Hi, Are there any tools in ChimeraX suitable for clustering/analyzing conformational ensembles, e.g. as generated by AFSAMPLE2 (https://www.nature.com/articles/s42003-025-07791-9) or similar approaches? The closest I could find was this: https://www.rbvi.ucsf.edu/chimerax/data/simstruct-doc-oct2024/simstruct.html The similarstructures interface looks great, but it seems like this can only analyze sets of structures generated in the tool - is there any way to use it to analyze an arbitrary set of structures? Basically I would like a way to (for a set of say 200 input structures of the same sequence) quickly align them and then cluster them, to identify and group representative conformational states in the ensemble of predictions. Cheers Oli
Looking at the daily builds, it seems like md cluster might do the trick, if the set of structures is opened as a trajectory - it is not exactly clear to me how to do that though, for a set of individual PDB files as opposed to an MD trajectory or NMR multi-model ensemble. Also, the PDBs generated by alphafold are in arbitrary orientations. Is there a way to open, align and cluster a large set (100s) of structures in arbitrary orientations? Cheers Oli
On Mar 7, 2026, at 3:38 PM, Oliver Clarke <olibclarke@gmail.com> wrote:
Hi,
Are there any tools in ChimeraX suitable for clustering/analyzing conformational ensembles, e.g. as generated by AFSAMPLE2 (https://www.nature.com/articles/s42003-025-07791-9) or similar approaches?
The closest I could find was this:
https://www.rbvi.ucsf.edu/chimerax/data/simstruct-doc-oct2024/simstruct.html
The similarstructures interface looks great, but it seems like this can only analyze sets of structures generated in the tool - is there any way to use it to analyze an arbitrary set of structures?
Basically I would like a way to (for a set of say 200 input structures of the same sequence) quickly align them and then cluster them, to identify and group representative conformational states in the ensemble of predictions.
Cheers Oli
Ok, so I figured out how to make an aligned ensemble - matchmaker, combine then split - and I now have an aligned 100-member ensemble (pseudo-NMR-like). However, when I try to analyze this using coordset (open ensemble.pdb coordsets true), nothing is displayed, and ChimeraX launches one instance of the coordset slider interface for each member of the ensemble! Is this the intended behavior, or am I doing something stupid? Putting the same ensemble pdb into "old" chimera and running ensemble clustering seems to work fine, so I don't think it is an issue with the ensemble itself Cheers Oli
On Mar 8, 2026, at 9:20 AM, Oliver Clarke <olibclarke@gmail.com> wrote:
Looking at the daily builds, it seems like md cluster might do the trick, if the set of structures is opened as a trajectory - it is not exactly clear to me how to do that though, for a set of individual PDB files as opposed to an MD trajectory or NMR multi-model ensemble.
Also, the PDBs generated by alphafold are in arbitrary orientations. Is there a way to open, align and cluster a large set (100s) of structures in arbitrary orientations?
Cheers Oli
On Mar 7, 2026, at 3:38 PM, Oliver Clarke <olibclarke@gmail.com> wrote:
Hi,
Are there any tools in ChimeraX suitable for clustering/analyzing conformational ensembles, e.g. as generated by AFSAMPLE2 (https://www.nature.com/articles/s42003-025-07791-9) or similar approaches?
The closest I could find was this:
https://www.rbvi.ucsf.edu/chimerax/data/simstruct-doc-oct2024/simstruct.html
The similarstructures interface looks great, but it seems like this can only analyze sets of structures generated in the tool - is there any way to use it to analyze an arbitrary set of structures?
Basically I would like a way to (for a set of say 200 input structures of the same sequence) quickly align them and then cluster them, to identify and group representative conformational states in the ensemble of predictions.
Cheers Oli
Hi Oli, In your file, each model has a different chain ID, so ChimeraX doesn’t find any residue of the second/third/etc. model to be the same as a residue of the first model, so creates one “trajectory” per model. I’ll try to improve this behavior and will update the ticket you created when I have something. My suggestion for now is to skip the combine/split steps: just guarantee the chain names are all the same after you do matchmaker (using the changechains command as needed) and then save as a multi-MODEL PDB file. --Eric Eric Pettersen UCSF Computer Graphics Lab
On Mar 8, 2026, at 6:56 AM, Oliver Clarke via ChimeraX-users <chimerax-users@cgl.ucsf.edu> wrote:
Ok, so I figured out how to make an aligned ensemble - matchmaker, combine then split - and I now have an aligned 100-member ensemble (pseudo-NMR-like).
However, when I try to analyze this using coordset (open ensemble.pdb coordsets true), nothing is displayed, and ChimeraX launches one instance of the coordset slider interface for each member of the ensemble! Is this the intended behavior, or am I doing something stupid?
Putting the same ensemble pdb into "old" chimera and running ensemble clustering seems to work fine, so I don't think it is an issue with the ensemble itself
Cheers Oli
On Mar 8, 2026, at 9:20 AM, Oliver Clarke <olibclarke@gmail.com> wrote:
Looking at the daily builds, it seems like md cluster might do the trick, if the set of structures is opened as a trajectory - it is not exactly clear to me how to do that though, for a set of individual PDB files as opposed to an MD trajectory or NMR multi-model ensemble.
Also, the PDBs generated by alphafold are in arbitrary orientations. Is there a way to open, align and cluster a large set (100s) of structures in arbitrary orientations?
Cheers Oli
On Mar 7, 2026, at 3:38 PM, Oliver Clarke <olibclarke@gmail.com> wrote:
Hi,
Are there any tools in ChimeraX suitable for clustering/analyzing conformational ensembles, e.g. as generated by AFSAMPLE2 (https://www.nature.com/articles/s42003-025-07791-9) or similar approaches?
The closest I could find was this:
https://www.rbvi.ucsf.edu/chimerax/data/simstruct-doc-oct2024/simstruct.html
The similarstructures interface looks great, but it seems like this can only analyze sets of structures generated in the tool - is there any way to use it to analyze an arbitrary set of structures?
