Supplementary Figure 5: Normalization of fluorescence trajectories.

Representative fluorescence trajectory before (a) and after (b) normalization using the emulateFRET program as described in the Methods section. Briefly, to normalize each trajectory we first computed the maximum non-outlying value in this trajectory, which is the 98th percentile value among the values within the trajectory. Each fluorescence value in the trajectory is divided by the maximum non-outlying value; it is further multiplied by a factor of .85 to ensure that the entire trace fits well within the 0–1 range. This resulting trajectory is smoothed with a five-point moving average. Any values in the trace which are still above 1 (these are not expected to occur with any significant frequency) are normalized down to equal 1 (though if this occurs, a small amount of noise is added to these few points, to avoid the trajectory equaling exactly 1 at any point). The second step in normalization determines the location of the new baseline. A histogram of the values from all trajectories collected during the first 30 seconds of the experiment is constructed. This histogram should have a peak at the value at which the baselines center, with a deviation corresponding to the variation in that baseline. The point which is two standard deviations above the mean of this peak is considered the new baseline for the traces. All values in the trace below this value are cut off, and set equal to this new baseline. To avoid the baseline remaining precisely flat (causing over-fitting in ebFRET), a small amount of noise (magnitude .005, at the most) is added throughout the traces. This new set of trajectories is then saved in a format, which ebFRET can read.