PERalign

PERalign predicts genomic alignments of transcript fragments from RNA-seq paired-end reads (PERs) and reconstructs splicing paths to improve detection of splice junctions and transcript structures.

Key Features:

• Probabilistic framework: PERalign employs a probabilistic model that considers exonic and spliced alignments of end reads from overlapping PERs to generate possible splicing paths connecting paired ends.
• Expectation-Maximization inference: It maximizes the likelihood of all PER alignments using an expectation maximization (EM) algorithm that assigns likelihood values to splice junctions and identifies the most probable alignments.
• Enhanced coverage and accuracy: On 2 x 35 bp PER datasets from MCF-7 and SUM-102, PERalign increased coverage threefold compared to aligning end reads alone and improved splice detection accuracy.
• qRT-PCR validation: Predicted exon-skipping alternative splicing events were validated by quantitative reverse transcription PCR (qRT-PCR) on eight events.
• Long-range splicing confirmation: PERalign confirmed 8 of 10 fusion events reported by Maher et al., 2009 in the MCF-7 cell line via long-range splicing analysis.

Scientific Applications:

• Cancer genomics: Detection and validation of alternative splicing events and fusion genes in cancer cell lines such as MCF-7 and SUM-102.
• Transcriptome reconstruction: Reconstruction of transcript structures and splice junction discovery from short 2 x 35 bp PER datasets.
• Splicing validation prioritization: Prioritizing splice junctions and exon-skipping events for experimental confirmation such as qRT-PCR.

Methodology:

Probabilistic model that considers exonic and spliced alignments of end reads from overlapping PERs, construction of potential splicing paths connecting paired ends, and maximization of PER alignment likelihoods via an expectation maximization (EM) algorithm assigning likelihood values to splice junctions.