A community effort to identify and correct mislabeled samples in proteogenomic studies

Area of Science:

• Genomics
• Proteomics
• Bioinformatics
• Computational Biology

Background:

• Sample mislabeling and misannotation are persistent challenges in scientific research.
• These errors are particularly common in complex, large-scale multi-omic studies.
• Automated quality control methods are urgently needed to detect and correct sample mix-ups in multi-omic data.

Purpose of the Study:

• To establish a framework for evaluating methods to identify and correct sample mislabels in integrative proteogenomic studies.
• To benchmark the performance of various computational approaches for detecting sample mislabeling.
• To foster the development of robust solutions for multi-omic data quality control.

Main Methods:

• A crowdsourced challenge (precisionFDA NCI-CPTAC Multi-omics Enabled Sample Mislabeling Correction Challenge) was organized.
• Data scientists submitted diverse methods for mislabel identification and correction.
• Performance of submitted methods was systematically evaluated on simulated and real multi-omic datasets.

Main Results:

• The challenge attracted numerous submissions from global participants.
• Performance varied significantly across the submitted mislabeling correction methods.
• Post-challenge collaboration resulted in the development of the open-source software COSMO.

Conclusions:

• COSMO demonstrates high accuracy and robustness in identifying and correcting sample mislabels.
• The developed software provides a valuable tool for quality control in proteogenomic studies.
• This work highlights the effectiveness of crowdsourced challenges in advancing multi-omic data integrity.