Studying the effects of correlation on protein selection in proteomics data

Area of Science:

• Statistical genomics and proteomics
• Bioinformatics and computational biology

Background:

• Efron's (2007) methods assess correlation's effect on false discovery rate (FDR) in large-scale testing, particularly for microarray data.
• While FDR procedures don't mandate test independence, correlation can lead to inaccurate estimations of significant findings.
• Accurate FDR assessment is crucial for identifying true discoveries in high-throughput biological experiments.

Purpose of the Study:

• To review Efron's (2007) FDR assessment methods.
• To apply these methods to smaller-scale spectrometry proteomics data.
• To investigate the impact of correlation on FDR and the identification of critical proteins in proteomics.

Main Methods:

• Review of statistical methodologies for FDR assessment.
• Application of Efron's correlation-adjusted FDR methods.
• Analysis of spectrometry proteomics data to evaluate FDR under correlated test statistics.

Main Results:

• Correlation was found to affect FDR values in the analyzed proteomics dataset.
• The number of proteins identified as statistically significant was influenced by the presence of correlation.
• The findings demonstrate the relevance of accounting for correlation even in smaller-scale proteomics studies.

Conclusions:

• Correlation significantly impacts the accuracy of false discovery rate calculations in biological data analysis.
• Efron's methods are applicable and necessary for reliable FDR assessment in proteomics.
• Ignoring correlation can lead to misinterpretation of results, over- or underestimating the number of significant proteins.