Basically I would like a way to (for a set of say 200 input structures of the same sequence) quickly align them and then cluster them, to identify and group representative conformational states in the ensemble of predictions.
Cheers Oli
_______________________________________________ ChimeraX-users mailing list -- chimerax-users@cgl.ucsf.edu To unsubscribe send an email to chimerax-users-leave@cgl.ucsf.edu Archives: https://mail.cgl.ucsf.edu/mailman/archives/list/chimerax-users@cgl.ucsf.edu/
Hi Oli, The ChimeraX Similar Structures capability is intended for looking at large sets (hundreds or thousands) of distant homolog structures found by FoldSeek structure-based searches. It offers clustering that uses x,y,z coordinates of specified residues (defaulting to the top several conserved residues) using UMAP. Because the intended use was with large numbers of homologs it uses both sequence alignments, and is optimized to use C-alpha coordinates provided by Foldseek instead of trying to load a thousand full structure models. In theory this could be used for your much simpler case of structures all having the same sequence. But currently there is no code to convert a trajectory into the data structures with sequence alignments and C-alpha coordinates that are currently used. It wouldn't take much code to adapt to your case. I'm not sure how useful it would be because it requires you to figure out which residues to use for clustering conformations. Using UMAP to reduce to reduce a high dimensional set of residue x,y,z coordinates to 2D was an experiment. It is not even clustering, just reducing to 2D for visualization. So I'm not sure this experimental approach designed for distant homologs is the best for your case where I guess many clustering algorithms have been developed. Tom
On Mar 7, 2026, at 12:38 PM, Oliver Clarke via ChimeraX-users <chimerax-users@cgl.ucsf.edu> wrote:
Hi,
Are there any tools in ChimeraX suitable for clustering/analyzing conformational ensembles, e.g. as generated by AFSAMPLE2 (https://www.nature.com/articles/s42003-025-07791-9) or similar approaches?
The closest I could find was this:
https://www.rbvi.ucsf.edu/chimerax/data/simstruct-doc-oct2024/simstruct.html
The similarstructures interface looks great, but it seems like this can only analyze sets of structures generated in the tool - is there any way to use it to analyze an arbitrary set of structures?
Basically I would like a way to (for a set of say 200 input structures of the same sequence) quickly align them and then cluster them, to identify and group representative conformational states in the ensemble of predictions.
Cheers Oli _______________________________________________ ChimeraX-users mailing list -- chimerax-users@cgl.ucsf.edu To unsubscribe send an email to chimerax-users-leave@cgl.ucsf.edu Archives: https://mail.cgl.ucsf.edu/mailman/archives/list/chimerax-users@cgl.ucsf.edu/
Hi Eric, Thanks that did the trick! I didn't realize that ChimeraX automatically saves as a multimodel PDB if you have multiple pdbs selected. matchmaker --> save as multimodel --> open with coordsets true --> cluster frames does the trick for visualization. However, I can't see how to save/export the results? E.g. in "old" chimera, if I run ensemble cluster and select a cluster, it has an option to either select all frames in the cluster or a single representative frame, which I can then save as separate PDBs. Is there similar functionality in coordsets in ChimeraX? Cheers Oli 
On Mar 9, 2026, at 4:52 PM, Eric Pettersen <pett@cgl.ucsf.edu> wrote:
Hi Oli, In your file, each model has a different chain ID, so ChimeraX doesn’t find any residue of the second/third/etc. model to be the same as a residue of the first model, so creates one “trajectory” per model. I’ll try to improve this behavior and will update the ticket you created when I have something. My suggestion for now is to skip the combine/split steps: just guarantee the chain names are all the same after you do matchmaker (using the changechains command as needed) and then save as a multi-MODEL PDB file.
--Erics
Eric Pettersen UCSF Computer Graphics Lab
On Mar 8, 2026, at 6:56 AM, Oliver Clarke via ChimeraX-users <chimerax-users@cgl.ucsf.edu> wrote:
Ok, so I figured out how to make an aligned ensemble - matchmaker, combine then split - and I now have an aligned 100-member ensemble (pseudo-NMR-like).
However, when I try to analyze this using coordset (open ensemble.pdb coordsets true), nothing is displayed, and ChimeraX launches one instance of the coordset slider interface for each member of the ensemble! Is this the intended behavior, or am I doing something stupid?
Putting the same ensemble pdb into "old" chimera and running ensemble clustering seems to work fine, so I don't think it is an issue with the ensemble itself
Cheers Oli
On Mar 8, 2026, at 9:20 AM, Oliver Clarke <olibclarke@gmail.com> wrote:
Looking at the daily builds, it seems like md cluster might do the trick, if the set of structures is opened as a trajectory - it is not exactly clear to me how to do that though, for a set of individual PDB files as opposed to an MD trajectory or NMR multi-model ensemble.
Also, the PDBs generated by alphafold are in arbitrary orientations. Is there a way to open, align and cluster a large set (100s) of structures in arbitrary orientations?
Cheers Oli
On Mar 7, 2026, at 3:38 PM, Oliver Clarke <olibclarke@gmail.com> wrote:
Hi,
Are there any tools in ChimeraX suitable for clustering/analyzing conformational ensembles, e.g. as generated by AFSAMPLE2 (https://www.nature.com/articles/s42003-025-07791-9) or similar approaches?
The closest I could find was this:
https://www.rbvi.ucsf.edu/chimerax/data/simstruct-doc-oct2024/simstruct.html
The similarstructures interface looks great, but it seems like this can only analyze sets of structures generated in the tool - is there any way to use it to analyze an arbitrary set of structures?
Basically I would like a way to (for a set of say 200 input structures of the same sequence) quickly align them and then cluster them, to identify and group representative conformational states in the ensemble of predictions.
