Summary of Revealing Data Leakage in Protein Interaction Benchmarks, by Anton Bushuiev et al.
Revealing data leakage in protein interaction benchmarks
by Anton Bushuiev, Roman Bushuiev, Jiri Sedlar, Tomas Pluskal, Jiri Damborsky, Stanislav Mazurenko, Josef Sivic
First submitted to arxiv on: 16 Apr 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: None
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary This paper investigates machine learning methods for predicting protein-protein interactions. Current approaches have focused on improving learning algorithms, but neglected evaluation strategies and data preparation. The authors reveal that commonly used splitting strategies introduce data leakage, leading to overoptimistic evaluations of model generalization and unfair benchmarking. To overcome this issue, the researchers propose constructing data splits based on 3D structural similarity of protein-protein interfaces and suggest corresponding algorithms. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Protein-protein interactions are important for many biological processes. Scientists have made progress in using machine learning to predict these interactions, but there’s a problem with how they evaluate their methods. Right now, people split their data into training and testing sets based on how similar the proteins are or what they do. However, this approach can make it seem like their models are better at predicting interactions than they really are. To fix this issue, researchers suggest creating new ways to split the data based on how the proteins fit together in space. This could help them develop more accurate and useful models for predicting protein-protein interactions. |
Keywords
» Artificial intelligence » Generalization » Machine learning
