dousatsu 0.2.1.2

Utility functions for proteomics data analysis

Dousatsu

Dousatsu is a Python library for the analysis of quantitative mass spectrometry-based proteomics data. It provides a set of tools for feature preprocessing, analysis, selection, and visualization, enabling a comprehensive workflow from raw data to biological insights.

The library is designed to be modular and easy to use, with a focus on integrating with the scientific Python ecosystem, including pandas, numpy, scikit-learn, and statsmodels.

Core Modules

Dousatsu is organized into four main modules, each addressing a specific step in the proteomics data analysis pipeline:

feature_preprocessing

This module provides a suite of tools for cleaning, normalizing, and transforming raw proteomics data into an analysis-ready format. Key functionalities include:

• Data Loading: Functions to load data from common proteomics software outputs like TRIC, Diann, and Spectronaut.
• Data Cleaning: Transformers to remove contaminants, non-proteotypic peptides, and low-quality data based on intensity and q-value cutoffs.
• Data Formatting: Tools to reshape data from wide to long format and to standardize column names.
• Normalization: Methods for median and quantile normalization to correct for systematic variations between samples.
• Missing Value Imputation: Strategies to handle missing values, a common issue in proteomics data.

The preprocessing steps are implemented as scikit-learn compatible transformers, allowing them to be chained together in a Pipeline.

feature_analysis

Once the data is preprocessed, this module offers functions for statistical analysis to identify differentially abundant proteins or peptides. Features include:

• Statistical Tests: Implementation of two-sample t-tests with corrections for multiple testing (e.g., Benjamini-Hochberg).
• Fold Change Calculation: Functions to calculate log2 fold changes between different conditions.
• Correlation Analysis: Tools to assess the correlation between technical replicates.

feature_selection

This module helps in identifying the most informative features (peptides or proteins) for building predictive models or for biomarker discovery. It includes:

• Recursive Feature Elimination (RFE): A cross-validated RFE implementation to select the most stable and predictive features.
• Visualization: Functions to visualize the results of the feature selection process.

feature_visualization

A picture is worth a thousand words. This module provides a wide range of visualization functions to explore the data and present the results of the analysis:

• Dimensionality Reduction: PCA plots to visualize sample clustering and identify batch effects.
• Differential Abundance: Volcano plots to visualize the results of statistical tests.
• Heatmaps: Clustermaps to visualize the expression patterns of proteins or peptides across samples.
• Data Quality: Plots for visualizing intensity distributions and missingness.

Development Environment

This repository is set up for development inside a Docker container to ensure a consistent and reproducible environment.

Requirements

• Docker

How to use

Initial setup

1. Clone the repository.
2. Build and start the development container:

   ./start_dev.sh
   

3. The first time you start the container, install the pre-commit hooks:

   pre-commit install
   

Developing

• The project directory is mounted inside the container at /App, so you can edit the files on your host machine with your favorite editor.
• Run all git commands from within the container.
• Install the package in editable mode to test your changes:

  pip install -e .
  

• To stop the container, run:

  ./stop_dev.sh
  

This will also remove the container, so you can start fresh the next time.