2024
Revealing data leakage in protein interaction benchmarks
BUSHUIEV, Anton, Roman BUSHUIEV, Jiri SEDLAR, Tomas PLUSKAL, Jiří DAMBORSKÝ et. al.Základní údaje
Originální název
Revealing data leakage in protein interaction benchmarks
Autoři
BUSHUIEV, Anton, Roman BUSHUIEV, Jiri SEDLAR, Tomas PLUSKAL, Jiří DAMBORSKÝ, Stanislav MAZURENKO a Josef SIVIC
Vydání
ICLR 2024 Workshop on Generative and Experimental Perspectives for Biomolecular Design, 13 s. 2024
Další údaje
Jazyk
angličtina
Typ výsledku
Stať ve sborníku
Obor
10608 Biochemistry and molecular biology
Utajení
není předmětem státního či obchodního tajemství
Organizační jednotka
Přírodovědecká fakulta
Příznaky
Mezinárodní význam, Recenzováno
Změněno: 24. 4. 2024 10:52, Mgr. Marie Šípková, DiS.
Anotace
V originále
In recent years, there has been remarkable progress in machine learning for protein-protein interactions. However, prior work has predominantly focused on improving learning algorithms, with less attention paid to evaluation strategies and data preparation. Here, we demonstrate that further development of machine learning methods may be hindered by the quality of existing train-test splits. Specifically, we find that commonly used splitting strategies for protein complexes, based on protein sequence or metadata similarity, introduce major data leakage. This may result in overoptimistic evaluation of generalization, as well as unfair benchmarking of the models, biased towards assessing their overfitting capacity rather than practical utility. To overcome the data leakage, we recommend constructing data splits based on 3D structural similarity of protein-protein interfaces and suggest corresponding algorithms. We believe that addressing the data leakage problem is critical for further progress in this research area.
Návaznosti
EF17_043/0009632, projekt VaV |
| ||
LM2023055, projekt VaV |
| ||
LM2023069, projekt VaV |
| ||
857560, interní kód MU (Kód CEP: EF17_043/0009632) |
| ||
90254, velká výzkumná infrastruktura |
|