A Comparison of Methods for Classifying Clinical Samples Based on Proteomics Data: A Case Study for Statistical and Machine Learning Approaches
2011

Comparing Methods for Classifying Clinical Samples Using Proteomics Data

publication Evidence: moderate

Author Information

Author(s): Sampson Dayle L., Parker Tony J., Upton Zee, Hurst Cameron P.

Primary Institution: Queensland University of Technology

Hypothesis

Can statistical and machine learning approaches effectively classify clinical samples based on proteomics data?

Conclusion

Both Partial Least Squares (PLS) and Support Vector Machines (SVM) are effective for classifying clinical proteomic datasets, with PLS providing additional interpretability.

Supporting Evidence

  • PLS and SVM demonstrated strong utility for proteomic classification problems.
  • PLS-based classifiers produced models with additional meaningful information.
  • SVMs were the most efficient classifier on most datasets tested.

Takeaway

This study looks at different ways to sort medical samples based on proteins. Some methods work better than others, and one method helps us understand the results better.

Methodology

The study compared statistical dimension reduction techniques (PLS and PCA) and machine learning methods (SVM) for classifying clinical proteomic datasets.

Potential Biases

Potential bias due to the selection of datasets and methods used for classification.

Limitations

The study's findings may not generalize across all datasets due to variability in performance among different classification methods.

Digital Object Identifier (DOI)

10.1371/journal.pone.0024973

Want to read the original?

Access the complete publication on the publisher's website

View Original Publication