-
Notifications
You must be signed in to change notification settings - Fork 8
Deep learning approaches for single cell data
We present scScope, a scalable deep-learning based approach that can accurately and rapidly identify cell-type composition from millions of noisy single-cell gene-expression profiles.
Link: https://www.biorxiv.org/content/early/2018/11/27/315556
Code: https://github.com/AltschulerWu-Lab/scScope
We propose a deep count autoencoder network (DCA) to denoise scRNA-seq datasets. DCA takes the count distribution, overdispersion and sparsity of the data into account using a zero-inflated negative binomial noise model, and nonlinear gene-gene or gene-dispersion interactions are captured. Our method scales linearly with the number of cells and can therefore be applied to datasets of millions of cells. We demonstrate that DCA denoising improves a diverse set of typical scRNA-seq data analyses using simulated and real datasets. DCA outperforms existing methods for data imputation in quality and speed, enhancing biological discovery.
Link: https://www.biorxiv.org/content/early/2018/04/13/300681
Code: https://github.com/theislab/dca
Tags: tensorflow, python, autoencoder
Single-cell transcriptome measurements can reveal unexplored biological diversity, but they suffer from technical noise and bias that must be modeled to account for the resulting uncertainty in downstream analyses. Here we introduce single-cell variational inference (scVI), a ready-to-use scalable framework for the probabilistic representation and analysis of gene expression in single cells. scVI uses stochastic optimization and deep neural networks to aggregate information across similar cells and genes and to approximate the distributions that underlie observed expression values, while accounting for batch effects and limited sensitivity. We used scVI for a range of fundamental analysis tasks including batch correction, visualization, clustering, and differential expression, and achieved high accuracy for each task.
Link: https://people.eecs.berkeley.edu/~jregier/publications/lopez2018deep.pdf
Code: https://github.com/YosefLab/scVI
Generative modeling and latent space arithmetics predict single-cell perturbation response across cell types, studies and species
We demonstrate that scGen learns cell type and species specific response implying that it captures features that distinguish responding from non-responding genes and cells. With the upcoming availability of large-scale atlases of organs in healthy state, we envision scGen to become a tool for experimental design through in silico screening of perturbation response in the context of disease and drug treatment.
Link: https://www.biorxiv.org/content/early/2018/12/14/478503
Code: https://github.com/theislab/scGen
SAUCIE, a deep neural network that leverages the high degree of parallelization and scalability offered by neural networks, as well as the deep representation of data that can be learned by them to perform many single-cell data analysis tasks, all on a unified representation
Link: https://www.biorxiv.org/content/early/2019/01/03/237065
Code: https://github.com/KrishnaswamyLab/SAUCIE/
Tags: tensorflow, python, autoencoder
We treated zeros as missing values and developed deep learning methods to impute the missing values, exploiting the dependence structure across genes and across cells. Specifically, our LATE (Learning with AuToEncoder) method trains an autoencoder directly on scRNA-seq data with random initial values of the parameters, whereas our TRANSLATE (TRANSfer learning with LATE) method further allows for the use of a reference gene expression data set to provide LATE with an initial set of parameter estimates.
Link: https://www.biorxiv.org/content/early/2018/12/29/504977
Code: https://github.com/audreyqyfu/LATE
Tags: tensorflow, python, autoencoder
VASC: dimension reduction and visualization of single cell RNA sequencing data by deep variational autoencoder
We proposed VASC (deep Variational Autoencoder for scRNA-seq data), a deep multi-layer generative model, for the unsupervised dimension reduction and visualization of scRNA-seq data. It can explicitly model the dropout events and find the nonlinear hierarchical feature representations of the original data.
Link: https://www.biorxiv.org/content/early/2017/10/06/199315
Code: https://github.com/wang-research/VASC
Here we are proposing Dhaka a variational autoencoder based single cell analysis tool to transform genomic data to a latent encoded feature space that is more efficient in differentiating between the hidden tumor subpopulations.
