Iterative Projection Algorithms for protein crystallography
A program for ab initio phase determination.
-
The program is capable of ab initio phase determination for protein crystals with high solvent content (solvent content > 70%)
-
It can also be used to generate plausible molecular envelopes for crystals with much lower solvent content.
Ab initio phase determination is treated as a constraint satisfaction problem, in which an image is sought that is consistent with both the diffraction data and generic constraints on the density distribution in the crystal. The problem is solved using iterative projection algorithms which have good global convergence properties, and can locate the correct solution without any initial phase information.
The only required inputs are the measured diffraction data (in a CCP4 mtz file) and an estimate for the solvent fraction of the crystal.
The current release (v1.1.0) contains the source code which is readily compilable under Mac OS 10.15, 11, 12 (Intel chip architecture). See the detailed instructions in the src directory (Install_MacOS.txt). Data and parameter file to allow solution of a test case (PDB ID 2w4m) are included in sub-directory test_case.
We hope to distribute Mac OS binary executables shortly and we are working on a Linux-compatible version.
The behavior of program IPA is controlled by a single experimental parameter file. Running the program without a parameter file will cause the program to generate a default parameter file, which the user can edit.
The only compulsory edits are:
(1) Adding details of the mtz file that contains the diffraction data.
The program expects estimates of the Structure Factor Amplitudes |F| and their standard deviations. Standard Bayesean treatment of weak data is anticipated. If the data have been corrected for significant anisotropy (e.g. as is optionally done via the STARANISO server), this will cause the estimation of the overall anisotropic B-factor within program IPA to perform unpredictably, with some bad downstream consequences ...
(2) Providing an estimate of the solvent fraction of the crystal.
The solvent fraction is generally estimated via analysis of crystal packing density (e.g. with the CCP4 program MATTHEWS_COEFF). If there is uncertainty about the solvent fraction, it is better to underestimate this quantity. If the solvent fraction is overestimated, ab initio phase determination will certainly fail (setting the solvent fraction too high is inconsistent with the solution, while setting it too low remains consistent with the solution).
For computational efficiency, the program breaks the problem of ab initio phase determination into two stages; initial approximation of the molecular envelope at low resolution, followed by subsequent phase determination using all of the data. At both stages, the algorithm is initiated with many different and random phase sets, which are evolved subject the constraints. A clustering procedure is used to identify consistent results across multiple runs, which are then averaged to generate consensus envelopes or phase sets. The emergence of highly consistent phase sets is diagnostic of success.
The results are stored in 4 subdirectories:
- envelope_determination
- envelope_consensus
- phase_determination
- phase_consensus
If successful, the program will output the proposed solution (both Fourier coefficients and maps) in directory phase_consensus, for inspection and interpretation. As with all ab initio phase determination methods, the procedure may generate the true solution or its inverse. The correct hand of the reconstructed image must be determined by inspection.
Version 1 of the program effectively automates the procedures described in:
Kingston, R.L. and Millane R.P. (2022) A general method for directly phasing diffraction data from high–solvent content protein crystals. IUCrJ, 9, https://doi.org/10.1107/S2052252522006996
Address correspondence to: r.l.kingston@auckland.ac.nz
Report issues to michael.barnett@auckland.ac.nz or r.l.kingston@auckland.ac.nz