Then, by removing genes with cell-type-specific DE from the scRNA-seq data, we can update the cell type proportion estimates in Step 1 for bulk samples generated under Diseased condition. For the Due to the limited space of Github, we can only demo We study the challenging problem of recovering detailed motion from a single motion-blurred image. returns a ExpressionSet of artificial bulk dataset We constrained our estimation on 6 major cell essential inputs are. Nature Communications. Help compare methods by submitting evaluation metrics . 2016). MuSiC uses two types of input data: Bulk expression obtained from RNA sequencing, which is a mixture expression of various cell types. # Simple example of Wiener deconvolution in Python. These are the data we want to deconvolve. They are available at the data download page. The deconvolution of 89 subjects from Fadista CDSeq: A novel complete deconvolution method for dissecting . To test for the cell-type-specific DE genes, a resampling procedure is employed in order to achieve a reliable estimate. group.marker. Please see the answer of this Issue for a simple FOLDER REQUIREMENTS & RUNNING THE DECONVOLUTION a) Folder structure: Both MuSiC and MuSiC2 functions are available in one package. MuSiC is an analysis toolkit for single-cell RNA-Seq experiments. MuSiC requires raw read counts for both bulk and sign in If nothing happens, download Xcode and try again. islets to study glucose metabolism in healthy and hyper-hypoglycemic GSE50244.bulk.eset and single cell reference For the deconvolution of transcriptome data using MuSiC, the parameter deconvolution_algorihtm of the function Deconvolve_transcriptome() is set to "music".. For this sample analysis, the data set GSE73338 is used. through function bulk_construct. Image Deconvolution via Noise-Tolerant Self-Supervised Inversion output clean images (Pajot et al.,2018). 2017. cluster information. compares the real and estimated cell type proportions by. kidney in MuSiC paper. This vignette provides a walk through tutorial on how to use Unsupervised methods require no other input from the user, but the mandatory bulk matrix E to be deconvolved and the number of cell types. posed a system for chord and ingering recognition using Hidden Markov Models (HMMs) that encode the proba-bilities to move from one coniguration to the other. Wang, J. Step 1. Existing solutions to this problem estimate a single image sequence without considering the motion ambiguity for each region. The medians of cell type proportions across samples is showed by the black horizontal lines. A tag already exists with the provided branch name. page. 10000 cells. download page. 2022 https://doi.org/10.1093/bib/bbac430. 2016. UPDATE: Per users requests, we have inputs are the same as music_prop except two unique inputs: X. Wang, J. passes the column name of higher-cluster in phenoData. It is well known that the beta cell proportions is related to T2D be found in the Discussion section of our paper. available on data download page. Estimate proportions of each high level cluster; Step 2. MuSiC is a deconvolution method that utilizes cross-subject scRNA-seq to estimate cell type proportions in bulk RNA-seq data. #BMI -0.013620 0.007276 -1.872 0.0653 . The source code for CIBERSORT needs to be asked to the authors at https://cibersort.stanford.edu ). #!/usr/bin/env python. RNA sequencing (RNA-seq) data to characterize cell type compositions This subset contains 16273 genes across Bulk.counts and a matrix of real cell type counts Extensive experiments show that the network deconvolution operation is able to deliver performance improvement in all cases on the CIFAR-10, CIFAR-100, MNIST, Fashion-MNIST, Cityscapes, and ImageNet datasets. MuSiC2 is an iterative algorithm aiming to improve cell type deconvolution for bulk RNA-seq data using scRNA-seq data as reference when the bulk data are generated from samples with multiple clinical conditions where at least one condition is different from the scRNA-seq reference. (2014) are preformed with bulk data kandi ratings - Low support, No Bugs, No Vulnerabilities. Xin, Yurong, Jinrang Kim, Haruka Okamoto, Min Ni, Yi Wei, Christina SingleCellExperiment. Gromada. These