iexcerno

iexcerno filters formalin‑fixed paraffin‑embedded (FFPE)‑induced C>T sequencing artifacts from next‑generation sequencing cancer genomics data to improve mutation‑calling accuracy and downstream analyses.

Key Features:

• Artifact Identification: Leverages the FFPE‑associated mutational signature to identify and score FFPE‑induced artifacts, particularly C>T point mutations.
• Bayesian Approach: Applies Bayes' formula to calculate the probability that a detected variant is an FFPE artifact for probabilistic filtering.
• Performance Dependence on Signature Similarity: Quantifies how sensitivity and specificity vary with the cosine similarity between baseline mutational signatures and the FFPE signature.
• Predictive Modeling: Uses a linear model with an interaction term to predict specificity and sensitivity, reporting R² values of 0.84 for specificity and 0.79 for sensitivity.
• Validation: Validated using simulated mutations combining FFPE‑specific mutations with each of the 60 COSMIC baseline mutational signatures and with real RNA sequencing data from six cancer samples.

Scientific Applications:

• Accurate mutation detection in FFPE samples: Reduces FFPE‑induced false positives in cancer genomics studies to improve reliability of variant calls from FFPE tissue.
• Support for precision oncology: Improves the quality of genomic data used for tumor biology interpretation and treatment decision‑making.

Methodology:

Uses the FFPE mutational signature, Bayes' formula for artifact probability scoring, cosine similarity to assess impact on sensitivity/specificity, a linear model with an interaction term to predict performance (R² reported), and validation via simulations across 60 COSMIC signatures plus FFPE mutations and RNA‑seq data from six cancer samples.