BALBOA

BALBOA assigns functional annotations to unclassified open reading frames (ORFs) by applying semi-supervised bicluster analysis to gene expression (microarray) data.

Key Features:

• Bicluster Analysis: Employs bicluster analysis (an unsupervised classification technique) to identify subsets of samples and genes with coherent expression patterns.
• Semi-Supervised Approach: Integrates existing ORF annotations into the biclustering framework to create a semi-supervised model combining annotated and unannotated data.
• Classifier Construction and Application: Constructs bicluster-based classifiers from a labeled training set of annotated ORFs and applies them to predict functions of unlabeled ORFs.
• Validation and Benchmarking: Evaluates classifier performance via cross-validation across three independent gene expression datasets and benchmarks against a multi-class k-Nearest Neighbour (kNN) classifier.
• Functional Annotation and Novel Module Prediction: Annotates unclassified yeast ORFs and predicts novel functional modules, with validation using experimental and protein sequence information.

Scientific Applications:

• ORF Functional Annotation: Assigns functional labels to unclassified ORFs using microarray expression patterns.
• Novel Module Discovery: Predicts novel co-regulated functional modules in yeast and proposes module membership for ORFs.
• Cross-Organism Expression Analysis: Applies the semi-supervised biclustering approach to other organisms where microarray gene expression data are available.
• Method Comparison: Provides comparative evaluation against supervised classifiers such as multi-class k-Nearest Neighbour (kNN).

Methodology:

Performs bicluster analysis to detect correlated expression subsets, integrates annotations to form a semi-supervised model, constructs bicluster classifiers from labeled annotated ORFs, applies classifiers to unlabeled ORFs, assesses performance by cross-validation across three independent gene expression datasets, benchmarks against a multi-class k-Nearest Neighbour (kNN) classifier, and validates predictions with experimental and protein sequence information.