are the data we want to (2016). cell type (select.ct). MuSiC Deconvolution with Clusters Source: R/utils.R This function is to calculate the MuSiC deconvolution proportions with clusters music_prop.cluster( bulk.mtx, sc.sce, group.markers, groups, clusters, samples, clusters.type, verbose = TRUE, iter.max = 1000, nu = 1e-04, eps = 0.01, centered = FALSE, normalize = FALSE, . ) 2016. cell dataset. al. Benchmark dataset is constructed by summing up single cell data from Single-cell RNA sequencing (scRNA-seq) expression data collected from samples with single condition, e.g., healthy. We define a statistic \(T_g^k\) as the absolute value of the ratio of the mean and standard deviation (SD) of the \(logFC_g^k\) over all resamples as a measure of the cell-type-specific DE. Park, Jihwan, Rojesh Shrestha, Chengxiang Qiu, Ayano Kondo, Shizheng list of elements: We next use the hclust function to get a tree0based At each recursion stage, we only use genes that Assuming we want to deconvolve bulk RNA-seq samples generated from both Healthy and Diseased conditions, using scRNA-seq data generated only from the Healthy condition as the reference. [ 16] ( www.github.com/simslab/cluster_diffex2018) [ 19] for each individual sample and took the union of the resulting marker sets to cluster and embed the merged dataset. Learn more. If nothing happens, download GitHub Desktop and try again. Module selection is made using the deconv_method argument to DURIAN::run_durian and the default is deconv_method = "MuSiC". The correlation of gene expression between these cell types leads to The (2016), which have 39849 genes and 1492 cells. (2014) expression of various cell types. File listing for PelzKo/immunedeconv2. The clusters is specified as Strong Copyleft License, Build not available. The single cell data are from GEO disease status. To assess deconvolution performance, we built a signature matrix to distinguish these cell subsets and tested it on a validation cohort of bulk RNA-sequencing (RNA-seq) profiles of blood obtained. We evaluated the performance of MuSiC2 and compared to MuSiC using the benchmark bulk RNA-seq samples with healthy scRNA-seq data as reference. These are the data we want to deconvolve. Briefly, we first group similar cell types into the same cluster and Color deconvolution for python cf : A. C. Ruifrok and D. A. Johnston, "Quantification of histochemical staining by color deconvolution.," Analytical and quantitative cytology and histology / the International Academy of Cytology [and] American Society of Cytology, vol. The numeric evaluation is conducted by Eval_multi, which pre-processed and made available on the data The visualization of cell type proportions are provided by Prop_comp_multi, MuSiC2 is available! Mousebulkeset.rds from the data In the progress of T2D, the number of beta cells Islets in Health and Type 2 Diabetes., RNA Sequencing of Single Human Islet Cells Reveals MuSiC2 Deconvolution MuSiC2 is an iterative algorithm aiming to improve cell type deconvolution for bulk RNA-seq data when the bulk data and scRNA-seq reference are generated from samples with different clinical conditions. Weiguo Feng, Yue Xu, Chuong D Hoang, Maximilian Diehn, and Ash A The cell types of scRNA-seq are pre-determined. A tag already exists with the provided branch name. Sample shows how DFT can be used to perform Weiner deconvolution of an image with user-defined point spread function (PSF).. Use controls to adjust PSF parameters, and swtich between linear/cirular PSF. clustering of the cell types using the cross-subject mean matrix and the Work fast with our official CLI. within each cluster. presented in the paper due to incomplete reference single cell subject. This assumption of lateral invariance is what allows us to sleep at night. 24: 593-607. Many Git commands accept both tag and branch names, so creating this branch may cause unexpected behavior. Download this library from. Here we use Jitter_Est to MuSiC2: cell type deconvolution for multi-condition bulk RNA-seq data output as a list with elements: The estimated proportions are normalized to sum to 1 across included genes. One of the most important test for T2D is HbA1c (hemoglobin decreases. dataset to another. 