Author Correction: Biophysical prediction of protein–peptide interactions and signaling networks using machine learning

2026-05-21 · Source: Machine learning : nature.com subject feeds · Field: Science & Research — Life Sciences & Biology, Artificial Intelligence & Machine Learning · Depth: Expert, quick

Summary

An author correction has been issued for the article "Biophysical prediction of protein–peptide interactions and signaling networks using machine learning," originally published on January 6, 2020, in Nature Methods. The correction addresses a critical error in the tyrosine kinase dataset used to train the model, which caused the model to inaccurately represent overall kinase binding preferences. The article has been thoroughly updated with a corrected dataset and a newly trained model. Consequently, Figures 1–6, the main text, and the Supplementary information have all been revised. Julia R. Rogers from Columbia University has also been added to the author list, recognizing her significant contributions to constructing the updated dataset, training the new model, designing analyses, performing computational experiments, and revising the manuscript. Both the HTML and PDF versions of the article now reflect these corrections, with the original uncorrected version available as supplementary material.

Key takeaway

For AI Scientists developing or validating machine learning models in biophysical domains, this correction underscores the critical importance of dataset integrity. You must rigorously verify your training data sources, especially for sensitive applications like kinase binding predictions. An identified data error necessitates a complete model retraining and a thorough update of all derived results and figures. Proactively implement robust data validation pipelines to prevent similar inaccuracies from compromising your research findings.

Key insights

Data quality is paramount for accurate machine learning models in biophysical predictions.

Principles

• Data integrity is foundational for model reliability.
• Model performance hinges on accurate training data.
• Comprehensive updates are needed post-data correction.

In practice

• Scrutinize datasets before model training.
• Re-evaluate models upon data source changes.
• Ensure all published artifacts reflect corrections.

Topics

• Machine Learning Models
• Protein-Peptide Interactions
• Biophysical Prediction
• Tyrosine Kinase Dataset
• Data Quality
• Model Retraining

Best for: Research Scientist, AI Scientist

Related on AIssential

• A flaw in using pretrained protein language models in protein–protein interaction inference models — Pretrained protein language models exhibit data leakage in protein-protein interaction inference, inflating performance metrics.
• Tools to Improve Training Data — Data quality issues can render high model accuracy meaningless, necessitating human-in-the-loop tools for robust ML.
• I-SAFE: Wasserstein Coherence Metrics for Structural Auditing of Scientific AI Models — Scientific AI model performance benchmarks are insufficient to confirm alignment with domain-relevant structural behavior.
• A world model for proteins is here — Biohub's open-source protein "world model" (ESMFold2, ESM Atlas) accelerates drug discovery by mapping, predicting, and designing proteins.
• GENEB: Why Genomic Models Are Hard to Compare — Genomic model comparison is fragmented; GENEB provides a unified benchmark revealing that architecture and pretraining often trump scale.
• Author Correction: Generalizable and scalable protein stability prediction with rewired protein generative models — A published article received a correction due to an outdated figure being uploaded.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Machine learning : nature.com subject feeds.