Multi-omics Autoencoder Integration (maui)

maui is an autoencoder-based framework for multi-omics data analysis. It consists of two main modules, The Maui Class, and Maui Utilities. For an introduction of the use of autoencoders for multi-omics integration, see Multi-modal Autoencoders.

Table of contents

• Multi-modal Autoencoders
  • Variational Autoencoders
  • Stacked Autoencoders
• Data and Normalization
• Filtering and Merging latent factors
  • Dropping unexplanatory latent factors
  • Merging similar latent factors
  • Supervised filtering of latent factors
• Saving and loading models
  • Saving a trained model to disk
  • Loading a model from disk
• The Maui Class
• Maui Utilities

Quickstart

The Maui class implements scikit-learn’s BaseEstimator. In order to infer latent factors in multi-omics data, first instantiate a Maui model with the desired parameters, and then fit it to some data:

from maui import Maui

maui_model = maui.Maui(n_hidden=[900], n_latent=70, epochs=100)
z = maui_model.fit_transform({'mRNA': gex, 'Mutations': mut, 'CNV': cnv})

This will instantiate a maui model with one hidden layer of 900 nodes, and a middle layer of 70 nodes, which will be traiend for 100 epochs. It then feeds the multi-omics data in gex, mut, and cnv to the fitting procedure. The omics data (gex et. al.) are pandas.DataFrame objects of dimension (n_features, n_samples). The return object z is a pandas.DataFrame (n_samples, n_latent), and may be used for further analysis.

In order to check the model’s convergance, the hist object may be inspected, and plotted:

maui_model.hist.plot()

For a more comprehensive example, check out our vignette.

Indices and tables

• Index
• Module Index
• Search Page