bayesynergy

bayesynergy models in-vitro drug combination dose-response data using a Bayesian semi-parametric framework to quantify synergistic and antagonistic interactions.

Key Features:

• Probabilistic Modelling: Uses a probabilistic framework to represent drug combination experiments and capture synergistic and antagonistic interactions.
• Flexible Interaction Modelling: Models the interaction effect with a Gaussian process formulation to flexibly represent additional combination effects beyond single-agent responses.
• Statistical Rigor: Employs Bayesian methods that explicitly incorporate replicates, handle missing data, and accommodate uneven concentration grids.
• Uncertainty Quantification: Provides posterior-based uncertainty quantification for inferred effects and interaction estimates.
• Stan Implementation: Implemented using the Stan programming language for Bayesian computation.
• Efficient Sampling: Utilizes a computationally efficient sampler for Bayesian inference.
• Variational Inference: Offers fast approximation of the posterior distribution through variational inference techniques.
• Parallel Processing: Supports parallel processing capabilities for large-scale drug combination screens.

Scientific Applications:

• Pre-clinical cancer research: Quantifies interaction effects in in-vitro cancer drug combination experiments to identify synergistic or antagonistic combinations.
• High-throughput screening analysis: Analyzes large compound libraries with uneven concentration grids and replicates to prioritize combinations for further study.

Methodology:

Observed dose-response curves are modelled as the sum of an expected response under a zero-interaction model and an additional interaction effect, with the interaction component modelled by a Gaussian process and inference performed in Stan using sampling or variational inference and optional parallel computation.