IsoBayes

IsoBayes performs Bayesian inference on single protein isoforms from liquid chromatography mass spectrometry (MS) data, inferring presence/absence, estimating abundance with uncertainty measures, and identifying isoforms where protein and transcript relative abundances significantly differ.

Key Features:

• Isoform presence/absence inference: Computes the posterior probability that a protein isoform is present in the sample.
• Abundance estimation with uncertainty: Estimates protein isoform abundance and provides a posterior credible interval.
• Multi-data-type compatibility: Accepts both PSM counts and intensities from MS data.
• Transcriptomics integration: Incorporates transcript isoform abundances (TPMs) as an informative prior for protein isoform relative abundance.
• Protein-transcript discordance detection: Identifies isoforms where the relative abundance of proteins and transcripts significantly differ.
• Two-layer latent variable model: Models peptide detection uncertainty and shared peptide ambiguity across multiple protein isoforms.

Scientific Applications:

• Single-isoform proteomics inference: Enables isoform-level analysis rather than gene-level or isoform-group abstraction in bottom-up mass spectrometry proteomics.
• Multi-omic data integration: Combines mass spectrometry proteomics and transcriptomics data in a unified Bayesian framework.
• Peptide detection uncertainty modeling: Accounts for erroneously detected peptides using posterior error probability (PEP).
• Shared peptide allocation: Distributes peptide abundance across multiple compatible protein isoforms.

Methodology:

IsoBayes uses a two-layer latent variable approach. In the first layer, it samples the presence/absence of each peptide based on its posterior error probability (PEP). In the second layer, for peptides estimated as present, it allocates their abundance across the protein isoforms they map to. TPMs are used to formulate an informative prior for protein isoform relative abundance when available.