Cheers Oli
_______________________________________________ ChimeraX-users mailing list -- chimerax-users@cgl.ucsf.edu To unsubscribe send an email to chimerax-users-leave@cgl.ucsf.edu Archives: https://mail.cgl.ucsf.edu/mailman/archives/list/chimerax-users@cgl.ucsf.edu/
Hi Oli, I am mystified. What in old Chimera allowed you to select all frames in a cluster (what does that even mean?). In the ChimeraX version, choosing a cluster will jump to the representative frame, which you could save using the usual PDB-saving functions in ChimeraX. Also, the Save button at the bottom of the dialog will allow you to save a list of the frames in each cluster, with the representative frame listed first. This is all the same as the Chimera version AFAIK. --Eric
On Mar 10, 2026, at 6:00 AM, Oliver Clarke via ChimeraX-users <chimerax-users@cgl.ucsf.edu> wrote:
Hi Eric,
Thanks that did the trick! I didn't realize that ChimeraX automatically saves as a multimodel PDB if you have multiple pdbs selected. matchmaker --> save as multimodel --> open with coordsets true --> cluster frames does the trick for visualization.
However, I can't see how to save/export the results? E.g. in "old" chimera, if I run ensemble cluster and select a cluster, it has an option to either select all frames in the cluster or a single representative frame, which I can then save as separate PDBs. Is there similar functionality in coordsets in ChimeraX?
Cheers Oli
<PastedGraphic-1.png>
On Mar 9, 2026, at 4:52 PM, Eric Pettersen <pett@cgl.ucsf.edu> wrote:
Hi Oli, In your file, each model has a different chain ID, so ChimeraX doesn’t find any residue of the second/third/etc. model to be the same as a residue of the first model, so creates one “trajectory” per model. I’ll try to improve this behavior and will update the ticket you created when I have something. My suggestion for now is to skip the combine/split steps: just guarantee the chain names are all the same after you do matchmaker (using the changechains command as needed) and then save as a multi-MODEL PDB file.
--Erics
Eric Pettersen UCSF Computer Graphics Lab
On Mar 8, 2026, at 6:56 AM, Oliver Clarke via ChimeraX-users <chimerax-users@cgl.ucsf.edu> wrote:
Ok, so I figured out how to make an aligned ensemble - matchmaker, combine then split - and I now have an aligned 100-member ensemble (pseudo-NMR-like).
However, when I try to analyze this using coordset (open ensemble.pdb coordsets true), nothing is displayed, and ChimeraX launches one instance of the coordset slider interface for each member of the ensemble! Is this the intended behavior, or am I doing something stupid?
Putting the same ensemble pdb into "old" chimera and running ensemble clustering seems to work fine, so I don't think it is an issue with the ensemble itself
Cheers Oli
On Mar 8, 2026, at 9:20 AM, Oliver Clarke <olibclarke@gmail.com> wrote:
Looking at the daily builds, it seems like md cluster might do the trick, if the set of structures is opened as a trajectory - it is not exactly clear to me how to do that though, for a set of individual PDB files as opposed to an MD trajectory or NMR multi-model ensemble.
Also, the PDBs generated by alphafold are in arbitrary orientations. Is there a way to open, align and cluster a large set (100s) of structures in arbitrary orientations?
Cheers Oli
On Mar 7, 2026, at 3:38 PM, Oliver Clarke <olibclarke@gmail.com> wrote:
Hi,
Are there any tools in ChimeraX suitable for clustering/analyzing conformational ensembles, e.g. as generated by AFSAMPLE2 (https://www.nature.com/articles/s42003-025-07791-9) or similar approaches?
The closest I could find was this:
https://www.rbvi.ucsf.edu/chimerax/data/simstruct-doc-oct2024/simstruct.html
The similarstructures interface looks great, but it seems like this can only analyze sets of structures generated in the tool - is there any way to use it to analyze an arbitrary set of structures?
Basically I would like a way to (for a set of say 200 input structures of the same sequence) quickly align them and then cluster them, to identify and group representative conformational states in the ensemble of predictions.
Cheers Oli
_______________________________________________ ChimeraX-users mailing list -- chimerax-users@cgl.ucsf.edu To unsubscribe send an email to chimerax-users-leave@cgl.ucsf.edu Archives: https://mail.cgl.ucsf.edu/mailman/archives/list/chimerax-users@cgl.ucsf.edu/
_______________________________________________ ChimeraX-users mailing list -- chimerax-users@cgl.ucsf.edu To unsubscribe send an email to chimerax-users-leave@cgl.ucsf.edu Archives: https://mail.cgl.ucsf.edu/mailman/archives/list/chimerax-users@cgl.ucsf.edu/
Hi Eric, See screenshot - in chimera there is an option in Ensemble Cluster to (1) color members of the cluster and (2) select either all members or a representative member in the model panel. Then one can click "write pdb" to save them. I don't see any similar option in ChimeraX? Just the option (which you mentioned) to write out a list in a text file, but this is not quite so convenient Cheers Oli 
On Mar 10, 2026, at 1:51 PM, Eric Pettersen <pett@cgl.ucsf.edu> wrote:
Hi Oli, I am mystified. What in old Chimera allowed you to select all frames in a cluster (what does that even mean?). In the ChimeraX version, choosing a cluster will jump to the representative frame, which you could save using the usual PDB-saving functions in ChimeraX. Also, the Save button at the bottom of the dialog will allow you to save a list of the frames in each cluster, with the representative frame listed first. This is all the same as the Chimera version AFAIK.
--Eric
On Mar 10, 2026, at 6:00 AM, Oliver Clarke via ChimeraX-users <chimerax-users@cgl.ucsf.edu> wrote:
Hi Eric,
Thanks that did the trick! I didn't realize that ChimeraX automatically saves as a multimodel PDB if you have multiple pdbs selected. matchmaker --> save as multimodel --> open with coordsets true --> cluster frames does the trick for visualization.
