LooplessFluxSampler: an efficient toolbox for sampling the loopless flux solution space of metabolic models.

Area of Science:

• Systems Biology
• Metabolic Engineering
• Computational Biology

Background:

• Uniform random sampling of metabolic flux solutions is crucial for unbiased network appraisal.
• Convex samplers often generate thermodynamically infeasible loops in large metabolic models.
• Existing methods for sampling loopless flux spaces are inefficient and lack theoretical guarantees.

Purpose of the Study:

• To develop an efficient algorithm for exploring the loopless, mass-balanced flux solution space of metabolic models.
• To address the limitations of current methods in handling thermodynamically infeasible loops.
• To provide a theoretically sound framework for statistical inference in metabolic network analysis.

Main Methods:

• Introduced LooplessFluxSampler, an algorithm based on Adaptive Directions Sampling on a Box (ADSB).
• ADSB is an adaptation of the Parallel Adaptive Direction Sampling (ADS) framework, ensuring theoretical convergence.
• The algorithm samples directions that adapt to the target distribution for efficient space traversal and faster mixing.

Main Results:

• LooplessFluxSampler efficiently explores the loopless mass-balanced flux solution space.
• ADSB guarantees targeting the uniform distribution over convex regions and converges on non-convex regions.
• Demonstrated efficient traversal and faster mixing compared to existing methods.

Conclusions:

• LooplessFluxSampler enables scalable statistical inference for large metabolic models.
• The algorithm provides efficient and reliable exploration of the non-convex loopless flux space.
• Includes a Markov Chain diagnostics suite for assessing sample quality and algorithm performance.