Skip to content

Releases: JeschkeLab/DeerLab

v1.1.4

03 Sep 17:56
34e5a3a
Compare
Choose a tag to compare

Removed Python 3.8 Support and requires Numpy 2.0

v1.1.3

31 Jul 19:24
178249e
Compare
Choose a tag to compare

What's Changed

Full Changelog: v1.1.2...v1.1.3

v1.1.2

03 Nov 11:21
cf7cefe
Compare
Choose a tag to compare
  • Fixes an issue with sophgrid (:pr:463).

  • Fixes an error in the normalisation of the rice models (:pr:459).

  • Removes the broken three spin example (:pr:427).

  • Fixes an error in the linear solver for linearly constrained not non-negative problems.

v1.1.1

28 Aug 07:15
f7e0340
Compare
Choose a tag to compare
  • Bug fix in FitResult.evaluate(): #454

v1.1.0

18 Aug 18:35
c1926f3
Compare
Choose a tag to compare
  • The definition of the dipolar time axis for RIDME has changed to match the one for 4-pulse DEER (#436).
  • The doc theme has moved to furo (#432).
  • Fixes errors in documentation (#429).
    • Changes the file name of figures 'modelling*.pngtomodeling*.png`. To keep all spelling consistent with American english.
    • Adds a missing ) in fitting_guide
    • Corrects the time axis in echo_crossing example.
    • Corrects plotting in ex_simulate_basic_4pdeer example.
  • Fixes an errors in tests (#429).
    • The test test_algorithms had an incorrect search range.
  • Fixes an errors in the general Gaussian distribution model (#435).
  • Removes the default font from the fit function due to conflicts on some systems (#429).
  • Methods of the FitResult class are now defined with the class (#440).
    • FitResult class now in file fitresults.py
    • fit function now in file fit.py
  • Added a RIDME example (#438).
  • Masked data points are no-longer included in gof statistics in FitResult.plot().

v1.0.1

22 Mar 11:20
5496cff
Compare
Choose a tag to compare

Release v1.0.1 - March 2023

|fix| Fixes some minor bugs in the documentation. The file modelling_guide.rst is renamed to modeling_guide.rst to keep spelling consistency. The "Simulating a two-pathway 5-pulse DEER signal" and "Simulating a three-pathway 4-pulse DEER signal" examples now run correctly.
|fix| Fixes issues with CVXOPT in tests. The testing will now use quadprog as the default solver. To account for the change tested values are now generated using the grid method
|fix| Updates GitHub actions to use latest packages.
|api| Hardcodes out Python 3.11 support. This will remain until quadprog is fixed.
|fix| Removes hard-wired RNG seeding

Release v1.0.0

19 Dec 17:02
da38704
Compare
Choose a tag to compare

Release v1.0.0 - December 2022

  • |feature| Introduces multi-spin dipolar EPR spectroscopy models (:pr:385).

    • Implements multi-spin models in dipolarmodel by taking into account all possible two-spin an three-spin interaction pathways in a multi-pin system. Add a new optional argument spins to specify the number of spins in a molecule. From that number and the npathways or experiment arguments, dipolarmodel automatically constructs the full set of multi-spin pathways (including permutations) within the pair-wise pathway factorization approximation. Adds a new keyword argument triangles, required for spins>3 to specify triads of indices of interspin vectors that are connected forming a triangle (must be specified by the user since it defines the geometry of the spins system).
    • Adds a new argument specperm to dipolarmodel, to enable/disable the assumption of spectral permutability of the multi-spin system. This option allows to account for multi-spin system with chemically or spectrally different spin-1/2 centers.
    • Adds and changes the optional argument nKnots in dipolarkernel and dipolarmodel to gridsize. Changing the specification from knots to grid points allows a finer control of the computation costs due to the highly non-linear scaling of grid points with number of knots in a spherical grid.
    • Implements a new utility function sophegrid to generate SOPHE grids based on a port of the Easyspin sphgrid function.
    • The multi-spin models in dipolarmodel have at the moment a hard-coded unimodal normal multi-variate distance distribution at its core. This could be adapted in the future to be modular to specify multi-modal distributions and other multi-variate basis functions.
  • |feature| Implements different parametrization strategies for dipolar EPR models (:pr:409). Implements a new optional argument parametrization for dipolarmodel, which specifies the parametrization strategy of the dipolar pathway refocusing times. Can be one of the following:

    • 'reftimes' - Each refocusing time is represented individually as a parameter. The current behavior.
    • ``'delays'``` - The pulse delays are introduced as parameters from which the refocusing times are computed
    • 'shift' - A time shift is introduced as a parameter to represent the variability of the refocusing times from their theoretical values.
  • |feature| Expands the toolset for asserting the goodness-of-fit of model estimates robustly (:pr:388).

