Continuous attractors as unreliable estimators Arvind Murugan - PowerPoint PPT Presentation

Continuous attractors as unreliable estimators Arvind Murugan Dept. of Physics

Regression using attractors High dim config space Regression Attractor dynamics

Circadian clocks as phase estimators External signal (Day-night cycle of light) Time -> Cyanobacteria Gorbunov 2001 clock state clock state Synechococcus Elongatus Justin Chew, Rust lab Time -> Time -> Phase estimate (clock state)

Circadian team Theory Experiment Justin Chew Wee Pittayakanchit Zhiyue Lu Michael Rust

The best estimator?

The best estimator? Internal unreliability

Two kinds of bugs Two kinds of estimators Synechococcus Elongatus Procholorococus Marinus Continuous Point attractor attractor (limit cycle) Equally good in noiseless cycling conditions Good Bad Atmospheric noise + Internal noise Really bad Bad

Limit cycle clocks Synechococcus Elongatus Continuous attractor (limit cycle) Leypunskiy et al. (eLife 2017) (Rust lab)

Point attractor clocks Synechococcus Elongatus Procholorococus Marinus Continuous Point attractor attractor (limit cycle)

Weather fluctuations External noise

Continuous attractors limit dark pulse delays Continuous attractor Point attractor 2.5 hr dark pulse during the day Delay: Delay:

Internal noise Synechococcus Elongatus(~ 3 um) Molecules are discrete ~ 10000 copies of KaiC Prochlorococcus Marinus (~ 0.5 um) ~ 400 copies of KaiC Justin Chew, Rust lab Particle Distribution Diffusion constant ~ 0 copies of KaiC

Internal noise Diffusion (internal noise) No external noise

Flatness encourages diffusion Curvature discourages diffusion Flat Curved

External + Internal noise Circadian clocks Prochlorococcus Synechococcus Marinus Elongatus Fixed External noise Gillespie simulation of KaiABC molecular model Internal noise

Different strokes for different folks

S. Elongatus ~ 40000 copies ~ 2500 copies Experiments by Justin Chew, Rust lab