However, I can't see how to save/export the results? E.g. in "old" chimera, if I run ensemble cluster and select a cluster, it has an option to either select all frames in the cluster or a single representative frame, which I can then save as separate PDBs. Is there similar functionality in coordsets in ChimeraX?
Cheers Oli
<PastedGraphic-1.png>
On Mar 9, 2026, at 4:52 PM, Eric Pettersen <pett@cgl.ucsf.edu> wrote:
Hi Oli, In your file, each model has a different chain ID, so ChimeraX doesn’t find any residue of the second/third/etc. model to be the same as a residue of the first model, so creates one “trajectory” per model. I’ll try to improve this behavior and will update the ticket you created when I have something. My suggestion for now is to skip the combine/split steps: just guarantee the chain names are all the same after you do matchmaker (using the changechains command as needed) and then save as a multi-MODEL PDB file.
--Erics
Eric Pettersen UCSF Computer Graphics Lab
On Mar 8, 2026, at 6:56 AM, Oliver Clarke via ChimeraX-users <chimerax-users@cgl.ucsf.edu> wrote:
Ok, so I figured out how to make an aligned ensemble - matchmaker, combine then split - and I now have an aligned 100-member ensemble (pseudo-NMR-like).
However, when I try to analyze this using coordset (open ensemble.pdb coordsets true), nothing is displayed, and ChimeraX launches one instance of the coordset slider interface for each member of the ensemble! Is this the intended behavior, or am I doing something stupid?
Putting the same ensemble pdb into "old" chimera and running ensemble clustering seems to work fine, so I don't think it is an issue with the ensemble itself
Cheers Oli
On Mar 8, 2026, at 9:20 AM, Oliver Clarke <olibclarke@gmail.com> wrote:
Looking at the daily builds, it seems like md cluster might do the trick, if the set of structures is opened as a trajectory - it is not exactly clear to me how to do that though, for a set of individual PDB files as opposed to an MD trajectory or NMR multi-model ensemble.
Also, the PDBs generated by alphafold are in arbitrary orientations. Is there a way to open, align and cluster a large set (100s) of structures in arbitrary orientations?
Cheers Oli
On Mar 7, 2026, at 3:38 PM, Oliver Clarke <olibclarke@gmail.com> wrote:
Hi,
Are there any tools in ChimeraX suitable for clustering/analyzing conformational ensembles, e.g. as generated by AFSAMPLE2 (https://www.nature.com/articles/s42003-025-07791-9) or similar approaches?
The closest I could find was this:
https://www.rbvi.ucsf.edu/chimerax/data/simstruct-doc-oct2024/simstruct.html
The similarstructures interface looks great, but it seems like this can only analyze sets of structures generated in the tool - is there any way to use it to analyze an arbitrary set of structures?
Basically I would like a way to (for a set of say 200 input structures of the same sequence) quickly align them and then cluster them, to identify and group representative conformational states in the ensemble of predictions.
Cheers Oli
_______________________________________________ ChimeraX-users mailing list -- chimerax-users@cgl.ucsf.edu To unsubscribe send an email to chimerax-users-leave@cgl.ucsf.edu Archives: https://mail.cgl.ucsf.edu/mailman/archives/list/chimerax-users@cgl.ucsf.edu/
_______________________________________________ ChimeraX-users mailing list -- chimerax-users@cgl.ucsf.edu To unsubscribe send an email to chimerax-users-leave@cgl.ucsf.edu Archives: https://mail.cgl.ucsf.edu/mailman/archives/list/chimerax-users@cgl.ucsf.edu/
Hi Oli, Trajectory clustering and ensemble clustering are separate tools/capabilities, though they both use the same underlying clustering library to perform the clustering. I have implemented trajectory clustering but have not yet implemented ensemble clustering. Since the underlying library is available, implementing ensemble clustering would be less work than otherwise, but nonetheless its implementation competes with a lot of other priorities as usual. I will open a feature-request ticket in our database for implementing ensemble clustering, but it probably will not happen in the kind of time frame that helps with your immediate problem. In the interim what you could do is take the file that the current Save button gives you and in a text editor (or using a script) change the lists of frame numbers into "save" commands, e.g. change: 7 2 4 14 17 into: save cluster1.pdb models #1.7,2,4,14,17 Then you would need to open your multi-model PDB not as a trajectory (i.e. omit "coords t"). Then the command file you made would work to save the clusters into separate PDB files. --Eric
On Mar 10, 2026, at 11:28 AM, Oliver Clarke <olibclarke@gmail.com> wrote:
Hi Eric,
See screenshot - in chimera there is an option in Ensemble Cluster to (1) color members of the cluster and (2) select either all members or a representative member in the model panel. Then one can click "write pdb" to save them. I don't see any similar option in ChimeraX? Just the option (which you mentioned) to write out a list in a text file, but this is not quite so convenient
Cheers Oli <PastedGraphic-1.png>
On Mar 10, 2026, at 1:51 PM, Eric Pettersen <pett@cgl.ucsf.edu> wrote:
Hi Oli, I am mystified. What in old Chimera allowed you to select all frames in a cluster (what does that even mean?). In the ChimeraX version, choosing a cluster will jump to the representative frame, which you could save using the usual PDB-saving functions in ChimeraX. Also, the Save button at the bottom of the dialog will allow you to save a list of the frames in each cluster, with the representative frame listed first. This is all the same as the Chimera version AFAIK.
--Eric
On Mar 10, 2026, at 6:00 AM, Oliver Clarke via ChimeraX-users <chimerax-users@cgl.ucsf.edu> wrote:
Hi Eric,
Thanks that did the trick! I didn't realize that ChimeraX automatically saves as a multimodel PDB if you have multiple pdbs selected. matchmaker --> save as multimodel --> open with coordsets true --> cluster frames does the trick for visualization.
