The code in this repository can be used to run our Multi-Omics Variational autoEncoder (MOVE) framework for integration of omics and clinical variabels spanning both categorial and continuous data. Our approach includes training ensemble VAE models and using in silico perturbation experiments to identify cross omics associations. The manuscript has been published in Nature Biotechnology:
Allesøe, R.L., Lundgaard, A.T., Hernández Medina, R. et al. Discovery of drug–omics associations in type 2 diabetes with generative deep-learning models. Nat Biotechnol (2023). https://doi.org/10.1038/s41587-022-01520-x
We developed the method based on a Type 2 Diabetes cohort from the IMI DIRECT project containing 789 newly diagnosed T2D patients. The cohort and data creation is described in Koivula et al. and Wesolowska-Andersen et al.. For the analysis we included the following data:
Multi-omics data sets:
Genomics
Transcriptomics
Proteomics
Metabolomics
Metagenomics
Other data sets:
Clinical data (blood measurements, imaging data, ...)
Questionnaire data (diet etc)
Accelerometer data
Medication data
MOVE is written in Python and can be installed using pip:
>>> pip install move-dlMOVE should run on any environmnet where Python is available. The variational autoencoder architecture is implemented in PyTorch.
The training of the VAEs can be done using CPUs only or GPU acceleration. If you do not have powerful GPUs available, it is possible to run using only CPUs. For instance, the tutorial data set consisting of simulated drug, metabolomics and proteomics data for 500 individuals runs fine on a standard macbook.
Note: The pip installation of
move-dldoes not setup your local GPU automatically
MOVE has five-six steps:
01. Encode the data into a format that can be read by MOVE
02. Finding the right architecture of the network focusing on reconstruction accuracy
03. Finding the right architecture of the network focusing on stability of the model
04. Use model, determined from steps 02-03, to create and analyze the latent space
05. Identify associations between a categorical and continuous datasets
05a. Using an ensemble of VAEs with the t-test approach
05b. Using an ensemble of VAEs with the Bayesian decision theory approach
06. If both 5a and 5b were run select the overlap between them
Please refer to our documentation for examples and tutorials on how to run MOVE.
Additionally, you can copy this notebook and follow its instructions to get familiar with our pipeline.
The data used in notebooks are not available for testing due to the informed consent given by study participants, the various national ethical approvals for the study, and the European General Data Protection Regulation (GDPR). Therefore, individual-level clinical and omics data cannot be transferred from the centralized IMI-DIRECT repository. Requests for access to summary statistics IMI-DIRECT data, including those presented here, can be made to DIRECTdataaccess@Dundee.ac.uk. Requesters will be informed on how summary-level data can be accessed via the DIRECT secure analysis platform following submission of appropriate application. The IMI-DIRECT data access policy is available here.
We have therefore provided two datasets to test the workflow: a simulated dataset and a publicly-available maize rhizosphere microbiome data set.
To cite MOVE, use the following information:
Allesøe, R.L., Lundgaard, A.T., Hernández Medina, R. et al. Discovery of drug–omics associations in type 2 diabetes with generative deep-learning models. Nat Biotechnol (2023). https://doi.org/10.1038/s41587-022-01520-x