Predicted structure length

[Ieremie]

Hello,

I have used the ESMfold to fold some proteins of interest. In certain cases, the resulting structure file contains a number of amino acids that does not correspond to the full length of the protein sequence.

I thought the generated structure should always have the same number of AA as the seq. Am I wrong here?

For example:

ATOM 76 N GLU A 10 -25.702 1.286 43.183 1.00 47.54 N
ATOM 77 CA GLU A 10 -25.949 2.224 44.274 1.00 51.91 C
ATOM 78 C GLU A 10 -24.642 2.669 44.925 1.00 47.47 C
ATOM 79 CB GLU A 10 -26.873 1.600 45.322 1.00 41.38 C
ATOM 80 O GLU A 10 -23.884 1.842 45.436 1.00 43.84 O
ATOM 81 CG GLU A 10 -28.318 2.067 45.224 1.00 41.88 C
ATOM 82 CD GLU A 10 -29.211 1.486 46.309 1.00 47.96 C
ATOM 83 OE1 GLU A 10 -28.880 0.410 46.857 1.00 48.51 O
ATOM 84 OE2 GLU A 10 -30.252 2.111 46.613 1.00 43.77 O
ATOM 85 N GLY A 12 -21.653 5.712 45.193 1.00 49.22 N
ATOM 86 CA GLY A 12 -20.821 6.605 45.983 1.00 49.72 C
ATOM 87 C GLY A 12 -20.314 7.800 45.197 1.00 49.04 C
ATOM 88 O GLY A 12 -19.696 7.638 44.143 1.00 46.00 O

Residue with index 11 doesn't appear in the final structure for some reason.

[Ieremie]

It seems that the inclusion of unknown AA such as O, U, B, Z leads to the prediction of structures with missing residues.