Lots of other predictions, tests. We are hoping this stimulates some further experiments to try to prove us wrong! In particular, our model would predict that cells should be attracted to a particular point in the device - and we don't see this immediately in the data (but can't rule it out).

Why does internalization matter for chemorepulsion? Internalization decreases the amount of bound receptor. If you inhibit it, you get more bound receptor - and you get attraction again!

We then propagate error from the noise in the ligand-receptor binding to the noise in the response, and work out the signal-to-noise. The triangle is the 0-500 ng/mL experiment - repelled! The square is the 0-100 ng/mL experiment - attracted!

Why does response switch from increasing to decreasing as you go toward larger chemoattractant (larger probability of bound receptor)? If A and I are nonlinear functions of the bound receptor, the ratio of the two R = A/I can easily switch between being increasing and decreasing- chemorepulsion!

Then we take an idea from earlier work on growth cones and assume bound receptor regulates the eventual response of the cell via a nonlinear feedforward loop.

Because we know internalization is important, we start with the simplest possible model for ligand-dependent internalization, which does reasonably at capturing the timescale for the experiments.

Ten years ago, I saw a paper with some data that has bothered me ever since: B cells in a 0-100 ng/mL gradient of CCL19 are attracted to CCL19, but B cells in 0-500 ng/mL are repelled (see movie, ignoring the big clusters for now!). Why? Here's our model! doi.org/10.1101/2025...

5/ In principle, this makes the cell's "effective" dissociation constant adapt perfectly to changes in the ligand concentration - ensuring ~50% of receptors are always bound,

4/ If an allosteric protein binds to the receptor, like in the first diagram, and this binding changes K_D to K_D/α, you want more of the allosteric protein to lower K_D to adapt to lower concentrations. So the idea is that bound receptors inactivate G and unbound receptors activate G.

1/n New preprint: how eukaryotic cells could potentially adjust to new environments with perfect adaptation of their receptors (but why they probably might not). doi.org/10.48550/arX...

Not quite sure why @aip-publishing.bsky.social has scummy advertising that pretends to be a download link. This isn't ideal for a scientific society. (Brought to you by Chrome no longer supporting adblockers!)

I was curious about these failure modes. Surprisingly robust that you get wrong answers but easy to fix. I suspect LLMs are most useful when solutions are easy to check but hard to generate

New work from Emiliano Perez Ipina + me: if cells respond to chemoattractant A and chemoattractant B, when do they go to source A vs source B? Or do they ever wander back and forth? We show a lot of different options are possible: arxiv.org/abs/2507.19341

Google scholar has a known bug you should look out for when generating bibliography entries. If you published a paper and earlier had a published abstract with the same title, Google Scholar will assume that the abstract is the correct version. This paper was not published in Biophys. J!

I'll be giving the Biological Physics / Physical Biology seminar this Friday (Zoom) on how to connect protein motion on the membrane to how cells respond to an applied electric field! Includes some unpublished work- some of the most exciting in my career! sites.google.com/view/bppb-se...