Link: https://www.biorxiv.org/content/early/2018/04/19/183863
Code: https://github.com/MicrosoftGenomics/Dhaka
We propose a novel variational auto-encoder-based method for analysis of single-cell RNA sequencing data (scVAE) data. It avoids data preprocessing by using raw count data as input and can robustly estimate the expected gene expression levels and a latent representation for each cell. We show for several scRNA-seq data sets that our method outperforms recently proposed scRNA-seq methods in clustering cells.
Link: https://www.biorxiv.org/content/early/2018/05/16/318295
Code: https://github.com/chgroenbech/scVAE
Extracting a biologically relevant latent space from cancer transcriptomes with variational autoencoders
Variational autoencoders (VAEs) are a deep neural network approach capable of generating meaningful latent spaces for image and text data. In this work, we sought to determine the extent to which a VAE can be trained to model cancer gene expression, and whether or not such a VAE would capture biologically-relevant features. We name our method “Tybalt”.
Link: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5728678/
Code: https://github.com/greenelab/tybalt
Visualizing and interpreting single-cell gene expression datasets with Similarity Weighted Nonnegative Embedding
We developed Similarity Weighted Nonnegative Embedding (SWNE), which enhances interpretation of datasets by embedding the genes and factors that separate cell states alongside the cells on the visualization, captures local structure better than t-SNE and existing methods, and maintains fidelity when visualizing global structure. SWNE uses nonnegative matrix factorization to decompose the gene expression matrix into biologically relevant factors, embeds the cells, genes and factors in a 2D visualization, and uses a similarity matrix to smooth the embeddings.
Link: https://www.biorxiv.org/content/early/2018/06/22/276261
Code: https://github.com/yanwu2014/swne
We propose a novel similarity-learning framework, SIMLR (single-cell interpretation via multi-kernel learning), which learns an appropriate distance metric from the data for dimension reduction, clustering and visualization applications.
Link: https://www.biorxiv.org/content/early/2017/02/28/052225
Code: https://github.com/BatzoglouLabSU/SIMLR
cellassign automatically assigns single-cell RNA-seq data to known cell types across thousands of cells accounting for patient and batch specific effects. Information about a priori known markers cell types is provided as input to the model in the form of a (binary) marker gene by cell-type matrix. cellassign then probabilistically assigns each cell to a cell type, removing subjective biases from typical unsupervised clustering workflows.
Code: https://github.com/irrationone/cellassign
Here we present a new generalizable method (scPred) for prediction of cell type(s), using a combination of unbiased feature selection from a reduced-dimension space, and machine-learning classification. scPred solves several problems associated with the identification of individual gene feature selection, and is able to capture subtle effects of many genes, increasing the overall variance explained by the model, and correspondingly improving the prediction accuracy.
Link: https://www.biorxiv.org/content/early/2018/12/03/369538
Code: https://github.com/IMB-Computational-Genomics-Lab/scPred
Integration of multiple single-cell transcriptomics datasets leveraging stable expression and pseudo-replication
To enable effective interrogation of multiple scRNA-Seq datasets, we have developed a novel algorithm, named scMerge, that removes unwanted variation by combining stably expressed genes and utilizing pseudo-replicates across datasets. Analysis of large collections of publicly available datasets demonstrates that scMerge performs well in multiple scenarios and enhances biological discovery, including inferring cell developmental trajectories.
Link: https://www.biorxiv.org/content/early/2018/08/16/393280
Code: https://github.com/SydneyBioX/scMerge
Dimensionality reduction of single-cell RNA-seq high-dimensional datasets is essential for visualization and analysis, but single-cell RNA-seq data is challenging for classical dimensionality reduction methods because of the prevalence of dropout events leading to zero-inflated data. Here we develop a dimensionality reduction method, (Z)ero (I)nflated (F)actor (A)nalysis (ZIFA), which explicitly models the dropout characteristics, and show that it improves performance on simulated and biological datasets.
