Hi Jason,

The ligand and APHE got pushed away from one another and off of their respective densities. Occupancies for both were reduced to 0.5 for the refinement. What am I doing wrong here? Does the ligand need to have an altloc as well? But I don’t see an alternative conformation.

Yes, the ligand also needs an altloc, and it should be different from PHE altloc. This way the refinement will know that these atoms don't see each other. You don't have to put two alternative conformations for the ligand, you can have just one with partial occupancy (<1.0) which will mean that the ligand is not always there. Best regards, Oleg Sobolev.

On Feb 28, 2018, at 5:35 PM, van den Bedem, Henry < [email protected]> wrote:

If the phe is a gatekeeper, shouldn’t the ligand be refined at partial occupancy; i.e. occupancies not necessarily have to ‘add up’ as you suggest? Maybe this link is helpful too: www.biorxiv.org/content/ early/2018/01/24/253419

H

*From: * on behalf of "Phan, Jason" < [email protected]> *Date: *Wednesday, February 28, 2018 at 1:30 PM *To: *"[email protected]" *Subject: *[phenixbb] Overlapping ligand-protein side chain density

All,

A piece of a ligand and the side chain of a “gate-keeper” Phe occupy the same space with well-defined features observed for both (resolution is 1.7 A). It looks about 50:50. How do you refine both ligand and protein side chain in this case? A couple of phenixbb suggestions for dealing with ligand-ligand overlapping density have been considered. With the first suggestion, the two entities still clashed and moved apart off of their respective densities. The second suggestion is not applicable in this case since the other molecule is not a ligand but part of the protein but it was tested anyway. Although there was no bumping, the occupancies don’t add up, resulting in a big blob of negative density around the ligand piece.

------------------------------ Only two comments:

1. At that resolution, constrained group occupancy refinement should work reasonably well (provided you can model the 2 entities). Then you also do not have clashes between the molecules, because Occ(A)+Occ(B)=1, meaning when one (A) is there, the other one (B) is not. This works with refmac (external keyword file); if you need more sophisticated occupancy re/constraints SHELXL may offer more opportunities. 2. There is no necessity for the two NCS copies of the binding site to look exactly the same (non-equivalent). Maybe there is a good reason/story (accessibility, contacts etc) for one site to be occupied differently than the other one.

Best, BR —————————————— Modeling two molecules that occupy overlapping binding sites in a structure simply involves designating them as alternate conformers, with the same chain and residue number, and an occupancy that sums to 1.0. For example, if you have an AMP and an ADP that occupy the same binding site, you would define them as

AAMP B 501 BADP B 501

and initially set the occupancies for the atoms in each conformer to the ratio (50:50, 30:70, etc.) that you observe in the density.

Refinement in this manner is straightforward in PHENIX.

Diana ----------------------------

_______________________________________________ phenixbb mailing list [email protected] http://phenix-online.org/mailman/listinfo/phenixbb Unsubscribe: [email protected]

P.S.: In case you want the occupancies to add up to 1, the constrained group will be looking like this: refinement {   refine {     occupancies {       constrained_group {         selection = chain A and resseq 123 and resname LIG         selection = chain B and resseq 890 and altloc A       }     }   } } Note two selections in the group, meaning that occupancy of (chain A and resseq 123 and resname LIG) will be refined and constraied between 0 and 1, occupancy of (chain B and resseq 890 and altloc A) will be refined and constraied between 0 and 1, and occupancy of (chain A and resseq 123 and resname LIG) plus occupancy of (chain B and resseq 890 and altloc A) will be 1. Pavel On 3/2/18 06:59, Pavel Afonine wrote:

Hi Jason,

Oleg is correct, you can assign the ligand and PHE different non-blanc altlocs (say A and B). In this case PHE and ligand will not 'see' each other via repulsion term of restraints. Also, their occupancies will be refined independently and will be constrained to be within 0 and 1, but it will not make sure that they add up to 1. If you want them to add up to 1, then make a parameter file with a constrained group as described in one of examples here:

"13 typical occupancy refinement scenarios and available options in phenix.refine" http://phenix-online.org/newsletter/

You can do it in the GUI too.

Let us know if you more questions or problems!

Pavel

On 3/2/18 02:57, Oleg Sobolev wrote:

...

Hi Jason,

The ligand and APHE got pushed away from one another and off of their respective densities. Occupancies for both were reduced to 0.5 for the refinement. What am I doing wrong here? Does the ligand need to have an altloc as well? But I don’t see an alternative conformation.

Yes, the ligand also needs an altloc, and it should be different from PHE altloc. This way the refinement will know that these atoms don't see each other. You don't have to put two alternative conformations for the ligand, you can have just one with partial occupancy (<1.0) which will mean that the ligand is not always there. Best regards, Oleg Sobolev.

...

