ScoreGene

ScoreGene evaluates DNA microarray gene expression data by comparing signal quantitation (SQ) algorithms and performing differential expression detection, supervised classification, and semi-supervised gene clustering to assess and improve signal-to-noise characteristics.

Key Features:

• Signal quantitation algorithm comparison: Compares SQ algorithms including dChip, RMAExpress, and Affymetrix MAS5 to determine their effects on downstream analyses.
• Signal-to-noise evaluation: Assesses the impact of SQ algorithms on signal-to-noise ratio within datasets.
• Differential expression detection: Identifies differentially expressed genes across classes using statistical significance tests.
• Supervised classification: Classifies experiments using supervised learning algorithms on processed expression data.
• Semi-supervised gene clustering: Clusters genes with semi-supervised methods to leverage partial annotation or constraints.
• Noise assessment via redundancy: Evaluates noise by leveraging redundancy within experimental datasets to test variability among SQ outputs.
• Signal assessment via overabundance testing: Evaluates signal by testing overabundance of differentially expressed genes using statistical significance tests.
• Benchmarking on repeated hybridizations: Performs comparisons on datasets containing pairs of repeated hybridizations and reports outcomes such as dChip robustness for ~60% of genes and RMAExpress improvements in signal-to-noise ratio for >95% of genes.

Scientific Applications:

• SQ algorithm selection for preprocessing: Guides selection of dChip, RMAExpress, or Affymetrix MAS5 for microarray preprocessing based on SNR and stability metrics.
• Improving differential expression reliability: Informs choice of preprocessing to increase reliability of detected differentially expressed genes.
• Benchmarking preprocessing methods: Provides empirical comparisons to validate and benchmark SQ algorithms on repeated-hybridization datasets.
• Assessing effects on classification and clustering: Evaluates how preprocessing choices affect supervised classification and semi-supervised gene clustering results.

Methodology:

Compares SQ algorithms (dChip, RMAExpress, Affymetrix MAS5) by assessing effects on signal-to-noise ratio using noise assessment via redundancy in experimental datasets and signal assessment via statistical significance tests of overabundant differentially expressed genes, including analyses on datasets of repeated hybridization pairs.