However, I can't see how to save/export the results? E.g. in "old" chimera, if I run ensemble cluster and select a cluster, it has an option to either select all frames in the cluster or a single representative frame, which I can then save as separate PDBs. Is there similar functionality in coordsets in ChimeraX?
Cheers Oli
<PastedGraphic-1.png>
On Mar 9, 2026, at 4:52 PM, Eric Pettersen <pett@cgl.ucsf.edu> wrote:
Hi Oli, In your file, each model has a different chain ID, so ChimeraX doesn’t find any residue of the second/third/etc. model to be the same as a residue of the first model, so creates one “trajectory” per model. I’ll try to improve this behavior and will update the ticket you created when I have something. My suggestion for now is to skip the combine/split steps: just guarantee the chain names are all the same after you do matchmaker (using the changechains command as needed) and then save as a multi-MODEL PDB file.
--Erics
Eric Pettersen UCSF Computer Graphics Lab
On Mar 8, 2026, at 6:56 AM, Oliver Clarke via ChimeraX-users <chimerax-users@cgl.ucsf.edu> wrote:
Ok, so I figured out how to make an aligned ensemble - matchmaker, combine then split - and I now have an aligned 100-member ensemble (pseudo-NMR-like).
However, when I try to analyze this using coordset (open ensemble.pdb coordsets true), nothing is displayed, and ChimeraX launches one instance of the coordset slider interface for each member of the ensemble! Is this the intended behavior, or am I doing something stupid?
Putting the same ensemble pdb into "old" chimera and running ensemble clustering seems to work fine, so I don't think it is an issue with the ensemble itself
Cheers Oli
On Mar 8, 2026, at 9:20 AM, Oliver Clarke <olibclarke@gmail.com> wrote:
Looking at the daily builds, it seems like md cluster might do the trick, if the set of structures is opened as a trajectory - it is not exactly clear to me how to do that though, for a set of individual PDB files as opposed to an MD trajectory or NMR multi-model ensemble.
Also, the PDBs generated by alphafold are in arbitrary orientations. Is there a way to open, align and cluster a large set (100s) of structures in arbitrary orientations?
Cheers Oli
> On Mar 7, 2026, at 3:38 PM, Oliver Clarke <olibclarke@gmail.com> wrote: > > Hi, > > Are there any tools in ChimeraX suitable for clustering/analyzing conformational ensembles, e.g. as generated by AFSAMPLE2 (https://www.nature.com/articles/s42003-025-07791-9) or similar approaches? > > The closest I could find was this: > > https://www.rbvi.ucsf.edu/chimerax/data/simstruct-doc-oct2024/simstruct.html > > The similarstructures interface looks great, but it seems like this can only analyze sets of structures generated in the tool - is there any way to use it to analyze an arbitrary set of structures? > > Basically I would like a way to (for a set of say 200 input structures of the same sequence) quickly align them and then cluster them, to identify and group representative conformational states in the ensemble of predictions. > > Cheers > Oli
_______________________________________________ ChimeraX-users mailing list -- chimerax-users@cgl.ucsf.edu To unsubscribe send an email to chimerax-users-leave@cgl.ucsf.edu Archives: https://mail.cgl.ucsf.edu/mailman/archives/list/chimerax-users@cgl.ucsf.edu/
_______________________________________________ ChimeraX-users mailing list -- chimerax-users@cgl.ucsf.edu To unsubscribe send an email to chimerax-users-leave@cgl.ucsf.edu Archives: https://mail.cgl.ucsf.edu/mailman/archives/list/chimerax-users@cgl.ucsf.edu/
Thanks Eric - in the meantime I can also just use the functionality in old Chimera for the clustering aspect, but would def be nice to have in ChimeraX, as we are dealing with ensembles more than we used to with increased use of structure prediction & design tools. Thanks for the help as always! Cheers Oli
On Mar 10, 2026, at 5:12 PM, Eric Pettersen <pett@cgl.ucsf.edu> wrote:
Hi Oli, Trajectory clustering and ensemble clustering are separate tools/capabilities, though they both use the same underlying clustering library to perform the clustering. I have implemented trajectory clustering but have not yet implemented ensemble clustering. Since the underlying library is available, implementing ensemble clustering would be less work than otherwise, but nonetheless its implementation competes with a lot of other priorities as usual. I will open a feature-request ticket in our database for implementing ensemble clustering, but it probably will not happen in the kind of time frame that helps with your immediate problem. In the interim what you could do is take the file that the current Save button gives you and in a text editor (or using a script) change the lists of frame numbers into "save" commands, e.g. change:
7 2 4 14 17
into:
save cluster1.pdb models #1.7,2,4,14,17
Then you would need to open your multi-model PDB not as a trajectory (i.e. omit "coords t"). Then the command file you made would work to save the clusters into separate PDB files.
--Eric
On Mar 10, 2026, at 11:28 AM, Oliver Clarke <olibclarke@gmail.com> wrote:
Hi Eric,
See screenshot - in chimera there is an option in Ensemble Cluster to (1) color members of the cluster and (2) select either all members or a representative member in the model panel. Then one can click "write pdb" to save them. I don't see any similar option in ChimeraX? Just the option (which you mentioned) to write out a list in a text file, but this is not quite so convenient
Cheers Oli <PastedGraphic-1.png>
On Mar 10, 2026, at 1:51 PM, Eric Pettersen <pett@cgl.ucsf.edu> wrote:
Hi Oli, I am mystified. What in old Chimera allowed you to select all frames in a cluster (what does that even mean?). In the ChimeraX version, choosing a cluster will jump to the representative frame, which you could save using the usual PDB-saving functions in ChimeraX. Also, the Save button at the bottom of the dialog will allow you to save a list of the frames in each cluster, with the representative frame listed first. This is all the same as the Chimera version AFAIK.