On Feb 28, 2018, at 5:35 PM, van den Bedem, Henry mailto:[email protected]> wrote:

If the phe is a gatekeeper, shouldn’t the ligand be refined at partial occupancy; i.e. occupancies not necessarily have to ‘add up’ as you suggest? Maybe this link is helpful too:www.biorxiv.org/content/early/2018/01/24/253419 http://www.biorxiv.org/content/early/2018/01/24/253419 H *From:*mailto:[email protected]> on behalf of "Phan, Jason" mailto:[email protected]> *Date:*Wednesday, February 28, 2018 at 1:30 PM *To:*"[email protected] mailto:[email protected]" mailto:[email protected]> *Subject:*[phenixbb] Overlapping ligand-protein side chain density All, A piece of a ligand and the side chain of a “gate-keeper” Phe occupy the same space with well-defined features observed for both (resolution is 1.7 A). It looks about 50:50. How do you refine both ligand and protein side chain in this case? A couple of phenixbb suggestions for dealing with ligand-ligand overlapping density have been considered. With the first suggestion, the two entities still clashed and moved apart off of their respective densities. The second suggestion is not applicable in this case since the other molecule is not a ligand but part of the protein but it was tested anyway. Although there was no bumping, the occupancies don’t add up, resulting in a big blob of negative density around the ligand piece. ------------------------------ Only two comments:

1. At that resolution, constrained group occupancy refinement should work reasonably well (provided you can model the 2 entities). Then you also do not have clashes between the molecules, because Occ(A)+Occ(B)=1, meaning when one (A) is there, the other one (B) is not. This works with refmac (external keyword file); if you need more sophisticated occupancy re/constraints SHELXL may offer more opportunities. 2. There is no necessity for the two NCS copies of the binding site to look exactly the same (non-equivalent). Maybe there is a good reason/story (accessibility, contacts etc) for one site to be occupied differently than the other one.

Best, BR —————————————— Modeling two molecules that occupy overlapping binding sites in a structure simply involves designating them as alternate conformers, with the same chain and residue number, and an occupancy that sums to 1.0. For example, if you have an AMP and an ADP that occupy the same binding site, you would define them as AAMP B 501 BADP B 501 and initially set the occupancies for the atoms in each conformer to the ratio (50:50, 30:70, etc.) that you observe in the density. Refinement in this manner is straightforward in PHENIX. Diana ----------------------------

_______________________________________________ phenixbb mailing list [email protected] mailto:[email protected] http://phenix-online.org/mailman/listinfo/phenixbb http://phenix-online.org/mailman/listinfo/phenixbb Unsubscribe: [email protected] mailto:[email protected]

_______________________________________________ phenixbb mailing list [email protected] http://phenix-online.org/mailman/listinfo/phenixbb Unsubscribe:[email protected]

Hi Jason, "I did not force the occupancies to add up to 1 but the program did it anyway." If you use constrained groups (as defined in the parameter file you quote), the occupancy sum will be constrained to 1. Best wishes, Dorothee On Thu, Mar 1, 2018 at 3:35 PM, Phan, Jason wrote:

Thank you, Edward and Oleg for the suggestions, and Pavel for the clarification. You need two selections in the constrained group and an altloc for the ligand that’s different from the protein side chain. I did not force the occupancies to add up to 1 but the program did it anyway. Phenix refined the ligand’s occupancy to 0.68 and the PHE side chain to 0.32.

This is what I did:

I did not make an alternate conformation for the PHE, so there’s no BPHE. I just added the lock to the side chain starting at carbon beta. So, I have APHE for the side chain and BLIG for all the ligand atoms and reduced their occupancies to < 1.

This is the parameter file:

refinement { refine { occupancies { constrained_group { selection = (chain B and resseq 890) and altloc A selection = resname LIG and altloc B } } } }

The rationale is in one population, the ligand is not bound and the PHE occupies the pocket. In the other, the ligand is bound the the PHE is out somewhere that I cannot observe and model, thus I did not model that PHE-out conformation. The refinement went to completion with the occupancies refined to 0.68 and 0.32 for BLIG and APHE, respectively. R-values dropped by a few tenths of a percent. Please let me know if there’s a better way to do this refinement.

@Pavel, there’s repulsion between the selections if I don’t put an altloc on the ligand as suggested by Oleg.

Jason

On Mar 1, 2018, at 4:59 PM, Pavel Afonine wrote:

Hi Jason,

Oleg is correct, you can assign the ligand and PHE different non-blanc altlocs (say A and B). In this case PHE and ligand will not 'see' each other via repulsion term of restraints. Also, their occupancies will be refined independently and will be constrained to be within 0 and 1, but it will not make sure that they add up to 1. If you want them to add up to 1, then make a parameter file with a constrained group as described in one of examples here:

"13 typical occupancy refinement scenarios and available options in phenix.refine" http://phenix-online.org/newsletter/ https://na01.safelinks.protection.outlook.com/?url=http%3A%2F%2Fphenix-online.org%2Fnewsletter%2F&data=02%7C01%7Cjason.phan%40vanderbilt.edu%7C40b308da75d345eb3cf408d57fc82098%7Cba5a7f39e3be4ab3b45067fa80faecad%7C0%7C1%7C636555419887467714&sdata=5VYe7ZcJpoBjUZVV7VTStHt2sRTUn1W4z%2FztPNHyGpE%3D&reserved=0

You can do it in the GUI too.

