Skip to main content
. 2020 Nov 3;9:e60404. doi: 10.7554/eLife.60404

Figure 3. Confusion matrices of DeepFRET classification based on the ground truth data test.

(a) Classification accuracy of data in the six categories for the ALEX-enabled model, or the ALEX-disabled model. The absolute number of frames is shown while the fractions for each classification is displayed in parentheses (as calculated row-wise for each true label). The diagonal percentages show the accurate classification of DeepFRET (b) per-trace classification accuracy based on accepting only traces that are classified as smFRET (static/dynamic), and non-FRET data.

Figure 3.

Figure 3—figure supplement 1. Precision-recall of the neural network and human participants.

Figure 3—figure supplement 1.

Precision is plotted against recall for all precision on a simulated dataset containing 46 true smFRET traces and 954 non-smFRET traces. Additionally, the precision-recall for the entire test set (20,000 traces) is plotted for the model. The error bars on the average participant performance represent the standard deviation of the three individual participants.
Figure 3—figure supplement 1—source data 1. Data underlying Figure 3—figure supplement 1.