8-25-08 BAL Angle for sp2 acceptors: different hbeTypes

Thinking about what the BAL angle should really be for sp2 acceptors.
According to quantum mechanics : lone-pairs should be at roughly 120 degrees for sp2 acceptors.
For side-chain sp2 acceptors we have used 135 degrees, based on histograms for BAH angles. If we continue this approach for helix and BB-other acceptors we should actually increase the lone-pair angles, since the BAH distributions are even more shifted towards larger angles for these types - "lone pairs" would have even less connection to chemical intuition if we did this. Turns also have a strange distribution, although in this case shifted more towards lower angles.

8-01-08 Refactoring

Refactoring hbonds_geom so that it will be straightforward to use LP potential for different hb types.

7-29-08 Option to use lone-pair potential

Added an option to use the lone pair potential.

Use the lone pair potential:
./simple_relax.macosgccrelease -in::file::s 1nkdH.cln.pdb -database ~/minirosetta_database -HBonds::use_LP

Don't use the lone pair potential (use standard):

./simple_relax.macosgccrelease -in::file::s 1nkdH.cln.pdb -database ~/minirosetta_database

Note that the standard potential is invoked by not specifying any value for the flag (it defaults to false). Specifying -HBonds::use_LP causes the lone pair potential to be invoked.

7-10-08 Relaxes on small set (~30 proteins)

Have run 10 relaxes (ClassicRelax^?) per protein on a small set of ~30 proteins from the top500. The proteins were chosen so as to have a mix of structural motifs. The relaxes were done using the standard mini hbond potential, the coupled noLP potential for side-chain (SC) sp2 hbonds (16 terms), and the LP potential for sp2 SC set) (sp2 SC : HbeType^? 5, 8, 9)

List of proteins in small set:

• small_set_list.txt: List of proteins in small set

HB dumps:

• hb_native_base_2008_07_07.txt.zip: hb dump for natives (small set)

• hb_relax_base_2008_07_07.txt.gz: hb dump for standard mini relaxes (small set)

• hb_relax_LP_2008_07_09.txt.gz: HBd dump for LP relaxes (small set)

• hb_relax_noLPcomb_2008_07_07.txt.gz: hb dump for relaxes coupled noLP (small set)

Making some 1-dimensional histograms for various structural features (for the SC HBonds) for these dumps, along with comparisons to top5200 natives. Also histograms for energies for the different relaxes. The histograms for distance(AH) are interesting in that the relaxes using the standard potential showa small number of long (>2.4 Å) AH distance HBonds which are not present in the other relaxes. HBonds with similaar AH distances are also found in the natives for the small set, but not in the top5200. The histogram of BAH angle (not volume normalized) shows the optimal BAH in the top5200 well.

• html histogram of BAH angle (not volume normalized)

• html_hist_chi.zip: html chi

• html_hist_cos_AHD.zip: html cosAHD

• html_hist_cos_ALH.zip: html cosALH

• html_hist_cos_BAH.zip: html cosBAH (small set)

• html_hist_cos_DHL.zip: html cosDHL

• html_hist_cos_DHL.zip: html cosDHL

• html_hist_dAH.zip: html dist(AH)

• html_hist_energy.zip: html energy

7-2-08 Relaxes

Running top500 relaxes with 1) regular mini parameters 2) coupled hbond term w/o LP (for side-chain) 3) coupled hbond term with LP (for side-chain). Relaxes have finished for about 90 of the 500 proteins.

hb_2008_07_01_hires_top500_relax_base.txt.zip: hb relaxes for 90 of top500 (regular mini)

hb_2008_07_01_hires_top500_relax_noLPcomb.txt.zip: hb relaxes for 90 of top500 (coupled no LP)

hb_2008_07_01_hires_top500_relaxLP.txt.zip: hb relaxes for 90 of top500 (Lone pair potential)

Abe-Go derivatives: which modes?

Another thing about the geometric solvation: It's not clear whether or not the Abe-Go calculations should be done with the LP potential if this is used for geometric solvation - Abe-Go is normally turned off for derivative calculations called from the geometric solvation code. If I change the types of "atoms" (including adding LP) used in the potential though should they be excluded or not? (I could probably get the B2 if needed.)

• Many modes need only the energy values and no derivatives; only modes that use continuous minimization need derivatives. At this level we should not need to know that -- we just know whether we are told to compute derivatives or not. -- JackSnoeyink - 28 Jun 2008

6-25-08: tempfactor

Created workspace tempfactor for a mini branch that can read in and preserve temperature factors. Newly constructed atoms are given temperature factor of 99.00. Branched from r23257 like with the other workspaces (can merge changes from main mini later.)

Files changed for tempfactor: mini/src/core/conformation/Atom.hh mini/src/core/io/pdb/file_data.cc mini/src/core/io/pdb/pdb_dynamic_reader.cc also the hbonds analysis script would need to be changed if temp factors are used.

