Loading...

Batch-Effect Correction of Single-Cell Feature Embedding Using Deep Learning

Bahrami, Mojtaba | 2020

1153 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 53607 (19)
  4. University: Sharif University of Technology
  5. Department: Computer Engineering
  6. Advisor(s): Rabiee, Hamid Reza
  7. Abstract:
  8. Single-cell RNA-sequencing has opened opportunities to unlock the cell transcriptome at cellular-level resolution and investigate cell population structures and cell-type-specific gene expression patterns. Previously, we were only able to perform bulk RNA sequencing regardless of the cell type composition of cell populations. Now, with the advancements in sequencing technologies, it is possible to extract cell-level and cell-type-specific sequences, related to each type of cell separately and in a scalable and high-throughput manner. For such large studies, logistical constraints inevitably dictate that data be generated separately i.e., at different times and with different operators. Therefore, data aggregation and interpretation are not a straightforward task due to the separate and parallel experiments and the difference in environmental conditions. The unwanted effects in subgroups of experiments that differ due to unequal environmental conditions is called batch effects. These batch effects lead to confounding non-biological signals on the final representation of the cells. Some previous works use linear processing to eliminate batch effects which are unable to detect and eliminate complex, nonlinear effects. On the other hand, despite the utilization of deep learning-based methods deal with batch effects, the way these networks are employed is still not optimal and needs to be improved. In this study we present single-cell Generative Adversarial Network (scGAN). Our main contribution is to introduce an adversarial network to predict batch effects using the embeddings from the variational autoencoder network, which does not only need to maximize the data likelihood of the raw scRNA-seq counts but also minimize the correlation between the latent embeddings and the batch effects. We demonstrate scGAN on three public scRNA-seq datasets and show that our method confers superior performance over the state-of-the-art methods in forming clusters of known cell types and identifying known psychiatric genes that are associated with major depressive disorder
  9. Keywords:
  10. Batch Effect Correction ; Deep Learning ; Representation Learning ; RNA Sequencing ; Adversarial Training ; Single Cell Sequencing

 Digital Object List

 Bookmark

...see more