Machine learning of cytoskeletal machines

 

Traditional bottom-up physical-mathematical models have longstanding popularity and success
in studying cytoskeleton and mechanochemical machines driving cell movements and division.
These models brought and will continue to bring mechanistic insights into cell migration.
However, such models are either too simple to embrace the complexity of the multiscale cell
processes or are hopelessly cumbersome and unwieldy to be used to nimbly test multiple
hypotheses. Machine learning and AI approaches have demonstrated immense strength in
identifying statistical patterns in cytoskeletal machines and in predicting cytoskeletal dynamics
from microscopy data. However, these data-driven approaches largely neglect the laws of
physics and chemistry needed to ground the discoveries in biological mechanisms. These
complementary strengths and weaknesses between the traditional modeling and modern data-
scientific approaches suggest a promising avenue forward: augmenting traditional models with
data-scientific and AI methods for the sake of building more complex traditional models that can
be directly connected with the enormous volumes of biological data of cytoskeletal machines.

 

This workshop will convene data scientists, experimental biologists, mathematical modelers and
biophysicists using or interested in starting to use ML to study cytoskeletal dynamics, cell
migration and mitosis. The goal is to foster an exchange of ideas between these research
communities. The workshop is structured to help participants identify the most promising
opportunities for developing and using ML tools to answer biological questions. The program
includes both overview and research talks, poster sessions and lightning talks by poster
presenters, and will have ample time for participants to get to know each other, exchange ideas
and foster collaborations.