--Eric
On Mar 10, 2026, at 6:00 AM, Oliver Clarke via ChimeraX-users <chimerax-users@cgl.ucsf.edu> wrote:
Hi Eric,
Thanks that did the trick! I didn't realize that ChimeraX automatically saves as a multimodel PDB if you have multiple pdbs selected. matchmaker --> save as multimodel --> open with coordsets true --> cluster frames does the trick for visualization.
However, I can't see how to save/export the results? E.g. in "old" chimera, if I run ensemble cluster and select a cluster, it has an option to either select all frames in the cluster or a single representative frame, which I can then save as separate PDBs. Is there similar functionality in coordsets in ChimeraX?
Cheers Oli
<PastedGraphic-1.png>
On Mar 9, 2026, at 4:52 PM, Eric Pettersen <pett@cgl.ucsf.edu> wrote:
Hi Oli, In your file, each model has a different chain ID, so ChimeraX doesn’t find any residue of the second/third/etc. model to be the same as a residue of the first model, so creates one “trajectory” per model. I’ll try to improve this behavior and will update the ticket you created when I have something. My suggestion for now is to skip the combine/split steps: just guarantee the chain names are all the same after you do matchmaker (using the changechains command as needed) and then save as a multi-MODEL PDB file.
--Erics
Eric Pettersen UCSF Computer Graphics Lab
On Mar 8, 2026, at 6:56 AM, Oliver Clarke via ChimeraX-users <chimerax-users@cgl.ucsf.edu> wrote:
Ok, so I figured out how to make an aligned ensemble - matchmaker, combine then split - and I now have an aligned 100-member ensemble (pseudo-NMR-like).
However, when I try to analyze this using coordset (open ensemble.pdb coordsets true), nothing is displayed, and ChimeraX launches one instance of the coordset slider interface for each member of the ensemble! Is this the intended behavior, or am I doing something stupid?
Putting the same ensemble pdb into "old" chimera and running ensemble clustering seems to work fine, so I don't think it is an issue with the ensemble itself
Cheers Oli
> On Mar 8, 2026, at 9:20 AM, Oliver Clarke <olibclarke@gmail.com> wrote: > > Looking at the daily builds, it seems like md cluster might do the trick, if the set of structures is opened as a trajectory - it is not exactly clear to me how to do that though, for a set of individual PDB files as opposed to an MD trajectory or NMR multi-model ensemble. > > Also, the PDBs generated by alphafold are in arbitrary orientations. Is there a way to open, align and cluster a large set (100s) of structures in arbitrary orientations? > > Cheers > Oli > > >> On Mar 7, 2026, at 3:38 PM, Oliver Clarke <olibclarke@gmail.com> wrote: >> >> Hi, >> >> Are there any tools in ChimeraX suitable for clustering/analyzing conformational ensembles, e.g. as generated by AFSAMPLE2 (https://www.nature.com/articles/s42003-025-07791-9) or similar approaches? >> >> The closest I could find was this: >> >> https://www.rbvi.ucsf.edu/chimerax/data/simstruct-doc-oct2024/simstruct.html >> >> The similarstructures interface looks great, but it seems like this can only analyze sets of structures generated in the tool - is there any way to use it to analyze an arbitrary set of structures? >> >> Basically I would like a way to (for a set of say 200 input structures of the same sequence) quickly align them and then cluster them, to identify and group representative conformational states in the ensemble of predictions. >> >> Cheers >> Oli >
_______________________________________________ ChimeraX-users mailing list -- chimerax-users@cgl.ucsf.edu To unsubscribe send an email to chimerax-users-leave@cgl.ucsf.edu Archives: https://mail.cgl.ucsf.edu/mailman/archives/list/chimerax-users@cgl.ucsf.edu/
_______________________________________________ ChimeraX-users mailing list -- chimerax-users@cgl.ucsf.edu To unsubscribe send an email to chimerax-users-leave@cgl.ucsf.edu Archives: https://mail.cgl.ucsf.edu/mailman/archives/list/chimerax-users@cgl.ucsf.edu/
Dear Eric, is it possible to save the frames of the trajectory without saving it as a pdb file and reopening it? Currently, my 64GB memory computer doesn't allow me to open such a large pdb file. Is it not possible to save just a selection of frames from the trajectory? I did not find a way to do it. Best regards Sasha -----Original Message----- From: Eric <chimerax-users@cgl.ucsf.edu> To: Oliver <olibclarke@gmail.com> Cc: ChimeraX <chimerax-users@cgl.ucsf.edu> Date: Tuesday, 10 March 2026 10:14 PM CET Subject: [chimerax-users] Re: Tools for clustering conformational ensembles? Hi Oli, Trajectory clustering and ensemble clustering are separate tools/capabilities, though they both use the same underlying clustering library to perform the clustering. I have implemented trajectory clustering but have not yet implemented ensemble clustering. Since the underlying library is available, implementing ensemble clustering would be less work than otherwise, but nonetheless its implementation competes with a lot of other priorities as usual. I will open a feature-request ticket in our database for implementing ensemble clustering, but it probably will not happen in the kind of time frame that helps with your immediate problem. In the interim what you could do is take the file that the current Save button gives you and in a text editor (or using a script) change the lists of frame numbers into "save" commands, e.g. change: 7 2 4 14 17 into: save cluster1.pdb models #1.7,2,4,14,17 Then you would need to open your multi-model PDB not as a trajectory (i.e. omit "coords t"). Then the command file you made would work to save the clusters into separate PDB files. --Eric On Mar 10, 2026, at 11:28 AM, Oliver Clarke <olibclarke@gmail.com> wrote: Hi Eric, See screenshot - in chimera there is an option in Ensemble Cluster to (1) color members of the cluster and (2) select either all members or a representative member