Intrinsic-Extrinsic Convolution and Pooling for Learning on 3D
Protein Structures

k 2021

Intrinsic-Extrinsic Convolution and Pooling for Learning on 3D Protein Structures

HERMOSILLA CASAJÚS, Pedro, Marco SCHÄFER, Matěj LANG, Gloria FACKELMANN, Pere-Pau VÁZQUEZ ALCOCER et. al.

Basic information

Original name

Intrinsic-Extrinsic Convolution and Pooling for Learning on 3D Protein Structures

Authors

HERMOSILLA CASAJÚS, Pedro (724 Spain), Marco SCHÄFER (276 Germany), Matěj LANG (203 Czech Republic, belonging to the institution), Gloria FACKELMANN (276 Germany), Pere-Pau VÁZQUEZ ALCOCER (724 Spain), Barbora KOZLÍKOVÁ (203 Czech Republic, belonging to the institution), Michael KRONE (276 Germany), Tobias RITSCHEL (276 Germany) and Timo ROPINSKI

Edition

International Conference on Learning Representations (ICLR), 2021

Other information

Language

English

Type of outcome

Prezentace na konferencích

Field of Study

10200 1.2 Computer and information sciences

Country of publisher

United States of America

Confidentiality degree

není předmětem státního či obchodního tajemství

References:

Scopus URL

RIV identification code

RIV/00216224:14330/21:00118838

Organization unit

Faculty of Informatics

Keywords in English

classification; bioinformatics

Abstract

V originále

The result is a paper (16 pages) at International Conference on Learning Representations. Although it is among the very best conferences in CS, since its proceedings do not have an ISBN or ISSN, the result cannot be transferred to the RIV database as a result of type D. The original abstract follows: Proteins perform a large variety of functions in living organisms and thus play a key role in biology. However, commonly used algorithms in protein learning were not specifically designed for protein data, and are therefore not able to capture all relevant structural levels of a protein during learning. To fill this gap, we propose two new learning operators, specifically designed to process protein structures. First, we introduce a novel convolution operator that considers the primary, secondary, and tertiary structure of a protein by using n-D convolutions defined on both the Euclidean distance, as well as multiple geodesic distances between the atoms in a multi-graph. Second, we introduce a set of hierarchical pooling operators that enable multi-scale protein analysis. We further evaluate the accuracy of our algorithms on common downstream tasks, where we outperform state-of-the-art protein learning algorithms.

Links

GC18-18647J, research and development project

Name: Vizuální analýza interakcí proteinů a ligandů (Acronym: PROLINT)

Investor: Czech Science Foundation

Citovat

HERMOSILLA CASAJÚS, Pedro, Marco SCHÄFER, Matěj LANG, Gloria FACKELMANN, Pere-Pau VÁZQUEZ ALCOCER, Barbora KOZLÍKOVÁ, Michael KRONE, Tobias RITSCHEL and Timo ROPINSKI. Intrinsic-Extrinsic Convolution and Pooling for Learning on 3D Protein Structures. In International Conference on Learning Representations (ICLR). 2021.

@proceedings{1731463,
   author = {Hermosilla Casajús, Pedro and Schäfer, Marco and Lang, Matěj and Fackelmann, Gloria and Vázquez Alcocer, PereandPau and Kozlíková, Barbora and Krone, Michael and Ritschel, Tobias and Ropinski, Timo},
   booktitle = {International Conference on Learning Representations (ICLR)},
   keywords = {classification; bioinformatics},
   language = {eng},
   title = {Intrinsic-Extrinsic Convolution and Pooling for Learning on 3D Protein Structures},
   url = {https://www.scopus.com/record/display.uri?origin=inward&eid=2-s2.0-85136230434},
   year = {2021}
}

TY  - CONF
ID  - 1731463
AU  - Hermosilla Casajús, Pedro - Schäfer, Marco - Lang, Matěj - Fackelmann, Gloria - Vázquez Alcocer, Pere-Pau - Kozlíková, Barbora - Krone, Michael - Ritschel, Tobias - Ropinski, Timo
PY  - 2021
TI  - Intrinsic-Extrinsic Convolution and Pooling for Learning on 3D Protein Structures
KW  - classification
KW  - bioinformatics
UR  - https://www.scopus.com/record/display.uri?origin=inward&eid=2-s2.0-85136230434
N2  - The result is a paper (16 pages) at International Conference on Learning Representations. Although it is among the very best conferences in CS, since its proceedings do not have an ISBN or ISSN, the result cannot be transferred to the RIV database as a result of type D. The original abstract follows: Proteins perform a large variety of functions in living organisms and thus play a key role in biology. However, commonly used algorithms in protein learning were not specifically designed for protein data, and are therefore not able to capture all relevant structural levels of a protein during learning. To fill this gap, we propose two new learning operators, specifically designed to process protein structures. First, we introduce a novel convolution operator that considers the primary, secondary, and tertiary structure of a protein by using n-D convolutions defined on both the Euclidean distance, as well as multiple geodesic distances between the atoms in a multi-graph. Second, we introduce a set of hierarchical pooling operators that enable multi-scale protein analysis. We further evaluate the accuracy of our algorithms on common downstream tasks, where we outperform state-of-the-art protein learning algorithms.
ER  -

HERMOSILLA CASAJÚS, Pedro, Marco SCHÄFER, Matěj LANG, Gloria FACKELMANN, Pere-Pau VÁZQUEZ ALCOCER, Barbora KOZLÍKOVÁ, Michael KRONE, Tobias RITSCHEL and Timo ROPINSKI. Intrinsic-Extrinsic Convolution and Pooling for Learning on 3D Protein Structures. In \textit{International Conference on Learning Representations (ICLR)}. 2021.

Detailed Information on Publication Record