subgraphquant

subgraphquant computes confidence ranges for transcript abundances to assess and quantify non-identifiability in RNA-seq transcript expression quantification.

Key Features:

• Graph Quantification: Generalizes traditional transcript quantification by accounting for incomplete reference transcriptomes and allowing unannotated transcripts to explain observed RNA-seq reads.
• Confidence Range Calculation: Computes a "confidence range of expression" for each transcript representing possible abundance levels across equally optimal estimates.
• Error Analysis: Application to the Human Body Map data indicates that 35%–50% of transcripts may have inaccurate quantification caused by non-identifiability.
• Isoform Comparison: Quantifies that 20%–47% inaccuracy can prevent reliable ranking of expression levels between a transcript and its sibling isoforms within a sample.
• Differential Expression Analysis: Identifies which detected differentially expressed transcripts across RNA-seq samples remain reliable and highlights exceptions affected by non-identifiability.

Scientific Applications:

• Reliability Assessment: Assess the reliability of transcript quantification in the presence of non-identifiable estimates.
• Isoform Expression Comparison: Compare and evaluate the ranking of isoform expression levels within samples or across sample groups while accounting for non-identifiability.
• Differential Expression Validation: Validate differential expression results across RNA-seq samples against potential errors introduced by non-identifiability.

Methodology:

Generalizes transcript quantification to a graph-based formulation that permits unannotated transcripts and computes per-transcript confidence ranges representing possible abundances across equally optimal estimates; applied to datasets such as the Human Body Map to quantify the fraction of affected transcripts.