Re: [ccp4bb] Maps on mobile phones.
Oo that looks handy! And it reminds me to ask the community: if you wanted to do outreach to high schoolers and get them to, say, look at a DNA structure, is there a PyMol-like program that works on tablets? I see PyMol for ipad has been discontinued. Thanks, Phoebe From: CCP4 bulletin board on behalf of Paul Emsley Reply-To: Paul Emsley Date: Sunday, March 27, 2022 at 11:11 AM To: "CCP4BB@JISCMAIL.AC.UK" Subject: Re: [ccp4bb] Maps on mobile phones. On 27/03/2022 00:17, Jon Cooper wrote: Hello, I have been trying to put together a thing for viewing small blocks of CCP4 electron density maps with a mobile web browser. If anyone is interested, the current state of it is here: http://ic50.org/jbctest14.html Have you seen/heard about uglymol? The name is an insult but the project is interesting. https://github.com/uglymol/uglymol Sorry, the link is not https yet, but nothing gets uploaded to the server! It seems to work OK on Android and iPhone and the maps (note: only maps; it doesn't do MTZ's, sorry) look similar when viewed in Coot (taken as the gold standard ;-), For the record, I've never much liked the contouring of Coot - too many close lines and tiny triangles. I've wanted to change it for a long time, but it's never been the most important thing to fix. but I have a few questions about the contouring algorithm that I have used. It is "surfacenets.js" from here: https://github.com/mikolalysenko/isosurface and a paper describing it is here: https://www.merl.com/publications/docs/TR99-24.pdf Unfortunately, my maths is not good enough to tell if it matters if you give it fractional coordinates, rather than orthogonal. I simply give it the electron density values on the CCP4 map grid coordinates, which will be on non-orthogonal axes for unit cells with non-90 degree angles. It seems to give qualitatively similar results to Coot in these cases, so I am cautiously optimistic, but not sure. uglymol makes the transformation that you need - so does CootVR for that matter https://github.com/hamishtodd1/hamishtodd1.github.io/tree/master/cvr Here's Chris Hassall playing with it: https://www.youtube.com/watch?v=-wfopgdN8o4 Just to be clear, this is running inside Firefox. (It looks like he discovered a contouring bug when he zooms out) Another thing is that the results of the contouring are sent out in groups of 3 points which are the vertices of triangles forming the surface. Hence, I orthogonalize them and get three.js to draw them as just that - triangles. My worry is that, since the triangles all have edges in common, nearly all of the contour lines (except the ones at the edges of the map box) get drawn twice, or at least are sent to three.js twice for drawing, which doesn't seem terribly efficient?! Coot used to remove double drawing. The large speed up in contouring in the 0.9 series is a result of removing that test and just drawing the lines twice. Is there a nicer way of doing this? I think it might be better to have FRODO-style contouring just on the 2D sections of the map, rather than having lots of diagonal lines? I think so too, especially as a larger Shannon sampling factor is now not much of an issue. Anyway, my 5 YO phone takes about 3 seconds to step from one residue to the next, so it seems not too bad, although not ideal! FWIW, Kevin Cowtan recently advertised a position to "develop next generation web based molecular graphics software" - sounds related. Finally, its been asked before, but is there a nice way in CCP4i2 to output maps that cover the coordinates of the structure, rather than the asymmetric unit? Saving maps in Coot gives the asymmetric unit, too, although using Export Map Fragment seems the best option. I know about doing this in the old gui with mapmask, or using phenix, so just wondering if I've missed a way of doing this in i2, etc? I know that suitably extended CCP4 maps are available from the PDBe EBI. I just use a script that runs mapmask - using a GUI seems like an overhead. Paul. To unsubscribe from the CCP4BB list, click the following link: https://www.jiscmail.ac.uk/cgi-bin/WA-JISC.exe?SUBED1=CCP4BB&A=1 To unsubscribe from the CCP4BB list, click the following link: https://www.jiscmail.ac.uk/cgi-bin/WA-JISC.exe?SUBED1=CCP4BB&A=1 This message was issued to members of www.jiscmail.ac.uk/CCP4BB, a mailing list hosted by www.jiscmail.ac.uk, terms & conditions are available at https://www.jiscmail.ac.uk/policyandsecurity/
Re: [ccp4bb] Maps on mobile phones.
