Benchmarking variational AutoEncoders on cancer transcriptomics data

PLoS ONE, vol. 18, pp. e0292126

Abstract

Deep generative models, such as variational autoencoders (VAE), have gained increasing attention in computational biology due to their ability to capture complex data manifolds which subsequently can be used to achieve better performance in downstream tasks, such as cancer type prediction or subtyping of cancer. However, these models are difficult to train due to the large number of hyperparameters that need to be tuned. To get a better understanding of the importance of the different hyperparameters, we examined six different VAE models when trained on TCGA transcriptomics data and evaluated on the downstream tasks of cluster agreement with cancer subtypes and survival analysis. We studied the effect of the latent space dimensionality, learning rate, optimizer, initialization and activation function on the quality of subsequent downstream tasks on the TCGA samples. We found β-TCVAE and DIP-VAE to have a good performance, on average, despite being more sensitive to hyperparameters selection. Based on these experiments, we derived recommendations for selecting the different hyperparameters settings. To ensure generalization, we tested all hyperparameter configurations on the GTEx dataset. We found a significant correlation (ρ = 0.7) between the hyperparameter effects on clustering performance in the TCGA and GTEx datasets. This highlights the robustness and generalizability of our recommendations. In addition, we examined whether the learned latent spaces capture biologically relevant information. Hereto, we measured the correlation and mutual information of the different representations with various data characteristics such as gender, age, days to metastasis, immune infiltration, and mutation signatures. We found that for all models the latent factors, in general, do not uniquely correlate with one of the data characteristics nor capture separable information in the latent factors even for models specifically designed for disentanglement.

Authors 6

1. Mostafa Eltager

   Delft University of Technology

   Affiliation as printed

   Delft Bioinformatics Lab, Delft University of Technology, Delft, The Netherlands

2. Tamim Abdelaal Aachen Department of Radiology, Leiden University Medical Center

   Leiden University Medical Center · Delft University of Technology

   Affiliation as printed

   Delft Bioinformatics Lab, Delft University of Technology, Delft, The Netherlands

   Department of Radiology, Leiden University Medical Center, Leiden, The Netherlands

3. Mohammed Charrout

   Delft University of Technology

   Affiliation as printed

   Delft Bioinformatics Lab, Delft University of Technology, Delft, The Netherlands

4. Ahmed Essam Mahfouz Aachen Department of Human Genetics Leiden Computational Biology Center

   Leiden University Medical Center · Delft University of Technology

   Affiliation as printed

   Delft Bioinformatics Lab, Delft University of Technology, Delft, The Netherlands

   Department of Human Genetics, Leiden University Medical Center, Leiden, The Netherlands

   Leiden Computational Biology Center, Leiden University Medical Center, Leiden, The Netherlands

5. Marcel J. T. Reinders Aachen Department of Human Genetics Leiden Computational Biology Center

   Leiden University Medical Center · Delft University of Technology

   Affiliation as printed

   Delft Bioinformatics Lab, Delft University of Technology, Delft, The Netherlands

   Department of Human Genetics, Leiden University Medical Center, Leiden, The Netherlands

   Leiden Computational Biology Center, Leiden University Medical Center, Leiden, The Netherlands

6. Stavros Makrodimitris corresponding

   Erasmus MC · Delft University of Technology

   Affiliation as printed

   Delft Bioinformatics Lab, Delft University of Technology, Delft, The Netherlands

   Department of Medical Oncology, Erasmus Medical Center, Rotterdam, The Netherlands