SWISS MADE: Standardized WithIn Class Sum of Squares to evaluate methodologies and dataset elements

Area of Science:

• Genomics and Bioinformatics
• Statistical Methods in Biology

Background:

• High-dimensional biological assays like mRNA microarrays involve complex data processing.
• Evaluating the impact of these processing steps on biological interpretation is often overlooked.
• Researchers lack straightforward methods to compare diverse data processing techniques.

Purpose of the Study:

• To introduce Standardized WithIn class Sum of Squares (SWISS), a statistical tool for comparing data processing methods.
• To enable researchers to assess how different methods cluster biological classes.
• To provide a quantitative approach for selecting optimal experimental and computational workflows.

Main Methods:

• Developed SWISS, a tool utilizing Euclidean distance to measure clustering performance.
• Applied SWISS to compare experimental methods, normalization techniques, and gene sets across four datasets.
• Utilized SWISS to compare microarray platforms using MicroArray Quality Control (MAQC) data.
• Employed SWISS to compare different technologies, specifically Agilent two-color microarrays and RNA-Seq.

Main Results:

• SWISS demonstrated its utility in comparing diverse gene expression data processing strategies.
• Analysis of MAQC data revealed platform-specific inter-site reproducibility differences.
• SWISS indicated that one lane of RNA-Seq achieves comparable data clustering by biological phenotype to a single Agilent two-color microarray.

Conclusions:

• SWISS offers a valuable method for evaluating and selecting optimal data processing pipelines in high-dimensional biology.
• The tool aids in understanding the influence of experimental design and technology choice on biological conclusions.
• SWISS provides insights into platform performance and the comparative capabilities of emerging technologies like RNA-Seq.