DeConveil 0.2.0

DeConveil 0.2.0

This release introduces major methodological extensions to DeConveil, including:

• an updated multiple testing strategy,
• simulation utilities, and
• a complementary NB regression with Bayesian inference framework.

New features

• Revised multiple testing/FDR control strategy
  Implemented single FDR control procedure using the R package stageR (via rpy2) and replaced the previous two-stage independent correction with a single Benjamini–Hochberg (BH) adjustment.
• Simulation utilities
  Added functions to simulate dosage-dependent differential gene expression under different copy-number scenarios.
• Complementary Negative Binomial regression (Stan-based)
  Introduced an optional Bayesian NB regression model implemented in Stan and accessed via cmdstanpy.
  This model is applied to tumor samples only and can be used independently to test dosage sensitivity and dosage compensation.

Dependency updates

• rpy2 is now a required dependency to support single FDR control via stageR.
• Stan support is optional and can be enabled with:

  pip install "deconveil[stan]"