cimcb

A package containing the necessary tools for the statistical analysis of untargeted and targeted metabolomics data.

Installation

Dependencies

cimcb requires:

• Python (>=3.5)
• Bokeh (>=1.0.0)
• Keras
• NumPy (>=1.12)
• Pandas
• SciPy
• scikit-learn
• Statsmodels
• TensorFlow
• tqdm

User installation

The recommend way to install cimcb and dependencies is to using conda:

conda install -c cimcb cimcb

or pip:

pip install cimcb

Alternatively, to install directly from github:

pip install https://github.com/cimcb/cimcb/archive/master.zip

API

For futher detail on the usage refer to the docstring.

cimcb.model

• PLS_SIMPLS: Partial least-squares regression using the SIMPLS algorithm.
• PLS_NIPALS: Partial least-squares regression using the NIPALS algorithm.
• PCR: Principal component regression.
• PCLR: Principal component logistic regression.
• RF: Random forest.
• SVM: Support vector machine.
• NN_LinearLinear: 2 Layer linear-linear neural network.
• NN_LinearSigmoid: 2 Layer linear-sigmoid neural network.
• NN_SigmoidSigmoid: 2 Layer sigmoid-sigmoid neural network.

cimcb.plot

• boxplot: Creates a boxplot using Bokeh.
• distribution: Creates a distribution plot using Bokeh.
• pca: Creates a PCA scores and loadings plot using Bokeh.
• permutation_test: Creates permutation test plots using Bokeh.
• roc_plot: Creates a rocplot using Bokeh.
• scatter: Creates a scatterplot using Bokeh.
• scatterCI: Creates a scatterCI plot using Bokeh.

cimcb.cross_val

• kfold: Exhaustitive search over param_dict calculating binary metrics using k-fold cross validation.
• holdout: Exhaustitive search over param_dict calculating binary metrics using hold-out set.

cimcb.bootstrap

• Perc: Returns bootstrap confidence intervals using the percentile boostrap interval.
• BC: Returns bootstrap confidence intervals using the bias-corrected boostrap interval.
• BCA: Returns bootstrap confidence intervals using the bias-corrected and accelerated boostrap interval.

cimcb.utils

• binary_metrics: Return a dict of binary stats with the following metrics: R2, auc, accuracy, precision, sensitivity, specificity, and F1 score.
• ci95_ellipse: Construct a 95% confidence ellipse using PCA.
• dict_mean: Calculate mean for all keys in dictionary.
• dict_median: Calculate median for all keys in dictionary.
• dict_perc: Calculate confidence intervals (percentile) for all keys in dictionary.
• dict_std: Calculate std for all keys in dictionary.
• knnimpute: kNN missing value imputation using Euclidean distance.
• load_dataCSV: Loads and validates the DataFile and PeakFile from CSV files.
• load_dataXL: Loads and validates the DataFile and PeakFile from a excel file.
• nested_getattr: getattr for nested attributes.
• scale: Scales x (which can include nans) with method: 'auto', 'pareto', 'vast', or 'level'.
• table_check: Error checking for DataTable and PeakTable (used in load_dataXL).
• univariate_2class: Creates a table of univariate statistics (2 class).
• wmean: Returns Weighted Mean. Ignores NaNs and handles infinite weights.

License

cimcb is licensed under the MIT license.

Authors

• Kevin Mendez
• David Broadhurst

Correspondence

Professor David Broadhurst, Director of the Centre for Integrative Metabolomics & Computation Biology at Edith Cowan University. E-mail: d.broadhurst@ecu.edu.au