Sensitivity&Specificity

[Ultramancc]

Hi,

I'm currently using scoary2 and it's a useful tool. I have a question about sensitivity&specificity
May I know the calculation way for those two values? As I have found there may be some conflicts with sensitivity and specificity, those genes were lowly detected in trait-positive but the results indicate they have a high sensitivity.

[MrTomRod]

Dear @Ultramancc

Sorry for responding so slowly, I was on holiday.

Good question, the cause is a recent change that was made to Scoary2 in response to this issue/pull request:

The original behaviour (Scoary1 and Scoary2 up until 0.0.4):

• The sensitivity of using the presence of this gene as a diagnostic test to determine trait-positivity: g+t+/(g+t++g-t+)
• The specificity of using the absence of this gene as a diagnostic test to determine trait-negativity: g-t-/(g+t-+g-t-)

The current behaviour (Socary2 starting with 0.0.5):

If gene and trait are positively correlated (original behaviour):

• The sensitivity of using the presence of this gene as a diagnostic test to determine trait-positivity: g+t+/(g+t++g-t+)
• The specificity of using the absence of this gene as a diagnostic test to determine trait-negativity: g-t-/(g+t-+g-t-)

If gene and trait are negatively correlated:

• The sensitivity of using the absence of this gene as a diagnostic test to determine trait-positivity: g-t+/(g+t++g-t+)
• The specificity of using the presence of this gene as a diagnostic test to determine trait-negativity: g+t-/(g+t-+g-t-)

Your data

I made a simple Excel that may make it easier to understand what is happening: SensSpecPlayground.xlsx

[Ultramancc]

Thank you so much for your detailed and helpful answer!

[MrTomRod]

No worries! 👍