On 27/03/2022 00:17, Jon Cooper wrote: Hello, I have been trying to put together a thing for viewing small blocks of CCP4 electron density maps with a mobile web browser. If anyone is interested, the current state of it is here: http://ic50.org/jbctest14.html Have you seen/heard about uglymol? The name is an insult but the project is interesting. https://github.com/uglymol/uglymol Sorry, the link is not https yet, but nothing gets uploaded to the server! It seems to work OK on Android and iPhone and the maps (note: only maps; it doesn't do MTZ's, sorry) look similar when viewed in Coot (taken as the gold standard ;-), For the record, I've never much liked the contouring of Coot - too many close lines and tiny triangles. I've wanted to change it for a long time, but it's never been the most important thing to fix. but I have a few questions about the contouring algorithm that I have used. It is "surfacenets.js" from here: https://github.com/mikolalysenko/isosurface and a paper describing it is here: https://www.merl.com/publications/docs/TR99-24.pdf Unfortunately, my maths is not good enough to tell if it matters if you give it fractional coordinates, rather than orthogonal. I simply give it the electron density values on the CCP4 map grid coordinates, which will be on non-orthogonal axes for unit cells with non-90 degree angles. It seems to give qualitatively similar results to Coot in these cases, so I am cautiously optimistic, but not sure. uglymol makes the transformation that you need - so does CootVR for that matter https://github.com/hamishtodd1/hamishtodd1.github.io/tree/master/cvr Here's Chris Hassall playing with it: https://www.youtube.com/watch?v=-wfopgdN8o4 Just to be clear, this is running inside Firefox. (It looks like he discovered a contouring bug when he zooms out) Another thing is that the results of the contouring are sent out in groups of 3 points which are the vertices of triangles forming the surface. Hence, I orthogonalize them and get three.js to draw them as just that - triangles. My worry is that, since the triangles all have edges in common, nearly all of the contour lines (except the ones at the edges of the map box) get drawn twice, or at least are sent to three.js twicefor drawing, which doesn't seem terribly efficient?! Coot used to remove double drawing. The large speed up in contouring in the 0.9 series is a result of removing that test and just drawing the lines twice. Is there a nicer way of doing this? I think it might be better to have FRODO-style contouring just on the 2D sections of the map, rather than having lots of diagonal lines? I think so too, especially as a larger Shannon sampling factor is now not much of an issue. Anyway, my 5 YO phone takes about 3 seconds to step from one residue to the next, so it seems not too bad, although not ideal! FWIW, Kevin Cowtan recently advertised a position to "develop next generation web based molecular graphics software" - sounds related. Finally, its been asked before, but is there a nice way in CCP4i2 to output maps that cover the coordinates of the structure, rather than the asymmetric unit? Saving maps in Coot gives the asymmetric unit, too, although using Export Map Fragment seems the best option. I know about doing this in the old gui with mapmask, or using phenix, so just wondering if I've missed a way of doing this in i2, etc? I know that suitably extended CCP4 maps are available from the PDBe EBI. I just use a script that runs mapmask - using a GUI seems like an overhead. Paul. To unsubscribe from the CCP4BB list, click the following link: https://www.jiscmail.ac.uk/cgi-bin/WA-JISC.exe?SUBED1=CCP4BB&A=1 This message was issued to members of www.jiscmail.ac.uk/CCP4BB, a mailing list hosted by www.jiscmail.ac.uk, terms & conditions are available at https://www.jiscmail.ac.uk/policyandsecurity/
Re: [ccp4bb] Maps on mobile phones.
Hi Jon, > Another thing is that the results of the contouring are sent out in groups of > 3 points which are the vertices of triangles forming the surface. Hence, I > orthogonalize them and get three.js to draw them as just that - triangles. My > worry is that, since the triangles all have edges in common, nearly all of > the contour lines (except the ones at the edges of the map box) get drawn > twice, or at least are sent to three.js twice for drawing, which doesn't seem > terribly efficient?! Is there a nicer way of doing this? I was also concerned about it when writing UglyMol. It was ~6 years ago. Drawing maps as lines was definitely affecting speed, because fps was lower when the number of lines was huge. So to reduce duplicated lines I calculated new tables for the marching cubes algorithms - with lines instead of triangles in the tables. This allowed me to reduce some (most of?) duplicates. I don't remember the details, but the script that I used for calculating tables is here: https://github.com/uglymol/uglymol/blob/master/tools/isolut.py Then I was disappointed, because it didn't change the fps. I don't know why. And I don't know how to check where the bottleneck in the graphics pipeline is (vertex shader, fragment shader or something else?). I guess optimizing map rendering would require someone with expertise in 3D graphics rendering. > Finally, its been asked before, but is there a nice way in CCP4i2 to output > maps that cover the coordinates of the structure, rather than the asymmetric > unit? The maps that cover asu are, on average, smaller. They contain the list of symmetry operators that expand the map to the whole unit cell. So such maps may actually be better. Or, even better, use MTZ files. I think UglyMol is still the only web viewer that can show maps from MTZ, but it works on mobile web browsers (it's not designed for mobile, but rendering works) and you can re-use parts of it. https://uglymol.github.io/view/ Best wishes, Marcin To unsubscribe from the CCP4BB list, click the following link: https://www.jiscmail.ac.uk/cgi-bin/WA-JISC.exe?SUBED1=CCP4BB&A=1 This message was issued to members of www.jiscmail.ac.uk/CCP4BB, a mailing list hosted by www.jiscmail.ac.uk, terms & conditions are available at https://www.jiscmail.ac.uk/policyandsecurity/