    • Adds new option gof to FitResults.plot() method to add plots to aid goodness-of-fit assessment. Besides the plot of the data and the model fit, now adds a plot of the residuals with its mean and noise level estimates shown as lines, a plot of the normalized residual histogram compared to the standard normal distribution, and a plot of the autocorrelation in the data along the confidence region expected from white noise.
    • Adds a new quantity, Residuals autocorr. to the results summary. The value of this quantity is computed as abs(2 - dDW), where dDW is the Durbin–Watson statistic. It allows a quick assessment of the autocorrelation in the fit residuals. This quantity is also returned by the fit function as FitResults.stats['autocorr']
    • Remove the AIC quantity specified for each dataset from the fit's summary.
    • Add a new automatic system that colorizes the chi2red and Residuals autocorr. values depending on their severity, indicating a wrong estimate. Yellow coloring alerts the user of potential failures and red of confidently wrong estimates.
  • |efficiency| |fix| Implements a new NNLS solver function qpnnls that uses the Goldfarb/Idnani dual algorithm implemented in the quadprog package (:pr:390). The resulting NNLS solver is more efficient than the current implementation with cvxopt without sacrificing any accuracy. Since NNLS problem-solving is the bottleneck of most applications of DeerLab, it results in a significant improvement in speed when analyzing any models with linear parameters. Since it substitutes cvxopt as the default solver engine, it removes the hard dependency on the cvxopt package and its issues related to the Apple M1 chip (:issue:407).

  • |efficiency| Implements a new, more efficient evaluation of multi-pathway dipolar models based to greatly enhance the performance of dipolar EPR spectroscopy analyses (:pr:393). Combined with the new NNLS solver, the analysis time of, e.g., 4-pulse and 5-pulse DEER multi-pathway models, has been reduced on average about 70-80%.

  • |enhancement| Improves the interface and definition of dipolar pathways in dipolarmodel (:pr:396).

    • Labels the parameters of the different dipolar pathways included in dipolarmodel via the experiment argument based on the pathway label numbers rather than numerically by order of specification.
    • Harmonizes the pathway ordering of the ex_rev5pdeer and ex_fwd_5pdeer models according to published literature.
    • Adds new figures to the ex_ model functions to show the table of dipolar pathways along a schematic illustration of their intramolecular contributions. Improved the docstrings of the ex_ models in general.
  • |enhancement| |fix| Improves and expands the documentation (:pr:397, :pr:408).

    • Adds multiple new examples, fixes errors in existing ones.
    • Fixes several graphical bugs in the website.
    • Thoroughly documents all public and private functions in the code.
    • Adds release instructions for maintainers.
    • Adds a Publications page to the documentation linking to all literature for concepts introduced by DeerLab.
  • |api| Removes the subpackage deerlab.utils and makes its functions part of the main package for simpler maintenance (:pr:408). Removes multiple unused private functions.

.. rubric:: fit

  • |enhancement| The function now returns a full uncertainty quantification for the normalization factor of any model parameter with a normalization condition (:pr:372).
  • |efficiency| |api| Removes the automatic computation of the modelUncert output containing the propagated uncertainty estimate of the model's response (:pr:401). This significantly speeds up the runtime of the function by disabling the automatic propagation of uncertainty to the model's response which could take from several seconds to several minutes in complex models (:issue:391).

