Some of the 20 amino acid types in proteins have atoms that "look" the same. So, for instance, the Cδ1 and Cδ2 atoms in phenylalanine look the same, and thus all phenylalanine residues would have two ways of naming the atoms:

Figure 24. Phe has two seemingly equivalent rotamers. Biophysically, these two rotamers are the same. But for a bioinformatician they can be different. Just think of superposing two structures with software that isn't smart enough to see that the two rotamers are equivalent...

Similar problems can occur in a series of residues, and WHAT_CHECK warns for all of them:

Figure 25. In Val the Cγ1 and Cγ2 cannot be swapped.

Figure 26. In Arg the Cγ1 and Cγ2 cannot be swapped.

Swappable atoms

Similar problems can occur in a series of residues in which atoms (or atom names) can be found swapped without there being any actual biophysical error (but it seems wise to correct these swapped situations anyway):
Residue Swappable atoms
Asp Oδ1 and Oδ2
Glu Oε1 and Oε2
Phe Cδ1 and Cδ2; Cε1 and Cε2
Ile Cγ1 and Cγ2
Leu Cδ1 and Cδ2;
Val Cγ1 and Cγ2
Tyr Cδ1 and Cδ2; Cε1 and Cε2
In all these cases there is nothing wrong biophysically speaking. But as such swaps make it difficult to optimally use the coordinates in (in silico) follow-up studies a correction would be nice.

Swaps that are worth an error message instead of a warning

Obviously, if for example, the Cδ1 and Cδ2 have been swapped in phenylalanine but the Cε1 and Cε2 have not been swapped then we still did not disobey any law of physics, but it gets even harder to use the coordinates sensibly in drug design, in protein engineering, etc.

In this section we show some swaps that actually constitute errors, or get very close to being a real error.


Figure 27. If the Cγ1 and Cγ2 are swapped in isoleucine, then many softwares will draw strange bonds. Like here residue 318 in PDB file 1ARX. In this case there still nothing wrong, biophysically speaking.

Figure 28. But in these two isoleucines, in PDB file 1AZE, the error is more than just administrative. these two isoleucines are also biophysically wrong.


Figure 29. In Thr the Oγ is called Oγ1 and the Cγ is called Cγ2. So it is possible to make two kinds of errors here: Numbering the C and O the wrong way around, and actually swapping them. The latter is discussed below.

Figure 30. However, with threonine things can also go very, very wrong. In this 1.2 Ångström  resolution structure two threonines are observed very close to each other, with opposite hands on their Cβ.

We contacted the depositors of this PDB file (5RXN) about the funny Thr. To show how difficult things can be, these authors wrote us back that "both threonines were nicely located in good density, so they don't see anything wrong and don't know what we were talking about".


In histidine we see a similar problem as in threonine. The δ-atoms should be called Nδ1 and Cδ2. Consequently, following the rules in organic chemistry for atom naming further, the ε-atoms should be called Cε1 and Nε2. Now two kinds of errors can be made. The first is swapping the names. This is inconvenient for the users of the coordinates. However, swapping the atoms for either only the δ-atoms, or only the ε-atoms makes it a serious error as is illustrated with a picture from one example found in the PDB file 1HCE.

Figure 31. In Histidine with the Nδ1 and the Cδ2 swapped (or the ε-atoms swapped, of course; that difference cannot be seen).

WHAT_CHECK reacts 'shocked' on this latter case. It reports weird bond lengths (details of the numbers in these tables are discussed later on in this WHAT_CHECK user course):

  27 HIS ( 27-)  A - CB   CG    1.68   13.3
  27 HIS ( 27-)  A - CG   ND1   1.48    6.4
  27 HIS ( 27-)  A - CG   CD2   1.52   15.2
  27 HIS ( 27-)  A - ND1  CE1   1.41    7.0

It reports weird bond angles:

  27 HIS ( 27-)  A - CG   ND1  CE1  32.66  -72.9
  27 HIS ( 27-)  A - ND1  CE1  NE2  26.96  -65.2
  27 HIS ( 27-)  A - CE1  NE2  CD2  41.22  -50.5
  27 HIS ( 27-)  A - NE2  CD2  CG   43.89  -62.6
  27 HIS ( 27-)  A - CD2  CG   ND1  35.53  -70.6
  27 HIS ( 27-)  A - CB   CG   CD2 154.10   19.2

It reports bumps:

  27 HIS ( 27-)  A - CG  <--> 28 HIS ( 28-)  A - N    0.32  2.68 INTRA
  27 HIS ( 27-)  A - CE1 <--> 28 HIS ( 28-)  A - CD2  0.28  2.92 INTRA

And it reports a weird packing environment:

  27 HIS   (  27-)  A  -  ---   31 HIS   (  31-)  A  -   -5.37

Atom names

The names of the atoms in most residues have been fixed by the PDB. The concept originally was that the atom in the main chain is called the α-atom; and for most residues -certainly the 20 canonical ones- that is then the Cα. The first atom into the side chain is called the β atom, etcetera. As it was not possible in the very early days of the PDB to use Greek characters in computers, it was decided to use the European equivalents of the Greek characters.

Figure 32. The Greek alphabet with the European equivalent characters indicated. (Figure obtained from

Unfortunately, the PDB doesn't always stick to this rule...

Figure 33. KCX (a NZ-carboxylated lysine), with its backbone atoms coloured bright green. In this residue the 'official' atom names are: N CA C O CB CG CD CE NZ CH OX1 OX2. The Greek alphabet rule thus is disobeyed. I can imagine that one decided to skip θ as its European equivalent has two characters (th), but the ξ now chosen instead seems a rather illogical choice.