Deconvolution of autoencoders to learn biological regulatory modules from single cell mRNA sequencing data

Abstract

Background: Unsupervised machine learning methods (deep learning) have shown their usefulness with noisy single cell mRNA-sequencing data (scRNA-seq), where the models generalize well, despite the zero-inflation of the data. A class of neural networks, namely autoencoders, has been useful for denoising of single cell data, imputation of missing values and dimensionality reduction. Results: Here, we present a striking feature with the potential to greatly increase the usability of autoencoders: With specialized training, the autoencoder is not only able to generalize over the data, but also to tease apart biologically meaningful modules, which we found encoded in the representation layer of the network. Our model can, from scRNA-seq data, delineate biological meaningful modules that govern a dataset, as well as give information as to which modules are active in each single cell. Importantly, most of these modules can be explained by known biological functions, as provided by the Hallmark gene sets. Conclusions: We discover that tailored training of an autoencoder makes it possible to deconvolute biological modules inherent in the data, without any assumptions. By comparisons with gene signatures of canonical pathways we see that the modules are directly interpretable. The scope of this discovery has important implications, as it makes it possible to outline the drivers behind a given effect of a cell. In comparison with other dimensionality reduction methods, or supervised models for classification, our approach has the benefit of both handling well the zero-inflated nature of scRNA-seq, and validating that the model captures relevant information, by establishing a link between input and decoded data. In perspective, our model in combination with clustering methods is able to provide information about which subtype a given single cell belongs to, as well as which biological functions determine that membership.

Improving gene function predictions using independent transcriptional components

“…For example, autoencoders are a neural network architecture currently used to infer a latent representation of the mRNA transcription patterns obtained from both bulk and single-cell samples 27 , 28 . This representation can be used as a regulatory barcode to generate gene function prediction scores within a GBA strategy.…”

Potential applications of deep learning in single-cell RNA sequencing analysis for cell therapy and regenerative medicine

“…Of course, the deep learning method can not only remove the batch effect in the scRNA‐seq data, or complete the clustering of cells, but also directly learn the inherent biological modules in the data, and then describe the biologically meaningful control data set modules and provide information about which modules are active for each unit. For example, the autoencoder after deconvolution processing, 42 it also has good adaptability to a large amount of “lost” data, which is useful and important for analyzing scRNA‐seq data. Kinalis et al found that after specific training, the autoencoder can establish a connection between the model's presentation layer and biological functions after deconvolution of the resulting model.…”

“…Kinalis et al found that after specific training, the autoencoder can establish a connection between the model's presentation layer and biological functions after deconvolution of the resulting model. The hidden cells in the model are then mapped to well‐defined modules to complete the signature decryption of genes or locations, so that the autoencoder can better outline the driving factors behind a given cellular effect 42 . This will analyze the stem cell population from another angle, which will help us to screen the subpopulations of stem cells we need, and it will also play a role in promoting the development of stem cell therapy.…”

scVAE: Variational auto-encoders for single-cell gene expression data

“…As mentioned above, it should also be possible to combine encodings of the cells in the latent space and produce in-between cells like Lotfollahi et al (2018). We would also like to extent our investigation of what dimensions of the latent variables encode (Kinalis et al, 2019). We note that it is possible to apply these models to data sets with multiple modalities such as RNA-seq and exome sequencing (Brouwer and Lió, 2017).…”