PhenoFeatureFinder

Linking developmental phenotypes to metabolic features

PhenoFeatureFinder is divided into three classes:

• PhenotypeAnalysis
• OmicsAnalysis
• FeatureSelection

PhenotypeAnalysis was designed to analyse development over time through progressive stages in multiple groups or treatments. This could for example be the development of insects through their larval stages over time in different environments, or disease scores of fungal infections in multiple host plants. These types of phenotyping analyses can be challenging, due to the many variables involved (e.g. time, developmental stages, replicates, treatments), especially for researchers whose strength does not lie in data analysis. PhenotypeAnalysis offers a set of functions to visualise the development while taking into account those different variables, and to perform the necessary data preprocessing steps. From the output, it is easy to manually assign binary phenotypes to your groups (treatments, genotypes, etc), if you want to use it as input for FeatureSelection.

With OmicsAnalysis, you can filter large untargeted metabolomics datasets and visualise the structure of the data. The filtered data and a corresponding set of binary phenotypes can then be used as input for FeatureSelection. With only a few lines of code, the best fitting pipeline to link the phenotypes to metabolic features is created using Automated Machine Learning with TPOT and scikit-learn.

Although OmicsAnalysis and FeatureSelection are designed for metabolomics data, they might also be used for other types of omics data. The user would have to keep in mind that the functions were written for the specifics of metabolomics data (high sparsity, strongly correlated features) and first assess the fit for other types of data.

Installation

$ pip install PhenoFeatureFinder

At this moment, PhenoFeatureFinder requires python 3.9.

Usage

For each of the classes, you can find a manual with an explanation for all of their functions in the manuals folder. Alternatively, you can find the documentation of the classes and their functions on Read the Docs.

If you want to see an example of how PhenoFeatureFinder can be used for real-world data, you can take a look at one of the two examples. The first example showcases the use of the PhenotypeAnalysis class for the analysis of the development of caddisfly larvae in four freshwater streams. In the second example, the OmicsAnalysis and FeatureSelection classes are used to analyse and select interesting features from a mass spectrometry dataset of a panel of bacterial species.

Dependencies

Required for all classes:

• NumPy
• pandas
• Matplotlib
• seaborn

Additionally required for PhenotypeAnalysis:

• SciPy

Additionally required for OmicsAnalysis:

• scikit-learn
• UpSetPlot

Additionally required for FeatureSelection:

• scikit-learn
• TPOT
• auto-sklearn (auto-sklearn is made for Linux operating systems. On macOS it needs to be installed manually with brew and pip. You can do this by following these instructions.)

Testing

Before using PhenoFeatureFinder to analyse your data, follow the manuals using the accompanying data to test the functionality. The obtained results should be identical to those in the manual. If you run into any errors, please contact the authors.

Citation

Insert citation option when ready

Contributing

Interested in contributing? Check out the contributing guidelines. Please note that this project is released with a Code of Conduct. By contributing to this project, you agree to abide by its terms.

Author contributions

PhenoFeatureFinder was created by Lissy-Anne Denkers and Marc Galland, with input from Annabel Dekker, Valerio Bianchi and Petra Bleeker.

License

This package is licensed under the terms of the Apache License 2.0 license.

Credits

PhenoFeatureFinder was created with cookiecutter and the py-pkgs-cookiecutter template.

Useful reading

• Auto-sklearn talks
• NumPy docstring examples