A universal biomolecular integral feedback controller for robust perfect adaptation

Heather posted this link on CSGnet, and I was just reading it too,
Article at Quanta Magazine
Nature: A universal biomolecular integral feedback controller for robust perfect adaptation (Aoki et al, 2019)

Pretty cool stuff! Always nice to see control theory being used in unexpected ways. Synthetic circuits for homeostasis inside of a cell!

I like the terminology in this paper; people often use very different terms for elements of a feedback loop; but they use these terms to mean the same things: controlled variable, regulated variable, output of interest (it is a sensed output, so like an output from an input function). They show an example of controlling “growth rate”, and how it is stable against temperature disturbances.

I don’t understand any of the biochemistry stuff, though. Integral feedback is pretty tough to tune in normal ‘big’ systems, like motors, they are quick to start oscillations, so I imagine it’s a bit more difficult to do it in the micro scale. Would be cool to see if it is a pure integral or a leaky integral.

Yes, I think agree this is better than many papers in its terminology. As Rick has pointed out, it’s still using the terminology of controlling output rather than input, so it’s not taking a PCT perspective. However, from the perspective of bringing other scientists towards PCT, I think that other work showing the power of negative feedback control can be very helpful. I also think that the findings in this study could be used to help understand some of the structural underpinnings of PCT systems.

I don’t think there is any difference between their concept of control and PCT’s. Every signal in a feedback loop is an output from one element and input into another element. They clearly specify what is what, and it is just a regular feedback loop.

The problem with “control of output” only exists when people try to explain behavior of organisms with models that pre-compute the necessary behavior. Outside of heavily stable environments in industrial settings, those kinds of systems don’t work very well or at all.

I don’t see that in this system. Seems to work quite well.

Were you guy talking about this article?

And referring to my review on CSGNet? If so, for the record, her’s my review.

RM: Nice article. But, via it’s focus on the nature of the output function in a control loop. it leaves out that aspect of negative feedback control that is most important for understanding the behavior of biological systems: the fact that what is being controlled in a control loop is a perceptual variable (also called a PERCEPTION). The little lego demo described in the video in the article is a good example of the problem. The article says that the robot is keeping its distance from the target constant, implying that the perception it is controlling (via integral output) is distance. This is possible but it’s also possible that the robot is controlling some other variable that is a correlate of distance, such as reflected sound or light intensity (I built a logo robot that appeared to control its distance from an object by controlling the amplitude of reflected light). It could be (though it’s probably not) controlling a perceptual variable that is even more interesting like the degree of focus of what is written on the paper placed in front of the robot. Such a robot could then be designed to do things based on what is written on the paper, so the robot would have to move to a distance where it gets the right focus so that it knows what to do.

RM: Anyway, what I think about this work is that, like all non-PCT applications of control theory to understanding the behavior of biological (living) systems, the focus on behavioral output leads them to miss what is most important about understanding the controlling done by these systems: what variable(s) the system is controlling. This, of course, is what Powers didn’t miss. And, coincidentally, it was the topic of the talk that Richard Kennaway and I gave at the IAPCT meeting last week.

RM: So I guess I’d have to say that, from my perspective, the paper is a net negative since it continues to encourage a view of control as a process of producing the appropriate output (in this case, integral output) rather than the intended perceptual input.

Yeah, that one and the nature article linked in the first post.