This package uses data-driven models to analyze SAXS data. The models are trained from a set SAXS spectra that have been analyzed and curated on the Citrination platform.
SAXS spectra employed in this package are attributed to the following sources:
- Wu, Liheng, et al. Nature 548, 197–201 (2017). doi: 10.1038/nature23308
As more spectra are added to the curated set, the models behind this package are expected to become more effective. If you have a SAXS data set that you would like to volunteer to add to the curated set, please contact the development team at ssrl-citrination@slac.stanford.edu or paws-developers@slac.stanford.edu.
This example profiles, parameterizes, and optimizes the fit of a scattering equation to a measured saxs spectrum.
Read intensivity n-by-2 array `q_I` from a csv file:
import numpy as np
q_i = np.genfromtxt ('my_data/sample_0.csv', delimiter=",")
Import saxskit:
import saxskit
Profile a saxs spectrum:
from saxskit.saxs_math import profile_spectrum features = profile_spectrum(q_i)
To predict scatters populations we can use SAXSKIT models (built on Sklearn) or Citrination models.
Using SAXSKIT models:
Initialize SaxsClassifier and predicted scatterer populations:
from saxskit.saxs_classify import SaxsClassifier m = SaxsClassifier() populations, propability = m.classify(features) for k,v in populations.items(): print(k, ":", v, " with propability: %1.3f" % (propability[k])) print()
Initialize SaxsRegressor and predict counting scatterer parameters:
from saxskit.saxs_regression import SaxsRegressor r = SaxsRegressor() params = r.predict_params(populations,features, q_i)
Initialize SaxsFitter and update scatterer parameters with intensity parametes:
from saxskit import saxs_fit sxf = saxs_fit.SaxsFitter(q_i,populations) params, report = sxf.fit_intensity_params(params) for k,v in params.items(): print(k, ":", end="") for n in v: print(" %10.3f" % (n))
Using Citrination models:
Create SaxsCitrination using Citrination credentials:
from saxskit.saxs_citrination import CitrinationSaxsModels api_key_file = 'path_to_my_api_key_txt' saxs_models = CitrinationSaxsModels(api_key_file,'https://slac.citrination.com')
Predict scatterer populations:
populations, uncertainties = saxs_models.classify(features) for k,v in populations.items(): print(k, ":", v, " with uncertainties: %1.3f" % (uncertainties[k])) print()
Predict counting scatterer parameters:
params,uncertainties = saxs_models.predict_params(populations, features, q_i) for k,v in params.items(): print(k, ":", v, " +/- %1.3f" % (uncertainties[k])) print()
Initialize SaxsFitter and update scatterer parameters with intensity parametes:
sxf = saxs_fit.SaxsFitter(q_i,populations) params, report = sxf.fit_intensity_params(params) for k,v in params.items(): print(k, ":", end="") for n in range(len(v)): print(" %10.3f" % (v[n]) ) print()
The full version of this code: https://github.com/scattering-central/saxskit/blob/dev/examples/predict.py
Output: https://github.com/scattering-central/saxskit/blob/dev/examples/output.png
There are some more detailed examples of predictions, training and updating of models in "examples" folder.
This package is hosted on PyPI. Install it by pip install pysaxs
To contribute code, please feel free to submit a pull request on this repository.
To contribute data for improving the models, please contact the development team at ssrl-citrination@slac.stanford.edu or paws-developers@slac.stanford.edu.
The 3-clause BSD license attached to this software can be found in the LICENSE file in the source code root directory.