in Computer Science (including subseries Lecture Notes in Bioinformatics); Vol. 15757).

Research output: Chapters, Conference Papers, Creative and Literary Works › RGC 32 - Refereed conference paper (with host publication) › peer-review

TY - GEN

T1 - Automated Prediction of Protein Pair Distances Based on Deep Learning

T2 - 21st International Symposium on Bioinformatics Research and Applications (ISBRA 2025)

AU - Feng, Duo

AU - Li, Shuaicheng

AU - Zhang, Yue

PY - 2026

Y1 - 2026

N2 - This study introduces the innovative Protein Distance Model (PDM) for predicting structural distances between protein pairs in the Structural Classification of Proteins (SCOP) database. SCOP’s hierarchical classification presents challenges for direct prediction, so PDM predicts distances between unknown proteins and those with known classifications to infer structural categories. Incorporating NLP’s Attention mechanism and a dual-tower network, PDM extracts high-level features from two protein sequences to predict their distance. Transitioning from regression to a multiclass task improved performance, achieving over 90% AUC and Accuracy in binary classification and over 80% F1-Score and Accuracy in multiclass scenarios. PDM shows strong generalization, laying a foundation for future research in protein structural classification. The code is publicly available on GitHub. © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2026.

AB - This study introduces the innovative Protein Distance Model (PDM) for predicting structural distances between protein pairs in the Structural Classification of Proteins (SCOP) database. SCOP’s hierarchical classification presents challenges for direct prediction, so PDM predicts distances between unknown proteins and those with known classifications to infer structural categories. Incorporating NLP’s Attention mechanism and a dual-tower network, PDM extracts high-level features from two protein sequences to predict their distance. Transitioning from regression to a multiclass task improved performance, achieving over 90% AUC and Accuracy in binary classification and over 80% F1-Score and Accuracy in multiclass scenarios. PDM shows strong generalization, laying a foundation for future research in protein structural classification. The code is publicly available on GitHub. © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2026.

KW - Attention mechanism (NLP)

KW - Deep Learning

KW - Dual-tower network

KW - Hierarchical classification

KW - Protein Structure

KW - SCOP

UR - https://www.scopus.com/pages/publications/105013462397

UR - https://www.scopus.com/pages/publications/105013462397#tab=impact

U2 - 10.1007/978-981-95-0695-8_33

DO - 10.1007/978-981-95-0695-8_33

M3 - RGC 32 - Refereed conference paper (with host publication)

SN - 978-981-95-0694-1

T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Bioinformatics)

SP - 411

EP - 426

BT - Bioinformatics Research and Applications

A2 - Tang, Jing

A2 - Lai, Xin

A2 - Cai, Zhipeng

A2 - Peng, Wei

A2 - Wei, Yanjie

PB - Springer

CY - Singapore

Y2 - 3 August 2025 through 5 August 2025

ER -

Feng D, Li S, Zhang Y. Automated Prediction of Protein Pair Distances Based on Deep Learning: A Novel Approach to Protein Structure Prediction in SCOP. In Tang J, Lai X, Cai Z, Peng W, Wei Y, editors, Bioinformatics Research and Applications: 21st International Symposium, ISBRA 2025, Helsinki, Finland, August 3–5, 2025, Proceedings, Part II. Singapore: Springer . 2026. p. 411-426. (Lecture Notes in Computer Science (including subseries Lecture Notes in Bioinformatics)). Epub 2025 Aug 1. doi: 10.1007/978-981-95-0695-8_33

Automated Prediction of Protein Pair Distances Based on Deep Learning: A Novel Approach to Protein Structure Prediction in SCOP

Abstract

Publication series

Conference

Research Keywords

Access Output

Other Access Options

Permanent Link

Other Files and Links

Fingerprint

Cite this