GapSplit: universal sampling for metabolic models

• GapSplit is a sampling algorithm designed to generate uniform, high-coverage sample points on any metabolic model
• regardless of convexity (i.e. logical/integer constraints).

Functions

sample(fname, n_points, lower_bounds=None, upper_bounds=None, n_update=100, n_secondary=0)

• Generate samples from a given input model.

INPUT:

• fname - str

  • String representing path to model file (see gurobipy.read() for acceptable file types).

• n_points - int

  • Number of desired sample points.

• lower_bounds - list/ndarray, optional

  • FVA minimums for model. Generated if not provided.

• upper_bounds - list/ndarray, optional

  • FVA maximums for model. Generated if not provided.

• n_update - int, optional

  • Refresh rate (in points) for console output of current model coverage and sample count.

• n_secondary - int, optional

  • Number of additional gaps targeted for splitting.

OUTPUT:

• samples - ndarray

  • n_points by n_reactions array of sample points.

Dependencies

• gurobipy: 7.0 and up (requires download and license from gurobi.com - license provided free for academic users)
• numpy: 1.14.5