Our bow-tie paper is finally out. Here’s the teaser:
How does evolution shape living organisms that seem so well adapted that they could be intelligently designed? Here, we address this question by analyzing a simple biochemical network that directs social behavior in bacteria; we find that it works analogously to a machine learning algorithm that learns from data. Inspired by new experiments, we derive a model which shows that natural selection—by favoring biochemical networks that maximize fitness across a series of fluctuating environments—can be mathematically equivalent to training a machine learning model to solve a classification problem. Beyond bacteria, the formal link between evolution and learning opens new avenues for biology: machine learning is a fast-moving field and its many theoretical breakthroughs can answer long-standing questions in evolution.
Read the full paper:
Bow-tie signaling in c-di-GMP: machine learning in a simple biochemical network
Jinyuan Yan, Maxime Deforet, Kerry E. Boyle, Rayees Rahman, Raymond Liang, Chinweike Okegbe, Lars E. P. Dietrich, Weigang Qiu and Joao B. Xavier. PLOS Computational Biology