Link: https://www.biorxiv.org/content/early/2015/06/14/019141
Code: https://github.com/epierson9/ZIFA
We introduce scMatch, which directly annotates single cells by identifying their closest match in large reference datasets. We used this strategy to annotate various single-cell datasets and evaluated the impacts of sequencing depth, similarity metric and reference datasets.
Code: ttps://github.com/forrest-lab/scmatch
We present scQuery, a web server which uses our neural networks and fast matching methods to determine cell types, key genes, and more.
Link: https://www.biorxiv.org/content/early/2018/05/31/323238
Code: https://github.com/AmirAlavi/scrna_nn
https://github.com/mruffalo/sc-rna-seq-pipeline
https://github.com/AmirAlavi/single_cell_deg
We develop and test a method based on neural networks (NN) for the analysis and retrieval of single cell RNA-Seq data. We tested various NN architectures, some of which incorporate prior biological knowledge, and used these to obtain a reduced dimension representation of the single cell expression data. We show that the NN method improves upon prior methods in both, the ability to correctly group cells in experiments not used in the training and the ability to correctly infer cell type or state by querying a database of tens of thousands of single cell profiles. Such database queries (which can be performed using our web server) will enable researchers to better characterize cells when analyzing heterogeneous scRNA-Seq samples.
Link: https://academic.oup.com/nar/article/45/17/e156/4056711
Here, we present scmap, a method for projecting cells from a scRNA-seq experiment onto the cell-types or individual cells identified in other experiments
Link: https://www.biorxiv.org/content/early/2017/11/29/150292
Code: https://github.com/hemberg-lab/scmap
We present a method based on low-rank approximation which successfully replaces these dropouts (zero expression levels of unobserved expressed genes) by nonzero values, while preserving biologically non-expressed genes (true biological zeros) at zero expression levels.
Link: https://www.biorxiv.org/content/early/2018/08/22/397588
Code: https://github.com/KlugerLab/ALRA
We therefore suggest a different approach for cell labeling, namely, Castle - classifying cells from scRNA-seq datasets by using a model transferred from different (previously labeled) datasets
Link: https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0205499
Code: https://github.com/yuvallb/CaSTLe
We show that a deep autoencoder coupled to a Bayesian model remarkably improves UMI-based scRNA-seq data quality by transfer learning across datasets. This new technology, SAVER-X, outperforms existing state-of-the-art tools. The deep learning model in SAVER-X extracts transferable gene expression features across data from different labs, generated by varying technologies, and obtained from divergent species.
Link: https://www.biorxiv.org/content/early/2018/11/01/457879
Code: http://singlecell.wharton.upenn.edu/saver-x/
We present rCASC a modular RNAseq analysis workflow allowing data analysis from counts generation to cell sub-population signatures identification, granting both functional and computation reproducibility
Link: https://www.biorxiv.org/content/early/2018/10/03/430967
Code: https://github.com/kendomaniac/rCASC
Fast Batch Alignment of Single Cell Transcriptomes Unifies Multiple Mouse Cell Atlases into an Integrated Landscape
BBKNN is a fast and intuitive batch effect removal tool that can be directly used in the scanpy workflow. It serves as an alternative to scanpy.api.pp.neighbors(), with both functions creating a neighbour graph for subsequent use in clustering, pseudotime and UMAP visualisation.
Link: https://www.biorxiv.org/content/early/2018/08/22/397042
Code: https://github.com/Teichlab/bbknn
Unlike available single-cell integration methods, Harmony can simultaneously account for multiple experimental and biological factors. We develop objective metrics to evaluate the quality of data integration. In four separate analyses, we demonstrate the superior performance of Harmony to four single-cell-specific integration algorithms.