I haven't changed occupancy (also ignored in main mini) here, since I don't need it.

Some of this code was in the main mini code, just commented out for some reason.

The files that will be relaxed (which I started running earlier) will not have the correct temperature factors - probably not important since there shouldn't be any atoms with random coordinates in the relaxed structures, unlike with the natives (where there are sometimes unknown coordinates).

June 24. 2008 update: geometric solvation

Changed LP and noLP potentials so that geometric solvation code calls traditional potential. This is done in part because the geometric_solvation code can call compute_hbond_energy directly (traditional) depending on flags and also for the LP potential the calls to hbond_energy_deriv from the solvation code are weird in that base==base2 and therefore you can't place lone pairs correctly (for sp2 acceptors).

• Oh, crap. I didn't know there would be messes like this. We need to get Andrew to fix this one. -- JackSnoeyink - 28 Jun 2008

June 24, 2008: coupled HBond potentials and clamping

From last Monday till Today I created miniRosetta versions for the coupled HBond potentials for sp2 side-chain (SC) HBonds (at least one atom is SC). Versions exist for both lone-pair (LP) and the three-traditional mini HBond parameters (noLP).

The potentials are from least-squares fits to top5200 good HBonds (low B-factors, good geometry). The nonLP potential has 16 terms, the LP potential 20 terms. The potentials were not derived with other terms. Both potentials are forced to zero to avoid non-physical minima at strange proton-acceptor distances - this is not done smoothly however. The LP potential is only forced to zero at very short distances where Lennard-Jones potentials would probably prevent it anyways (<= 1.4 Å). The noLP potential is forced to zero at both extremely short distances (<= 1.45 Å) and at longer distances (>= 2.5 Å) and I'm a little worried about the latter since the polynomial didn't always go positive there, but it doesn't seem to be a problem with relax.

I'm running some relaxes in mini on the highest resolution proteins from the top500 with the traditional mini potential (Kortemme 3-parameters), the LP potential, and the noLP potential. Surprisingly the discontinuities in the potentials (from cutoffs or from protons jumping between LPs) don't seem to give error messages during minimization - hopefully it isn't such a problem - overall energies seem quite similar between all potentials (note that they are all scaled similarly).

files changed: hbonds_geom.cc, hbonds_geom.hh, HBEnergy.hh (for LP), hbtrie/HBAtom.hh (for LP - see note below)

Also I fixed a bus-error problem with my hbond analysis script which appeared when the mini code was updated. If I want to look at only low B-factor hbonds I need to change a few files - I have the changed files locally, but not in a workspace.

• B-factors are used to cover multitude of sins, according to Jane Richardson. Don't put too much interpretation on them. -- JackSnoeyink - 28 Jun 2008

Notes to self:

In order to always get the correct potential in the LP case to be called I added some extra data members (and accessors/mutators) to the HBatom class (hbonds/hbtrie/HBAtom.hh). Specifically a utility/pointer/owning_ptr to a const conformation::Residue and an int for an atom_id - it needs to be an owning pointer rather than an access or raw pointer or you get a bus error (different than the analysis one). Also added some lines in HBEnergy.hh to set these new members and to call the LP potential in a function used by the trie.

On emerald, use -R blade option for bsub to always get right machine on emerald (bsub -R blade -o foo.log < foo.csh).

On emerald, moved top5200 protein tar from scratch to ~/ms.

• hb_2008_07_01_hires_top500_relax_base.txt.zip: hb relaxes for 90 of top500 (regular mini)

• hb_2008_07_01_hires_top500_relax_noLPcomb.txt.zip: hb relaxes for 90 of top500 (coupled no LP)

• hb_2008_07_01_hires_top500_relaxLP.txt.zip: hb relaxes for 90 of top500 (Lone pair potential)

• hb_native_base_2008_07_07.txt.zip: hb dump for natives (small set)

• hb_relax_base_2008_07_07.txt.gz: hb dump for standard mini relaxes (small set)

• hb_relax_LP_2008_07_09.txt.gz: HBd dump for LP relaxes (small set)

• hb_relax_noLPcomb_2008_07_07.txt.gz: hb dump for relaxes coupled noLP (small set)

• html_hist_chi.zip: html chi

• html_hist_cos_AHD.zip: html cosAHD

• html_hist_cos_ALH.zip: html cosALH

• html_hist_cos_BAH.zip: html cosBAH (small set)

• html_hist_cos_DHL.zip: html cosDHL

• html_hist_cos_DHL.zip: html cosDHL

• html_hist_dAH.zip: html dist(AH)

• html_hist_energy.zip: html energy

• small_set_list.txt: List of proteins in small set

• html_histBAH.zip: html histogram of BAH angle (not volume normalized)

• NewestPresentation.ppt.zip: PPT

to top