SGSeq

SGSeq identifies and quantifies splice events from RNA-seq data, including unannotated and complex splicing patterns, to construct genome-wide splice graphs and measure splice variant usage.

Key Features:

• Genome-Guided Prediction: Predicts splice junctions and exons by mapping RNA-seq reads to a reference genome for downstream assembly.
• Splice Graph Construction: Generates genome-wide splice graphs from existing annotations or predicts them de novo directly from RNA-seq data.
• Recursive Event Identification: Identifies splice events recursively within the constructed splice graph to capture complex splicing patterns.
• Local Quantification: Quantifies splice event usage locally by analyzing reads that extend across the start or end points of each splice variant.
• Aligner Impact Assessment: Evaluates the influence of read aligners GSNAP, HISAT2, STAR, and TopHat2 on prediction accuracy.
• Comparative Quantification Validation: Validates quantification using simulated data and compares results with methods such as MISO and Cufflinks.

Scientific Applications:

• Analysis of Complex Splicing Events: Detects complex and unannotated splice events for detailed characterization of alternative splicing.
• Validation Across Datasets and Tissues: Validated using simulated data and RNA-seq from 16 normal human tissues to assess prediction and quantification performance.
• Discovery of Novel Internal Exons: Identified 249 internal exons within known genes in the Illumina Body Map 2.0 dataset, with validation by RT-PCR and paired-end RNA-seq in independent samples.

Methodology:

Reads are mapped to a reference genome to predict junctions and exons and assemble splice graphs (from annotations or de novo), splice events are identified recursively and quantified using reads spanning variant boundaries, and prediction/quantification performance is assessed with different aligners (GSNAP, HISAT2, STAR, TopHat2) and compared to MISO and Cufflinks using simulated data.