mCOPA: analysis of heterogeneous features in cancer expression data

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Cancer outlier profile analysis (COPA) effectively identifies gene fusion events but is limited to over-expressed outliers.
• Existing COPA implementations do not detect under-expressed outliers, restricting comprehensive cancer expression data analysis.
• There is a need for a versatile tool to identify both up- and down-regulated outliers in cancer genomics.

Purpose of the Study:

• To present mCOPA, a modified outlier detection method for cancer expression data.
• To enable the identification of both over- and under-expressed outliers.
• To provide a freely available tool applicable to any expression dataset.

Main Methods:

• Developed mCOPA with refinements to the outlier detection algorithm.
• Applied mCOPA to prostate cancer expression data.
• Integrated outlier analysis with clustering, pathway analysis, and tumor suppressor identification.

Main Results:

• mCOPA identifies both over- and under-expressed outliers, unlike previous methods.
• mCOPA selects more informative features than differential expression or variance-based approaches.
• The method successfully identifies known and novel prostate cancer tumor suppressors and aids in discovering molecular mechanisms of cancer heterogeneity.

Conclusions:

• mCOPA offers advantages over differential expression and variance methods in selecting outlier features for subtype discovery.
• Outlier analysis reveals distinct biological insights compared to differential expression or variance analysis.
• mCOPA is a valuable tool for exploring cancer datasets, discovering new subtypes, and understanding cancer mechanisms.