PCOILS

For sake of clarity we have to mention that this is our implementation of PCOILS, and we cannot guarantee that the original PCOILS program works in the same way, since the authors does not show the explicit algorithm [2]. However, in case of our PCOILS, all the machinery described above still remains untouched, with the exception of function $f$ (Eq. 5). Since we are dealing with evolutionary information computed from a given multiple alignment, instead of the single-sequence $s$ we work with the profile $P_k(a)$, that represents the frequency of residue $a$ in position $k$ of the alignment. In this case the PCOILS score is still defined by an equation similar to Eq.4, but with the new function:

$$x_i=(\prod_{h=1}^W f(S,P,h)^{e_h})^{1/N}$$ (6)

and

$$f(S,P,h)=<S^h,P_{i+h}>=\sum_{a \in \{Residues\}} S^h(a)\cdot P_{i+h}(a)$$ (7)