1
$\begingroup$

For my research I'm using a tool called seeSUMO, which predicts sumoylation sites in a protein based on sequence features. When it reports your results, it gives you a level of confidence for each individual lysine residue, that is predicted to be a sumoylation site, based on sensitivity.

This confidence is defined as: confidence = 1 - sensitivity (for + predictions)

and 1 - specificity (for - predictions)

How to interpret this confidence level? For example for + predictions: if your sensitivity is very low, this means that you have a very stringent test, thus giving you a high confidence if you eventually get a positive result?

However, I still find this interpretation strange, since if I would devise a very bad test or a random test with essentially a very low sensitivity, wouldn't this also supposedly give me high confidence in this prediction?

Improve this question
$\endgroup$
2
  • $\begingroup$ This sounds very odd. You can only be confident of a test result if it has good diagnostic accuracy. Generally use the Spin/Snout rule: Sp-p-in "specific test when positive rules in" Sn-n-out "sensitive test when negative rules in". Sounds like maybe a mistranslation? $\endgroup$
    – tristan
    Commented Mar 6, 2015 at 20:01
  • $\begingroup$ And low sensitivity doesn't mean stringent test, that would be high specificity, which is often associated with lower sensitivity. $\endgroup$
    – tristan
    Commented Mar 6, 2015 at 20:02

0

Reset to default

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.