Output.md

Output Interpretation

During prediction of new cells, several output files are generated in the specified output folder:
prediction.tsv includes the predicted class assignment (0 or 1) together with the class 1 probability for each cell. Values close to 1 indicate high-quality cells, values close to 0 indicate low-quality cells. The decision boundary for the class assignment is 0.5. Values between 0.3 and 0.7 might need a second look to verify the class assignment and are again listed in prediction_critical.tsv.
If an annotation file was specified to compare the predicted class assignments, prediction_accuracy.txt provides performance information along with wrongly predicted cells.

The structure of the output files is as follows, linked are the output files that can be generated for the test data:

prediction.tsv:

cell	prediction	probability
cell name	class prediction (0 or 1)	class 1 probability

prediction_critical.tsv:

cell	probability
cell name	class 1 probability (value between 0.3 and 0.7)

prediction_accuracy.txt:

• list of false positive predicted cells
• list of false positive predicted cells with critical value
• list of false negative predicted cells
• list of false negative predicted cells with critical value
• accuracy
• f1-score
• number of true positive (tp), true negative (tn), false positive (fp), false negative (fn) predicted cells