PennSeq

PennSeq estimates isoform-specific gene expression from RNA sequencing (RNA-Seq) data using a non-parametric statistical model that accounts for isoform-specific, non-uniform read distributions and sequencing biases.

Key Features:

• Non-parametric modeling: Employs a non-parametric statistical approach that avoids parametric assumptions about read distributions.
• Isoform-specific read distributions: Allows each isoform to have its own non-uniform read distribution.
• Bias accommodation: Accounts for hexamer priming bias, local sequence bias, positional bias, RNA degradation, mapping bias, and other unknown factors.
• Empirical fragment sampling probabilities: Estimates the probability that a fragment is sampled from particular regions based on aligned RNA-Seq data.
• Validation datasets: Evaluated on simulated datasets with known ground truth and on real Illumina RNA-Seq datasets including qRT-PCR measurements.
• Performance advantage: Provides improved isoform-expression estimation relative to existing methods, particularly under severe non-uniformity.

Scientific Applications:

• Isoform-level quantification: Estimating isoform-specific gene expression from RNA-Seq data.
• Bias-sensitive analyses: Improving accuracy of expression estimates in studies affected by sequencing and sample biases.
• Disease-associated gene studies: Enabling more accurate identification of genes associated with disease susceptibility through better isoform quantification.
• Method benchmarking: Benchmarking expression-estimation methods using simulated ground truth and qRT-PCR validation.

Methodology:

PennSeq models isoform-specific, non-uniform read distributions using a non-parametric statistical approach and estimates empirical fragment sampling probabilities from aligned RNA-Seq data; performance was evaluated on simulated datasets and Illumina RNA-Seq datasets with qRT-PCR measurements.