in the model panel. Then one can click "write pdb" to save them. I don't see any similar option in ChimeraX? Just the option (which you mentioned) to write out a list in a text file, but this is not quite so convenient Cheers Oli <PastedGraphic-1.png> On Mar 10, 2026, at 1:51 PM, Eric Pettersen <pett@cgl.ucsf.edu> wrote: Hi Oli, I am mystified. What in old Chimera allowed you to select all frames in a cluster (what does that even mean?). In the ChimeraX version, choosing a cluster will jump to the representative frame, which you could save using the usual PDB-saving functions in ChimeraX. Also, the Save button at the bottom of the dialog will allow you to save a list of the frames in each cluster, with the representative frame listed first. This is all the same as the Chimera version AFAIK. --Eric On Mar 10, 2026, at 6:00 AM, Oliver Clarke via ChimeraX-users <chimerax-users@cgl.ucsf.edu> wrote: Hi Eric, Thanks that did the trick! I didn't realize that ChimeraX automatically saves as a multimodel PDB if you have multiple pdbs selected. matchmaker --> save as multimodel --> open with coordsets true --> cluster frames does the trick for visualization. However, I can't see how to save/export the results? E.g. in "old" chimera, if I run ensemble cluster and select a cluster, it has an option to either select all frames in the cluster or a single representative frame, which I can then save as separate PDBs. Is there similar functionality in coordsets in ChimeraX? Cheers Oli <PastedGraphic-1.png> On Mar 9, 2026, at 4:52 PM, Eric Pettersen <pett@cgl.ucsf.edu> wrote: Hi Oli, In your file, each model has a different chain ID, so ChimeraX doesn’t find any residue of the second/third/etc. model to be the same as a residue of the first model, so creates one “trajectory” per model. I’ll try to improve this behavior and will update the ticket you created when I have something. My suggestion for now is to skip the combine/split steps: just guarantee the chain names are all the same after you do matchmaker (using the changechains command as needed) and then save as a multi-MODEL PDB file. --Erics Eric Pettersen UCSF Computer Graphics Lab On Mar 8, 2026, at 6:56 AM, Oliver Clarke via ChimeraX-users <chimerax-users@cgl.ucsf.edu> wrote: Ok, so I figured out how to make an aligned ensemble - matchmaker, combine then split - and I now have an aligned 100-member ensemble (pseudo-NMR-like). However, when I try to analyze this using coordset (open ensemble.pdb coordsets true), nothing is displayed, and ChimeraX launches one instance of the coordset slider interface for each member of the ensemble! Is this the intended behavior, or am I doing something stupid? Putting the same ensemble pdb into "old" chimera and running ensemble clustering seems to work fine, so I don't think it is an issue with the ensemble itself Cheers Oli On Mar 8, 2026, at 9:20 AM, Oliver Clarke <olibclarke@gmail.com> wrote: Looking at the daily builds, it seems like md cluster might do the trick, if the set of structures is opened as a trajectory - it is not exactly clear to me how to do that though, for a set of individual PDB files as opposed to an MD trajectory or NMR multi-model ensemble. Also, the PDBs generated by alphafold are in arbitrary orientations. Is there a way to open, align and cluster a large set (100s) of structures in arbitrary orientations? Cheers Oli On Mar 7, 2026, at 3:38 PM, Oliver Clarke <olibclarke@gmail.com> wrote: Hi, Are there any tools in ChimeraX suitable for clustering/analyzing conformational ensembles, e.g. as generated by AFSAMPLE2 (https://www.nature.com/articles/s42003-025-07791-9) or similar approaches? The closest I could find was this: https://www.rbvi.ucsf.edu/chimerax/data/simstruct-doc-oct2024/simstruct.html The similarstructures interface looks great, but it seems like this can only analyze sets of structures generated in the tool - is there any way to use it to analyze an arbitrary set of structures? Basically I would like a way to (for a set of say 200 input structures of the same sequence) quickly align them and then cluster them, to identify and group representative conformational states in the ensemble of predictions. Cheers Oli _______________________________________________ ChimeraX-users mailing list -- chimerax-users@cgl.ucsf.edu To unsubscribe send an email to chimerax-users-leave@cgl.ucsf.edu Archives: https://mail.cgl.ucsf.edu/mailman/archives/list/chimerax-users@cgl.ucsf.edu/ _______________________________________________ ChimeraX-users mailing list -- chimerax-users@cgl.ucsf.edu To unsubscribe send an email to chimerax-users-leave@cgl.ucsf.edu Archives: https://mail.cgl.ucsf.edu/mailman/archives/list/chimerax-users@cgl.ucsf.edu/
Hi Sasha, You could go to a particular frame of trajectory with the ‘coordset’ command and then save just that frame as a PDB file. Anything more than a few frames would be tedious though (except possibly as a script). You can save all the frames of a trajectory as a DCD file (e.g. save ~/traj.dcd), which as a binary file is considerably smaller than the corresponding PDB file, and I imagine uses much less memory (though I haven’t explicitly checked the latter assertion). --Eric Eric Pettersen UCSF Computer Graphics Lab
On Mar 12, 2026, at 8:40 AM, Alexandra Zahradnikova via ChimeraX-users <chimerax-users@cgl.ucsf.edu> wrote:
Dear Eric, is it possible to save the frames of the trajectory without saving it as a pdb file and reopening it? Currently, my 64GB memory computer doesn't allow me to open such a large pdb file. Is it not possible to save just a selection of frames from the trajectory? I did not find a way to do it. Best regards Sasha
From: Eric <chimerax-users@cgl.ucsf.edu> To: Oliver <olibclarke@gmail.com> Cc: ChimeraX <chimerax-users@cgl.ucsf.edu> Date: Tuesday, 10 March 2026 10:14 PM CET Subject: [chimerax-users] Re: Tools for clustering conformational ensembles?
