Overview

BioSwarm is an OpenStep framework for computational biological modeling and simulation. It is meant to accompany BioCocoa's bioinformatic capability with computational modeling. Currently BioSwarm includes:

• Gene network inference from microarray data. Providing a linear regression-based optimization algorithm described in this research paper. Preliminary support exists for nonlinear regression as well.

• Flux-balance analysis (FBA) for genome-scale metabolic models. Support for compound, reaction and model databases are provided in BioCocoa, but BioSwarm provides the linear programming capability to do FBA.

• Variety of computational biology modeling classes including ODE reaction, stochastic reaction, PDE reaction-diffusion, subcellular element method, and hybrid models that combine methods. Models are specified using property list files.

• Support classes for managing multi-scale 2D and 3D spatial grids.

• Generation of CUDA GPU code for the computational biology modeling classes which allows for massively parallel simulation. Earlier versions of this effort is described in this research paper.

BioSwarm is ongoing research project with the goal to integrate the approach of machine learning to discover models from biological data, with the approach of computational in silico experimentation to elucidate the performance objectives of those biological models. BioSwarm represents the advancement of computational tools to allow the study of more sophisticated and predictive computational models of biological phenomena.