To understand a mechanism, you have to poke it. A child taking apart an alarm clock doesn't just stare at the gears: they turn the wheel and listen to how the ticking changes. Scientists do the same: instead of passive observation, they feed specially chosen signals into the system and watch the response.
This new study proposes a method that automatically selects the most "tricky" input signal — one that makes different hypotheses about the system yield maximally different predictions. It's like an exam: the teacher doesn't ask the easiest questions but those that best reveal the difference between a strong and a weak student. Stochastic modeling here is like a map: the more accurate the map, the better you can navigate.
The authors tested their approach on numerical examples and then applied it in an experiment with ion channels — proteins in the cell membrane that open in response to light and let charged particles through. They were able to confidently select the best model among several candidates.
This approach resembles the work of an investigator: to find the truth, they ask questions to which different versions of events give fundamentally different answers. The greater the divergence in answers, the faster incorrect versions are eliminated. The method promises to speed up the search for correct models in biology, neuroscience, and other fields where data are scarce and hypotheses are many.
The light-sensitive proteins studied by the authors are used in optogenetics — a technique that allows neurons to be switched on and off with light.