A Julia package for the structure determination from single molecule X-ray scattering experiments with down to only three photons. This package has not been officially released in the Julia Repository.
This is not meant for 'out-of-box' usage (yet). It is a proof-of-principle demonstration of the method described in the paper "Structure determination from single molecule X-ray scattering with three photons per image".
First, clone the repository:
git clone https://github.com/h4rm/ThreePhotons.jl.git
Then make sure the environment variables are set correctly:
#Make sure julia finds the module
export JULIA_LOAD_PATH=/path/to/ThreePhotons.jl/src
#Make sure the s2kit is found
export DYLD_LIBRARY_PATH=$DYLD_LIBRARY_PATH:/path/to/ThreePhotons.jl/src/sh
#Tell julia where the structure determination runs
export THREEPHOTONS_PATH=/path/to/ThreePhotons.jl/Next, compile s2kit and the s2kit interface (part of this framework):
cd src/sh
makeStarting from a fresh Julia installation, you may run:
Pkg.init()
Pkg.add("CUDArt")
Pkg.add("Distributions")
Pkd.add("ProgressMeter")
Pkg.add("Optim")
Pkg.add("CurveFit")
Pkg.add("PyPlot")to initialize Julia with all required packages.
The package implements three major aspects of the structure determination method:
- 3D description of electron densities, Fourier densities and intensities
- synthetic image data generation and subsequent histograming of the two- and three-photon correlation
- structure determination runs starting from correlation histograms
This package provides functions to describe 3D structures on a cubic grid or on a spherical grid, each shell expanded in spherical harmonics. PDB structures can be loaded directly.
using ThreePhotons
LMAX = 25 #Maximum expansion order of spherical harmonics expansion
KMAX = 30 #Maximum number of shells used in the expansion
#Description of the Crambin electron density, Fourier density, and Fourier intensity expanded in Spherical Harmonics
density,fourier,intensity = createSphericalHarmonicsStructure("data/structures/crambin.pdb", LMAX, KMAX, float(KMAX))
#Same Crambin structure on a cubic grid
densityCube,fourierCube,intensityCube = createCubicStructure("data/structures/crambin.pdb", 2*KMAX+1, float(KMAX))This package covers the generation of synthetic scattering images (with and without noise) and subsequent two-photon and three-photon histogramming. For computational reasons, the scattering images are not cached.
using ThreePhotons
include("jobs/runs.jl")
generate_histograms(;
num_images = Integer(1e6), #Number of images to generate,
max_triplets = Integer(0), #Alternatively, the maximum number of triplets can be limited
Ncores = 8, #Number of CPU cores used for the data generation
N = 32, #Alpha/beta discretization
photons_per_image = 10, #Photons per image
batchsize = round(Int64,1e6/8),
successive_jobs = 1,
use_cube = false, #Use cubic or spherical harmonics description for data generation
qcut_ratio = 1.0, #Fraction of maximum wave number
K = 38, #total number of shells
rmax = float(38), #maximum radius in real space
histogram_method = "histogramCorrelationsInPicture_alltoall",
structure_pdb_path= "data/structures/crambin.pdb"
)jobs/runs.jl includes helper functions to spawn data generation and structure determination runs in various environments (including cluster systems).
Given a histogrammed two- and three-photon correlation, the structure can be retrieved de novo:
using ThreePhotons
include("jobs/runs.jl")
num_images::Int64 = Integer(1e6) #number of images
KMAX::Int64 = 38 #Maximum shell number used for two-photon inversion
N::Int64 = 32 #Alpha/beta discretization
L::Int64 = 18 #Maximum expansion order
K::Int64 = 26 #Number of shells used for structure determination
ppi::Int64 = 10 #Photons per image used for the histogram
rmax = float(KMAX)#Maximum radius of the reference structures
name = histogram_name("", ppi, N, KMAX, float(KMAX), img, "") for img in image_list) #histogram file name
run_determination(
"runname", #Name of the run
histograms = name, #Path to the histogram file
#Expansion parameters (see above)
K = K,
L = L,
KMAX = KMAX,
rmax = rmax,
N = N,
#Monte Carlo simulated annealing parameters
initial_stepsize = pi/4.0,
optimizer = "rotate_all_at_once",
initial_temperature_factor=0.1,
temperature_decay = 0.99998,
stepsizefactor = 1.01
measure = "Bayes",
#Misc parameters
range = 1000:1019,
postprocess = true,
gpu = true,
Ncores = 20,
)- FFTW 3.3 (for SH calculations in s2kit)
- Gnu Science Library (GSL) (for spherical harmonics basis functions (Ylm) and Wigner-3j symbols)