Hi Oli, Trajectory clustering and ensemble clustering are separate tools/capabilities, though they both use the same underlying clustering library to perform the clustering. I have implemented trajectory clustering but have not yet implemented ensemble clustering. Since the underlying library is available, implementing ensemble clustering would be less work than otherwise, but nonetheless its implementation competes with a lot of other priorities as usual. I will open a feature-request ticket in our database for implementing ensemble clustering, but it probably will not happen in the kind of time frame that helps with your immediate problem. In the interim what you could do is take the file that the current Save button gives you and in a text editor (or using a script) change the lists of frame numbers into "save" commands, e.g. change:
7 2 4 14 17
into:
save cluster1.pdb models #1.7,2,4,14,17
Then you would need to open your multi-model PDB not as a trajectory (i.e. omit "coords t"). Then the command file you made would work to save the clusters into separate PDB files.
--Eric
On Mar 10, 2026, at 11:28 AM, Oliver Clarke <olibclarke@gmail.com> wrote:
Hi Eric,
See screenshot - in chimera there is an option in Ensemble Cluster to (1) color members of the cluster and (2) select either all members or a representative member in the model panel. Then one can click "write pdb" to save them. I don't see any similar option in ChimeraX? Just the option (which you mentioned) to write out a list in a text file, but this is not quite so convenient
Cheers Oli <PastedGraphic-1.png>
On Mar 10, 2026, at 1:51 PM, Eric Pettersen <pett@cgl.ucsf.edu> wrote:
Hi Oli, I am mystified. What in old Chimera allowed you to select all frames in a cluster (what does that even mean?). In the ChimeraX version, choosing a cluster will jump to the representative frame, which you could save using the usual PDB-saving functions in ChimeraX. Also, the Save button at the bottom of the dialog will allow you to save a list of the frames in each cluster, with the representative frame listed first. This is all the same as the Chimera version AFAIK.
--Eric
On Mar 10, 2026, at 6:00 AM, Oliver Clarke via ChimeraX-users <chimerax-users@cgl.ucsf.edu> wrote:
Hi Eric,
Thanks that did the trick! I didn't realize that ChimeraX automatically saves as a multimodel PDB if you have multiple pdbs selected. matchmaker --> save as multimodel --> open with coordsets true --> cluster frames does the trick for visualization.
However, I can't see how to save/export the results? E.g. in "old" chimera, if I run ensemble cluster and select a cluster, it has an option to either select all frames in the cluster or a single representative frame, which I can then save as separate PDBs. Is there similar functionality in coordsets in ChimeraX?
Cheers Oli
<PastedGraphic-1.png>
On Mar 9, 2026, at 4:52 PM, Eric Pettersen <pett@cgl.ucsf.edu> wrote:
Hi Oli, In your file, each model has a different chain ID, so ChimeraX doesn’t find any residue of the second/third/etc. model to be the same as a residue of the first model, so creates one “trajectory” per model. I’ll try to improve this behavior and will update the ticket you created when I have something. My suggestion for now is to skip the combine/split steps: just guarantee the chain names are all the same after you do matchmaker (using the changechains command as needed) and then save as a multi-MODEL PDB file.
--Erics
Eric Pettersen UCSF Computer Graphics Lab
On Mar 8, 2026, at 6:56 AM, Oliver Clarke via ChimeraX-users <chimerax-users@cgl.ucsf.edu> wrote:
Ok, so I figured out how to make an aligned ensemble - matchmaker, combine then split - and I now have an aligned 100-member ensemble (pseudo-NMR-like).
However, when I try to analyze this using coordset (open ensemble.pdb coordsets true), nothing is displayed, and ChimeraX launches one instance of the coordset slider interface for each member of the ensemble! Is this the intended behavior, or am I doing something stupid?
Putting the same ensemble pdb into "old" chimera and running ensemble clustering seems to work fine, so I don't think it is an issue with the ensemble itself
Cheers Oli
On Mar 8, 2026, at 9:20 AM, Oliver Clarke <olibclarke@gmail.com> wrote:
Looking at the daily builds, it seems like md cluster might do the trick, if the set of structures is opened as a trajectory - it is not exactly clear to me how to do that though, for a set of individual PDB files as opposed to an MD trajectory or NMR multi-model ensemble.
Also, the PDBs generated by alphafold are in arbitrary orientations. Is there a way to open, align and cluster a large set (100s) of structures in arbitrary orientations?
Cheers Oli
On Mar 7, 2026, at 3:38 PM, Oliver Clarke <olibclarke@gmail.com> wrote:
Hi,
Are there any tools in ChimeraX suitable for clustering/analyzing conformational ensembles, e.g. as generated by AFSAMPLE2 (https://www.nature.com/articles/s42003-025-07791-9) or similar approaches?
The closest I could find was this:
https://www.rbvi.ucsf.edu/chimerax/data/simstruct-doc-oct2024/simstruct.html
The similarstructures interface looks great, but it seems like this can only analyze sets of structures generated in the tool - is there any way to use it to analyze an arbitrary set of structures?
Basically I would like a way to (for a set of say 200 input structures of the same sequence) quickly align them and then cluster them, to identify and group representative conformational states in the ensemble of predictions.
Cheers Oli _______________________________________________ ChimeraX-users mailing list -- chimerax-users@cgl.ucsf.edu To unsubscribe send an email to chimerax-users-leave@cgl.ucsf.edu Archives: https://mail.cgl.ucsf.edu/mailman/archives/list/chimerax-users@cgl.ucsf.edu/ _______________________________________________ ChimeraX-users mailing list -- chimerax-users@cgl.ucsf.edu To unsubscribe send an email to chimerax-users-leave@cgl.ucsf.edu Archives: https://mail.cgl.ucsf.edu/mailman/archives/list/chimerax-users@cgl.ucsf.edu/ _______________________________________________ ChimeraX-users mailing list -- chimerax-users@cgl.ucsf.edu To unsubscribe send an email to chimerax-users-leave@cgl.ucsf.edu Archives: https://mail.cgl.ucsf.edu/mailman/archives/list/chimerax-users@cgl.ucsf.edu/
participants (4)
-
Alexandra Zahradnikova -
Eric Pettersen -
Oliver Clarke -
Tom Goddard