⟩ How you calculate sensitivity and selectivity of Blast?
Suppose the Blast search returned 100 hits. Of these, 17
were false positives and we knew that there were 165
sequences in
the database which should have returned a hit with our
sequence.
To calculate the sensitivity and selectivity, we must
determine the number of true positives (ntp), the number of
false
positives (nfp) and the number of false negatives (nfn). We
are told that the number of false positives was 17, hence
the
number true positives must have been 100-17 = 83, as there
were 100 hits. Therefore we know that the search algorithm
found
83 of the 165 sequences it should have found, hence the
number of false negatives was 165-83 = 82. So, we know that
ntp = 83,
nfp = 17 and nfn=82. Using the equations in the notes, we
can calculate:
Sensitivity = ntp/(ntp+nfn) = 83/(83+82) = 83/165 = 0.50 (2
d.p)
Selectivity = ntp/(ntp+nfp) = 83/(83+17) = 83/100 = 0.83