Link: https://www.biorxiv.org/content/early/2018/11/05/461954
Code: https://github.com/immunogenomics/harmony
We present a new GAN called the Manifold-Aligning GAN (MAGAN) that aligns two manifolds such that related points in each measurement space are aligned together. We demonstrate applications of MAGAN in single-cell biology in integrating two different measurement types together. In our demonstrated examples, cells from the same tissue are measured with both genomic (single-cell RNA-sequencing) and proteomic (mass cytometry) technologies.
Link: https://arxiv.org/abs/1803.00385
Code: https://github.com/KrishnaswamyLab/MAGAN
Here we describe a protocol for successful exploratory data analysis using t-SNE. They include PCA initialisation, multi-scale similarity kernels, exaggeration, and downsampling-based initialisation for very large data sets. We use published single-cell RNA-seq data sets to demonstrate that this protocol yields superior results compared to the naive application of t-SNE.
Link: https://www.biorxiv.org/content/early/2018/10/25/453449
Code: https://github.com/berenslab/rna-seq-tsne
We introduce net-SNE, which trains a neural network to learn a high quality visualization of single cells that newly generalizes to unseen data. While matching the visualization quality of t-SNE on 14 benchmark data sets of varying sizes, from hundreds to 1.3 million cells, net-SNE also effectively positions previously unseen cells, even when an entire subtype is missing from the initial data set or when the new cells are from a different sequencing experiment.
Link: https://www.biorxiv.org/content/early/2018/03/27/289223
Code: https://github.com/hhcho/netsne
Uniform Manifold Approximation and Projection (UMAP) is a recently-published non-linear dimensionality reduction technique. Another such algorithm, t-SNE, has been the default method for such task in the past years. Herein we comment on the usefulness of UMAP high-dimensional cytometry and single-cell RNA sequencing, notably highlighting faster runtime and consistency, meaningful organization of cell clusters and preservation of continuums in UMAP compared to t-SNE.
Link: https://www.biorxiv.org/content/early/2018/04/10/298430
Code: https://github.com/lmcinnes/umap/
Code: https://github.com/alabarga/Parametric-t-SNE
scIVA is an interactive web-tool for visualisation and analysis of single cell RNA-Seq data.
Code: https://github.com/IMB-Computational-Genomics-Lab/scIVA
Tags: R, shiny
scClustViz provides an interactive interface for visualisation, assessment, and biological interpretation of cell-type classifications in scRNAseq experiments that can be easily added to existing analysis pipelines, enabling customization by bioinformaticians while enabling biologists to explore their results without the need for computational expertise.
Link: https://f1000research.com/articles/7-1522/v1
Code: https://github.com/BaderLab/scClustViz
Tags: R, shiny
iS-CellR is a web-based Shiny application designed to provide a comprehensive analysis of single-cell RNA sequencing data. iS-CellR provides a fast method for filtering and normalization of raw data, dimensionality reductions (linear and non-linear) to identify cell types clusters, differential gene expression analysis to locate markers, and inter-/intra-sample heterogeneity analysis.
Link: https://academic.oup.com/bioinformatics/article/34/24/4305/5048937
Code: https://github.com/immcore/iS-CellR
Tags: R, shiny
Shiny app for visualization, exploration of mouse brain single cell gene expression
Link: http://sci-hub.tw/https://doi.org/10.1016/j.cell.2018.07.028
Code: https://github.com/broadinstitute/dropviz
Tags: R, shiny
We have recently shown how Locality Sensitive Hashing (LSH) improves speed and accuracy of cell type clustering (14). CellAtlasSearch implements LSH on the powerful GPU architecture to attain an unmatched speed in archiving and querying expression data. Hashing based low dimensional encoding of expression profiles makes data transactions efficient and inexpensive, thus future-proof.
Link: https://academic.oup.com/nar/article/46/W1/W141/5000022
Code: http://www.cellatlassearch.com/
too-many-cells is a suite of tools, algorithms, and visualizations focusing on the relationships between cell clades. This includes new ways of clustering, plotting, choosing differential expression comparisons, and more