spotSegmentation

spotSegmentation segments microarray image spots and estimates foreground and background intensities to correct for inner holes, artifacts, and blank spots in cDNA microarray analysis.

Key Features:

• Automatic Gridding Method: Performs automatic gridding to align spots on the microarray slide.
• Model-Based Clustering: Applies model-based clustering to pixel intensity distributions using the Bayesian Information Criterion (BIC) to select up to three groups.
• Spatial Component Analysis: Identifies large spatially connected components within each cluster to address inner holes and mitigate artifacts.
• Inferential Blank-Spot Decision: Uses BIC-based model selection to infer blank spots by deciding between one, two, or three groups.
• Histogram and Spatial Integration: Combines histogram-based and spatial analysis to leverage pixel intensity distributions and spatial connectivity.

Scientific Applications:

• HIV cDNA Microarray Analysis: Applied to cDNA microarray images from HIV infection studies to estimate spot intensities.
• Replicate Stability Assessment: Demonstrates superior stability across replicates compared to fixed-circle segmentation and seeded region growing in SPOT software.
• Unbiased Differential Expression Estimation: Estimates intensities of differentially expressed genes without introducing bias.

Methodology:

Performs automatic gridding, model-based clustering of pixel intensities with BIC (implemented via the MCLUST R package) selecting up to three clusters, identification of large spatially connected components per cluster, and integration of histogram-based and spatial approaches.