puma

puma estimates gene, exon, and isoform expression levels from Affymetrix 3' arrays and Exon arrays and propagates measurement uncertainties through downstream analyses.

Key Features:

• Gene expression estimation from Affymetrix 3' arrays: Calculates gene expression levels from Affymetrix 3' arrays using refined probe-level analysis to stabilize estimates for low-expression genes.
• Isoform and exon-level quantification from Exon arrays: Quantifies expression across exons and estimates isoform-level expression from Affymetrix Exon arrays (exon microarrays).
• Uncertainty determination and propagation: Determines uncertainties from Affymetrix GeneChip data and propagates these uncertainties through downstream analyses.
• Gamma model for exon arrays: Implements a gamma model for exon arrays that computes isoform and gene expression measurements and associated uncertainties using multi-mapping between probes, isoforms, and genes.
• Efficient differential expression detection: Provides a more computationally efficient method for detecting differential expression compared to previous approaches within the package.
• Model-based clustering with replicate-aware variance: Uses an improved model-based clustering approach that accounts for variance across technical and biological replicates.
• Comprehensive downstream analyses: Supports summarization, differential expression detection, clustering, and principal component analysis (PCA) while incorporating measurement uncertainty.
• Cross-platform applicability: Applies downstream analytical methods to other expression quantification platforms such as RNA-Seq when uncertainty information is available.

Scientific Applications:

• Alternative splicing and isoform-level expression analysis: Enables quantification and comparison of exon and isoform expression to study alternative splicing.
• Gene regulation and function studies: Supports analyses aimed at understanding gene regulation and function at granular resolution by integrating measurement uncertainties.
• Uncertainty-aware cross-platform expression analysis: Facilitates downstream analyses that incorporate uncertainty estimates across platforms (e.g., RNA-Seq) when such uncertainty information is provided.

Methodology:

Computational methods include probe-level analysis; determination and propagation of uncertainties from Affymetrix GeneChip data; a gamma model for exon arrays incorporating multi-mapping between probes, isoforms, and genes; model-based clustering accounting for variance across technical and biological replicates; a computationally efficient differential expression detection method; summarization; and principal component analysis (PCA).