Agent-based models, open source, integrating HPC+ABMs+AI

Submitted by jacklyn.ebiasah on Fri, 03/13/2020 - 12:17

Hello!

I lead PhysiCell, an open source, cross-platform agent-based modeling platform for multicellular systems. 

This platform can simulate 10^5 or more cells with 5+ diffusing substrates (e.g., oxygen, metabolic factors, signaling factors, drugs) in 2D or 3D. PhysiCell uses off-lattice agents, where each cell agent has its own cell cycle status, death models, motility rules, secretion/uptake of chemical factors, basic mechanics, and growth rules. Each cell agent can have user-defined rules and custom data. 

On top of the agent modeling platform, we are developing a software ecosystem. Any PhysiCell model can be rapidly converted into a cloud-hosted, interactive model using xml2jupyter, which opens up new possibilities for public outreach, education, and scientific communication and dissemination. Examples: 

PhysiCell models can also be run on HPC (high performance computing) environments for massive model exploration, and machine learning can help to accelerate and interpret these investigations.  

We are planning our first ever PhysiCell hackathon for July 2020, with some NCI funding available for students and early career researchers. We also have funds for two 6-week visitors for long-term PhysiCell projects in summer 2020. 

References:

  1. PhysiCell method paper
  2. PhysiCell wins PLoS Computational Biology Research Prize for Public Impact 
  3. PhysiCell in high throughput on HPC
  4. Using HPC and Active Learning for accelerated discovery in immuno-oncology

Contact: macklinp@iu.edu

Tags
agent-based modeling
Cancer
systems biology
Machine learning and Artificial Intelligence
HPC
multicellular systems biology
open source