Imagine you're sitting at a pond, listening to the din of croaking frogs. You want to know how many frogs are in the pond, but you can't pick out the individual croaks—only the combined sound rising and falling in volume as frogs start and stop communicating.
But what if you were able to examine these volume changes to figure out how many frogs are in the pond?
That's the idea behind a new method developed by the Funke Lab at Janelia to count the individual molecules contained in a single spot of light detected by a fluorescence microscope—a quantity important for understanding the underlying biology of a living system.
Just like the chatter of the frogs at the pond, one spot of light captured by a fluorescence microscope is made up of many individuals—in this case, fluorescently labeled molecules. The resolution of the microscope is limited by the physics of light and the system itself, so the instrument can only detect the sum of these individual contributions—their combined brightness. Like the volume of croaks at the pond, the compound spot varies in intensity over time as molecules blink on and off.
To create their new method, the team first modeled the entire path of light through the system—from the photons leaving the fluorophore to the creation of the spot detected by the microscope. From this, they produced a "trace," or plot, of the spot's intensity over time that depends on all the system parameters, both known and unknown.
Armed with this machine learning model, the researchers then turned to figuring out the unknown parameters in an actual image of a spot of light detected by a microscope. To do this, they fit their model to a trace of that spot's intensity over time. By tweaking the parameters of their model to reproduce the real trace, they were able to infer the unknown quantities, including the number of individual molecules contained in the single spot.
Unlike previous counting methods, which yield a single number of individual molecules contained in a spot, the new method, blinx, gives a probability distribution of all possible answers. This enables researchers to see the confidence of the model's predictions and decide if they need more data.
"Sometimes the data just doesn't support a single answer. There might be so much fluctuation that the information is just not there," says Janelia Group Leader Jan Funke. "This model has the capability to tell you: I really don't know."
blinx can also count more individual molecules than similar methods, making it potentially useful for identifying individual proteins in a sample, according to the researchers. Proteins are made up of varying amounts of different amino acids, so being able to detect the exact numbers of a few key amino acids could help identify which proteins are present.
Funke says he hopes that not only will biologists start using blinx, but that other researchers will improve the new method. "I think this is laying the groundwork for a new generation of these kinds of algorithms," he says.
The researchers say that the development of blinx required a research institute like Janelia.
"It's a super ambitious project: there's a relatively low chance that the thing is going to work, and it takes a lot of multidisciplinary expertise," says Alex Hillsley, a former postdoc in the Funke and Stern labs who led the project. "You need someone that really knows the chemistry, you need someone that really knows super-resolution imaging, you need someone who really knows the modeling, and I can't imagine that coalescing except at a place like Janelia."
