MBA

MBA performs automated assignment and Bayesian uncertainty assessment of protein backbone NMR resonances to map atoms to spectral peaks and quantify assignment confidence.

Key Features:

• Bayesian statistical model: Employs a Bayesian statistical model that accounts for sources of uncertainty and yields posterior distributions over assignments.
• Tree-based stochastic optimization: Uses a tree-based stochastic optimization algorithm to explore the space of assignment solutions consistent with the data.
• Posterior standard deviations: Quantifies uncertainty for individual resonance assignments using posterior standard deviations.
• Posterior distribution of plausible assignments: Assesses information content through posterior distributions of plausible assignments to indicate support for alternative solutions.
• Overall plausibility measure: Computes a measure of overall plausibility for assignment sets to evaluate reliability.
• Incorporation of spectral stochasticity: Explicitly incorporates random variation in NMR spectra into the inference to avoid deterministic bias.
• Automatable inference framework: Provides an automatable framework for Bayesian inference of backbone resonance assignments.
• Implementation: Implemented in Java.

Scientific Applications:

• Protein backbone resonance assignment: Assigns and assesses uncertainty in backbone resonance assignments for protein NMR spectroscopy.
• Experimental validation: Applied to experimental data from Human Ubiquitin and Cold-shock protein A from E. coli.
• Simulation studies: Uses simulations to illustrate how different experimental conditions impact assignment uncertainty.
• Reliability assessment: Supports evaluation of the reliability and interpretability of protein structure studies derived from NMR assignments.

Methodology:

Performs model-based inference using a Bayesian statistical model, tree-based stochastic optimization, posterior distributions and posterior standard deviations to characterize assignment uncertainty.