.. rubric:: dipolarkernel

  • |feature| Implements multi-spin dipolar pathways up to three-spin interactions (:pr:385). The function takes now a list of distance vectors [r1,r2,...,rQ] for multi-spin kernels.
  • |feature| Expands the function to be able to account for arbitrary experimental time coordinates (:pr:385). Now a list of time vectors [t1,t2,...,tD] can be specified to construct a D-dimensional dipolar kernel.
  • |enhancement| : Refactors most code in the function (:pr:385). THe code should now be more logically ordered using mathematical symbols for clearer equations.
  • |api| Introduces a new and clearer syntax for defining dipolar pathways (:pr:385). Now, instead of specifying a list of pathways, where each pathway is a list of values (being the amplitude, refocusing time, and harmonic in that order), now pathways are specified as a list of dictionaries, e.g. pathways = [{'amp':0.5}, 'reftime':0, 'harmonic':1].
  • |feature| |efficiency| Adds a new optional argument tinterp to construct a dipolar kernel for a pathway and interpolate other pathways from that one (:pr:393).

.. rubric:: dipolarbackground

  • |feature| Implements multi-spin dipolar pathways up to three-spin interactions (:pr:385).
  • |feature| Expands the function to be able to account for arbitrary experimental time coordinates (:pr:385). Now a list of time vectors [t1,t2,...,tD] can be specified to construct a D-dimensional dipolar background function.
  • |api| Introduces the same new syntax for defining dipolar pathways as in dipolarkernel (:pr:385).

.. rubric:: correctphase

  • Adds a new optional argument offset to enable a numerical optimization of the phase while accounting for a non-zero imaginary component offset (:issue:392, :pr:395).

.. rubric:: snlls

  • Adds an optional argument modeluq to enable /disable the model uncertainty propagation (:pr:401).

v0.14.5

09 Dec 15:33
fcdb450
Compare
Choose a tag to compare
v0.14.5 Pre-release
Pre-release

Release v0.14.5 - December 2022

  • |fix| The distribution of DeerLab through Anaconda and its conda manager has been deprecated as of this release (#400).
  • |fix| Fix errors in the background function plots used in the examples showing 4-pulse DEER analyses.

fit

  • |fix| Expose the cores option of bootstrap_analysis to parallelize bootstrap analysis from the fit function.
  • |fix| Correct behavior of masking during fitting (#394). When using the mask option of the fit function, certain steps such as noise estimation and goodness-of-fit assessment were not taking into account the mask during the analysis.

bootstrap_analysis

  • |fix| Fix error prompted when analyzing scalar variables (#402).

v0.14.4

10 Aug 15:29
56856d9
Compare
Choose a tag to compare
v0.14.4 Pre-release
Pre-release

Release v0.14.4 - August 2022

  • |feature| The experiment model functions such as ex_4pdeer now take an additional optional keyword argument pulselength to specify the longest durations of a pulse during an experiment. The value is then used to more accurately set the boundaries of refocusing time parameters in dipolar models.
  • |enhancement| The parameter table displayed when printing DeerLab models has been expanded to show the start values and the frozen values, if present.
  • |fix| Added missing documentation for certain attributes of the UQResult objects related to the bootstrap and profile-likelihood methods.
  • |fix| Behavior of the documention on minimized browser windows and mobile phones.
  • |fix| Multiple issues with the incorrect dark theming of the webpage.

fit

  • |fix| Added multiple missing optional keyword arguments to the documentation of the function.

dd_randcoil

  • |fix| Fixed the erronously switched descriptions of the model parameters.

v0.14.3

30 Jul 22:15
11f4b1d
Compare
Choose a tag to compare
v0.14.3 Pre-release
Pre-release

Release v0.14.3 - July 2022

  • |api| Deprecated support for Python 3.6 and 3.7.

  • |feature| Added multiple quality of life improvements to the modelling system.

    • Add new method paramA.setas(paramB) for Parameter objects to copy the full metadata content from paramB into paramA.
    • Expand the FitResult summary to report on the regularization parameter and penalty weights when used in the analysis.
    • Improve the report of incorrect attribute requests in FitResult objects and provide close matches as suggestions.
    • Improve the report of errors during the evaluation of Model objects.
  • |fix| Fix bug in the fit function unfreezing all frozen model parameters upon fitting if any model parameter included a normalization constraint.

  • |fix| Corrected two minor mathematical errors in the physical background models bg_homfractal and bg_homfractal_phase.

  • |fix| Fixed display of the online documentation in browsers with an enabled dark theme that made certain menus and text sections unreadable. The documentation will now default to a light theme even for dark themed browser.