Automated model building (cryo-EM)

Why

Determining the structure of a macromolecule typically requires building an atomic model that fits the cryo-EM map. If the structure of the molecule or its components are unknown, the model has to be built from scratch into the cryo-EM map. This task is challenging because molecules are typically very large and chain tracing is difficult at low resolution. Therefore, manual interpretation of cryo-EM maps is time-consuming and error-prone. For many cases, however, automated methods can be used to build models for both proteins and nucleic acids.

After an atomic model is generated, it is real-space refined using secondary-structure restraints (e.g., see phenix.real_space_refine).

How

The program phenix.map_to_model interprets a cryo-EM map and automatically builds an atomic model. First, the map is sharpened using phenix.auto_sharpen. The unique parts of the structure are then identified by taking the reconstruction symmetry into account. In addition, the procedure identifies which parts of the map correspond to protein or RNA/DNA. To save runtime, you can use phenix.auto_sharpen to sharpen the map and phenix.map_box to cut out the unique part before you run phenix.map_to_model.

The phenix.map_to_model tool requires you to input a cryo-EM map with CCP4-style (MRC, etc) or MTZ map coefficients and a sequence file. For optimal model-building results, the resolution of the map should be 4.5 Å or better. The program outputs a PDB file with the resulting PDB map that is superimposed on the original map.

How to use the phenix.autobuild GUI: Click here

Phenix reference manual for phenix.map_to_model

Common issues

• One or more sub-processes crash: If you have a very large structure, your computer may not have enough memory to run phenix.map_to_model. The easiest solution is to run the tool on just the unique part of the structure. You can also cut out boxes and work on each section individually. Or your can cut back on the resolution to save memory.
• Want to re-run without all the steps: Partially-completed jobs can occur when your queueing system crashes during a run or one or more sub-processes crash. If you have a partially-completed or completed run, you can carry on from where you left off by using the keyword carry_on=True. You can also use this keyword to build up your model in pieces or to run many jobs in parallel.

Related programs

• phenix.auto_sharpen: This program performs map sharpening by optimizing the detail and connectivity of a cryo-EM map.
• phenix.map_box: This program cuts out a box around selected atoms from a larger map (MTZ, CCP4, or X-plor format).
• phenix.fix_insertions_deletions: This program fixes register errors in the model by comparing the density of side-chain positions to a sequence file.
• phenix.trace_and_build: This tool rapidly builds protein models by tracing the path of a polypeptide chain, finding the CB positions, and building the atomic model. Normally, this is done by running phenix.map_to_model, but you can also run phenix.trace_and_build directly.