[phenixbb] Introducing ISOLDE

Tristan Croll tic20 at cam.ac.uk
Tue May 1 01:05:17 PDT 2018

Dear members of the structural biology community,

It is with great pleasure that I announce the release of version 1.0b1 
of ISOLDE, a new, free for academic/nonprofit use, environment for 
interactive rebuilding of macromolecular models against low-medium 
resolution experimental crystallographic or cryo-EM maps.

A paper describing ISOLDE is online at 
http://scripts.iucr.org/cgi-bin/paper?S2059798318002425, and you can 
also get an overview from my demonstration video at 

In brief, though, ISOLDE is designed to help overcome the substantial 
challenge involved in building high-quality models into “difficult” maps 
(e.g. those with worse than 3Å resolution) by re imagining the task as 
one akin to interacting with a real, “living” molecule. At its heart is 
the use of interactive molecular dynamics: all atoms feel van der Waals 
and electrostatic interactions with their neighbours as you interact 
with them, avoiding clashes and constantly settling towards their 
nearest low-energy state. In addition to this, ISOLDE helps interpret 
what you're looking at by performing real-time rotamer, Ramachandran and 
peptide plane validation, flagging up potential problems directly on the 
model to give you instant feedback on the results of your actions.

Specific features:
- support for protein, nucleic acid, water and some small molecules/ions 
(with substantially expanded support to come in later versions)
- start an interactive simulation covering anything from a few residues 
to your entire model (depending on your hardware) at the click of a 
- “always-on” validation updates every time coordinates change (results 
are shown directly as markup on the model itself, and also accessible 
via more traditional tables/plots)
- Interactively apply peptide plane or cis/trans flips and add/remove 
position, distance, rotamer and secondary structure restraints via a 
simple graphical interface
- Smoothly shift any continuous stretch of backbone forward/backward 
along the chain to easily recover from register shift errors
- Save a checkpoint in a running simulation and return to it with a 
button click, allowing you to experiment with tricky regions without 
- Real-time visualisation of crystal symmetry contacts: when the master 
atoms move, they move.
- One-button masking of maps to cover any arbitrary atom selection – 
particularly useful for diagnosing large-scale problems.
- Extensive GUI and API documentation, available by clicking the “Help” 
button in the ISOLDE GUI window.

ISOLDE is built as a plugin to UCSF ChimeraX 
(http://preview.cgl.ucsf.edu/chimerax/download.html) and is currently 
available for modern Linux distros and MacOS, and can be installed from 
within ChimeraX via its ToolShed (Tools/More Tools...). It should run on 
any Linux version that supports ChimeraX itself. NOTE: due to some 
recent changes in the ChimeraX API, the current build will only work 
with the ChimeraX 0.6 release (not the version 0.7 daily builds).

In terms of hardware, ISOLDE is designed based on the understanding that 
not everybody has access to high-performance computing facilities, so 
its minimum requirements are as modest as possible. The most important 
requirement is a GPU capable of OpenCL 1.2 or CUDA processing. My 
“go-to” machine for performance testing is a 2016-model gaming laptop 
equipped with a Nvidia GTX 1070 GPU – this achieves ~20 coordinate 
updates and >30 graphics updates per second with live validation on the 
included 229-residue demo model, indistinguishable from (or slightly 
faster than) my Xeon workstation. At the other end of the scale, it runs 
at usable speeds on my MacBook Air for simulations of a few dozen 
residues, sufficient for many tasks.

I hope you find it useful, and enjoy using it as much as I've enjoyed 
making it. The MolProbity score to beat on the demo model is 1.02 (after 
refinement in your favourite package – ID is in the PDB header so you 
can download reflections from the wwPDB), and bonus points if you can 
identify the mystery ligand (I can't). If you recognise it as your model 
I apologise – I picked it out at random, I swear!

Best regards,


More information about the phenixbb mailing list