Figure 1

Workflow of the AS detection method and exemplary gene with observed hybridization profile. (A) Workflow of the supervised two-stage AS detection method. In the first stage, SVMs are employed to classify a given gene segment as an exon or intron, based on the hybridization signal observed in a single sample. The resulting SVM scores are transformed to probabilistic confidences to estimate the probability that this segment is included in mature mRNA transcripts. In the second stage, these inclusion probabilities are combined across samples by another SVM classifier to detect alternative splicing. (B) Example gene with hybridization data. The figure illustrates the annotated structure and measured hybridization data for the anti-fungal resistance gene RLM3 (AT4G16990), which was among the top 1% in our genome-wide screen. The logarithmized probe intensities for different samples are plotted against the relative genomic positions of the probes. The highlighted intron shows a characteristic hybridization pattern for differential alternative splicing.