Specificity in cuffcomare

[ZhengXia]

Hi Folks,
cuffcompare is a tool in Cufflinks package which will compare the predicted transcripts with the reference annotation. Cuffcompare will produce two statistics, sensitivity and specificity in the exon, intron and transcript level.
The definition of sensitivity and specificity are:
specificity = TN/(TN+FP)
sensitivity = TP/(TP+FN)
But in the source code of cuffcompare.cpp, the specificity and sensitivity are defined as
" fprintf(fout, "#--------------------| Sn | Sp | fSn | fSp \n");
double sp=(100.0*(double)ps->baseTP)/(ps->baseTP+ps->baseFP);
double sn=(100.0*(double)ps->baseTP)/(ps->baseTP+ps->baseFN);"
The calculation of sensitivity is the same with the standard definition. But the calculation of specificity is different. According to this calculation, it should be precision or positive prediction value.

If I'm wrong, can anyone correct me? Appreciate.

Best,