Skip to content

v1.0.3

Compare
Choose a tag to compare
@kopeckylukas kopeckylukas released this 04 Mar 14:42
· 27 commits to main since this release
v1.0.3
569269d

Overeview

This version of the package was used for data analysis reported in a manuscript entitled 'MAMSI: Integration of multi-assay liquid chromatography – mass spectrometry metabolomics data using multi-view machine learning', that has been submitted to ACS Analytical Chemistry journal on 4th March 2025 (date of this release).

There were previous releases of the package published on PyPI: https://pypi.org/project/mamsi/.

Version History

Version history of releaases

v1.0.3

New Features

  • k-fold cross-validation implemented as a method .kfold_cv() that can be used for model performance evaluation. This method includes GroupKFold option.
  • Monte Carlo cross-validaton (MCCV), also nown as 'random sampling cross-validation' implemented as a method .montecarlo_cv() that can be used for model performance evaluation.
  • .estimate_lv() method now allows to choose between k-fold CV and MC-CV using parameter method

Bug Fixes and Behavioural Changes

  • Plot title for .block_importance() fixed.
  • For regression analysis, MSE metric changed to RMSE
  • For .estimate_lv() method, parameter y_continuous=False was replaced with classification=True

v1.0.2

New Features

  • New method 'MamsiPls.block_importance()': Calculate the block importance for each block in the multiblock PLS model and plot the results.

Minor Bug Fixes and Behaviour Changes

  • Behavioural changes for MamsiPls.mb_vip(): The MB-VIP plot is now printed by default, scores are not returned by default. New default arguments (plot=True, get_scores=False).
  • Argument changes for MamsiPls.estimate_lv(): Old Arguments (no_folds, n_components) changed to (n_slplits, max_components) respectively.
  • Plots: 'Verdana' is no longer the default font. The default font changed to Matplotlib default 'DejaVu Sans'.
  • Updates to MamsiStructSearch class to comply with future warnings - Pandas 3.0.

v1.0.1

Minor Bugs Update

  • Fixes instances where flattened correlation clusters were misaligned to structural clusters.
  • Readme licence badge links directly to GitHub licence file (URL).
Assets 3
Loading