Supplement section 1.9.7 of Jumper et al. (2021), p. 38. The error in
the position of residue j when prediction and truth are aligned
on residue i is predicted as a distribution over distance bins;
the reported PAE is that distribution's mean.
Usage
Alfpae(z, w, bins = NULL)
Arguments
- z
Pair representation, an n x n x cz array.
- w
Projection to the bin logits.
- bins
Bin centres; defaults to 64 bins of width 0.5 A over 0-31.5 A.
Value
A list with the pae matrix, the distributions p,
estimate, n and method.
Details
Unlike the distogram head this one does not symmetrise the pair
representation: aligning on i and reading off j is a
different question from the reverse, so the matrix is deliberately
non-symmetric.
References
Jumper et al (2021) Nature 596:583-589, Suppl. section 1.9.7
Examples
set.seed(7)
z <- array(rnorm(3 * 3 * 4), c(3, 3, 4))
r <- Alfpae(z, w = matrix(rnorm(64 * 4), 64, 4))
round(r$pae, 2)
#> [,1] [,2] [,3]
#> [1,] 13.22 11.76 11.98
#> [2,] 12.07 19.69 14.66
#> [3,] 14.05 11.47 13.56