Let us know if you more questions or problems!

Pavel

On 3/2/18 02:57, Oleg Sobolev wrote:

Hi Jason,

...

The ligand and APHE got pushed away from one another and off of their respective densities. Occupancies for both were reduced to 0.5 for the refinement. What am I doing wrong here? Does the ligand need to have an altloc as well? But I don’t see an alternative conformation.

Yes, the ligand also needs an altloc, and it should be different from PHE altloc. This way the refinement will know that these atoms don't see each other. You don't have to put two alternative conformations for the ligand, you can have just one with partial occupancy (<1.0) which will mean that the ligand is not always there.

Best regards, Oleg Sobolev.

...

On Feb 28, 2018, at 5:35 PM, van den Bedem, Henry < [email protected]> wrote:

If the phe is a gatekeeper, shouldn’t the ligand be refined at partial occupancy; i.e. occupancies not necessarily have to ‘add up’ as you suggest? Maybe this link is helpful too: www.biorxiv.org/content/e arly/2018/01/24/253419 https://na01.safelinks.protection.outlook.com/?url=http%3A%2F%2Fwww.biorxiv.org%2Fcontent%2Fearly%2F2018%2F01%2F24%2F253419&data=02%7C01%7Cjason.phan%40vanderbilt.edu%7C40b308da75d345eb3cf408d57fc82098%7Cba5a7f39e3be4ab3b45067fa80faecad%7C0%7C1%7C636555419887467714&sdata=%2BeF9t7TNZ72bkVoY8GqaBvuNd1EOp%2BaRRmhlcB8OZ1g%3D&reserved=0

H

*From: * on behalf of "Phan, Jason" < [email protected]> *Date: *Wednesday, February 28, 2018 at 1:30 PM *To: *"[email protected]" *Subject: *[phenixbb] Overlapping ligand-protein side chain density

All,

A piece of a ligand and the side chain of a “gate-keeper” Phe occupy the same space with well-defined features observed for both (resolution is 1.7 A). It looks about 50:50. How do you refine both ligand and protein side chain in this case? A couple of phenixbb suggestions for dealing with ligand-ligand overlapping density have been considered. With the first suggestion, the two entities still clashed and moved apart off of their respective densities. The second suggestion is not applicable in this case since the other molecule is not a ligand but part of the protein but it was tested anyway. Although there was no bumping, the occupancies don’t add up, resulting in a big blob of negative density around the ligand piece.

------------------------------ Only two comments:

1. At that resolution, constrained group occupancy refinement should work reasonably well (provided you can model the 2 entities). Then you also do not have clashes between the molecules, because Occ(A)+Occ(B)=1, meaning when one (A) is there, the other one (B) is not. This works with refmac (external keyword file); if you need more sophisticated occupancy re/constraints SHELXL may offer more opportunities. 2. There is no necessity for the two NCS copies of the binding site to look exactly the same (non-equivalent). Maybe there is a good reason/story (accessibility, contacts etc) for one site to be occupied differently than the other one.

Best, BR —————————————— Modeling two molecules that occupy overlapping binding sites in a structure simply involves designating them as alternate conformers, with the same chain and residue number, and an occupancy that sums to 1.0. For example, if you have an AMP and an ADP that occupy the same binding site, you would define them as

AAMP B 501 BADP B 501

and initially set the occupancies for the atoms in each conformer to the ratio (50:50, 30:70, etc.) that you observe in the density.

Refinement in this manner is straightforward in PHENIX.

Diana ----------------------------

_______________________________________________ phenixbb mailing list [email protected] http://phenix-online.org/mailman/listinfo/phenixbb https://na01.safelinks.protection.outlook.com/?url=http%3A%2F%2Fphenix-online.org%2Fmailman%2Flistinfo%2Fphenixbb&data=02%7C01%7Cjason.phan%40vanderbilt.edu%7C40b308da75d345eb3cf408d57fc82098%7Cba5a7f39e3be4ab3b45067fa80faecad%7C0%7C1%7C636555419887467714&sdata=nIHAg7svGzV23ChHuKdhLwdoEow358Mz7wadZ3TczWk%3D&reserved=0 Unsubscribe: [email protected]

_______________________________________________ phenixbb mailing [email protected]http://phenix-online.org/mailman/listinfo/phenixbb https://na01.safelinks.protection.outlook.com/?url=http%3A%2F%2Fphenix-online.org%2Fmailman%2Flistinfo%2Fphenixbb&data=02%7C01%7Cjason.phan%40vanderbilt.edu%7C40b308da75d345eb3cf408d57fc82098%7Cba5a7f39e3be4ab3b45067fa80faecad%7C0%7C1%7C636555419887467714&sdata=nIHAg7svGzV23ChHuKdhLwdoEow358Mz7wadZ3TczWk%3D&reserved=0 Unsubscribe: [email protected]

_______________________________________________ phenixbb mailing list [email protected] http://phenix-online.org/mailman/listinfo/phenixbb Unsubscribe: [email protected]