Table 1. Features used in training the neural network in addition to the electron-density map resolution.

Mean, SD, highest and lowest were calculated for the features when applicable and each was used as a separate feature.

Feature	Mean	SD	Highest	Lowest	Categorical values (0 or 1)	Single value
Ramachandran angles in favoured regions					Yes
Ramachandran angles in allowed regions					Yes
Log-likelihood score (LLK)	Yes	Yes	Yes	Yes
Density score	Yes	Yes		Yes
Root-mean-square deviation (r.m.s.d.)						Yes
Is a tripeptide at the start of a chain?					Yes
Is a tripeptide in the middle of a chain?					Yes
Is a tripeptide at the end of a chain?					Yes