Ligand params file proton_chi parameters

The numbers following "EXTRA" are the additional sampling (plus minus) that one would get in situations where you invoke the "extra" rotamers.

Backing up a moment, for protein residues, Rosetta uses the Dunbrack rotamer library. This is derived from the sidechain conformations seen in the PDB. Roland Dunbrack and lab took a set of a large number of such observed sidechain conformations and attempted to condense these highly varied sets of strucutres into a discrete set of "rotamers".  What you end up with for each amino acid is a small number of such "rotamers" which represent the structural clusters of these sidechain conformations.

But you don't just get out a single number for each sidechain dihedral. Because it's a cluster of varying structures, you also get a (roughly gaussian) distribution of the frequency of each sidechain dihedral, representing the variation seen. So the Dunbrack rotamer library actually records both the mean and the standard distribution of these clusters.

When Rosetta samples based on the Dunbrack rotamer library, by default it just used the center of these clusters. However, there are flags to enable "extra" sampling of these rotamers. This is benefitial because sometimes you'll get an "off rotamer" conformation which aleviates a clash or has a better interaction energy versus the center of the gaussian. If you're packing with fixed rotamers (e.g. you don't allow for the rotamers to minimize while you optimize the rotamers), that can help out.

Typical flags to control this extra sampling are `-ex1` and `-ex2`, which state that in addition to the center of the gaussian, you should also sample plus or minus a standard deviation for chi 1 and chi 2 (respectively). There's additional flags to control other chis, how many additional samples and where, and when you invoke these extra samples (as extra sampling on the surface of the protein usually isn't all that productive.)

Back to ligands. While ligands don't use the same Dunbrack rotamers, the framework (especially for PROTON_CHI) is basically the same. You have the center of the torsional well specified. As you surmised, the first number after SAMPLES is the number of different wells you have, and the values following are the actual values (in degrees) which are given for each individual rotamer.  The numbers after EXTRA are similar, except that they're the equivalent to the standard deviation in the Dunbrack library.

Using your example, PROTON_CHI 1 SAMPLES 2 0 180 EXTRA 1 20  This means that this chi has two base samples, of 0 and 180 degrees. In cases where extra chi samples are enabled, there's two additional (one plus and one minus) samples for each base sample, so you'll have sampling of -20, 0, 20, 160, 180, and 200 degrees. Likewise with something like PROTON_CHI 2 SAMPLES 3  -60 60 180 EXTRA 2 5 10 you'd have base samples of -60, 60 and 180 normally, and then in cases where the extra chis are enabled you'd get samples of -70 -65 -60 -55 -50 50 55 60 65 70 170 175 180 185 and 190 degrees.

Looking at the code, it looks like the extra chi determination for ligands is a little iffy -- if you're manually setting which torsions you wish to sample (particularly for heavyatoms), I'd recommend just setting them in the SAMPLES section, and just set EXTRA 0

A final note - be careful on how many samples you set up. The total number of rotamers is multiplicative for each torsion. That is, if you have six torsions, each with five samples a piece, you'll get 5 * 5 * 5 * 5 * 5 * 5 = 15625 total rotamers, which is quite a few. If you're not careful, a ligand with a large number of torsions can quickly explode.  That's why you typically use the pre-generated conformers - you're able to take advantage of correlations between the different torsions and avoid the combinitorial explosion.