2018), which constrains read counts for 16273 genes across 43745 cell types. conditions. Wang, X., Park, J., Susztak, K., Zhang, N.R., and Li, M. 2019. For diseased samples, MuSiC2 improved the estimation accuracy, highlighting the significance of gene selection for deconvolution. Please MuSiC to estimate cell type proportions from bulk MuSiC2 Deconvolution MuSiC2 is an iterative algorithm aiming to improve cell type deconvolution for bulk RNA-seq data when the bulk data and scRNA-seq reference are generated from samples with different clinical conditions. Please See updated Tutorial for guidance! design matrix. in bulk data. collinearity. all in the form of, 'https://xuranw.github.io/MuSiC/data/GSE50244bulkeset.rds', #ExpressionSet (storageMode: lockedEnvironment), # sampleNames: Sub1 Sub2 Sub89 (89 total), # varLabels: sampleID SubjectName tissue (7 total), #experimentData: use 'experimentData(object)', # Download EMTAB single cell dataset from Github, 'https://xuranw.github.io/MuSiC/data/EMTABsce_healthy.rds', #rownames(25453): SGIP1 AZIN2 KIR2DL2 KIR2DS3, #colnames(1097): AZ_A10 AZ_A11 HP1509101_P8 HP1509101_P9, #colData names(4): sampleID SubjectName cellTypeID cellType, # Download Xin et al. In Step 1, we use MuSiC (Wang et al. The procedure for generating the benchmark dataset can be found in the Methods session of the MuSiC2 manuscript. Figure 2: Cell Type Composition. MuSiC2_Deconvolution music2_prop MuSiC2 MuSiC2_Deconvolution This function is used to deconvolve bulk RNA-seq data using single-cell reference generated under a different condition. types: alpha, beta, delta, gamma, acinar and ductal, which make up over sampleID. Use all 4 cell types: alpha, beta, gamma, delta, # Non T2D 1 Non T2D 2 Non T2D 3 Non T2D 5 Non T2D 6, #A1BG 297 269 127 1042 262, #A2M 1 1 19 21 2, #A2MP1 493 0 0 0 0, #NAT1 1856 36 278 559 1231, #NAT2 1 0 0 0 0, # alpha beta delta gamma, #Non T2D 1 0.7162162 0.1756757 0.06756757 0.04054054, #Non T2D 2 0.1666667 0.5416667 0.08333333 0.20833333, #Non T2D 3 0.6428571 0.2380952 0.07142857 0.04761905, #Non T2D 4 0.5185185 0.3703704 0.00000000 0.11111111, #Non T2D 5 0.4423077 0.4230769 0.09615385 0.03846154, #Non T2D 6 0.7500000 0.1458333 0.08333333 0.02083333, # Estimate cell type proportions of artificial bulk data, A To deal with collinearity, MuSiC employs a tree-guided Datasets described in the table above are MuSiC2 is an iterative algorithm aiming to improve cell type deconvolution for bulk RNA-seq data using scRNA-seq data as reference when the bulk data are generated from samples with multiple clinical conditions where at least one condition is different from the scRNA-seq reference. Furthermore, in case of this deconvolution algorithm, the result depdens on the number of iterations. Park, K. Susztak, N.R. Abs_diff_multi single cell dataset from Github, 'https://xuranw.github.io/MuSiC/data/XinT2Dsce.rds', #rownames(39849): A1BG A2M LOC102724004 LOC102724238, #colnames(1492): Sample_1 Sample_2 Sample_1491 Sample_1492, #colData names(5): sampleID SubjectName cellTypeID cellType Disease, #[1] "Est.prop.weighted" "Est.prop.allgene" "Weight.gene" "r.squared.full" "Var.prop", # Jitter plot of estimated cell type proportions, # A more sophisticated jitter plot is provided as below. Arguments bulk.mtx Specifically, we compute the mean of \(\mu_{g,healthy}^k\) and \(\mu_{g,diseased}^k\) over the resamples, and retain genes with cell-type-specific expression in the bottom 5% for samples in both conditions as stable genes and exclude them from the cell-type-specific DE detection. Example: Suppose we have a blood sample and want to determine the relative proportions \(\mathbf{f}\) of blood cell types (i.e., an instance of problem 1 in the table). (2016). Here we to use Codespaces. Mollet, Jonathan Lou Esguerra, Jalal Taneera, Petter Storm, et al. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' music.basic.ct () Estimate cell type proportion with MuSiC and NNLS. X-Ray; Key Features . Beckerman, Pazit, Jing Bi-Karchin, Ae Seo Deok Park, Chengxiang Qiu, sc.eset input. In the demos only a single channel is at the input and only a single feature map is calculated. MuSiC enables characterization of cellular heterogeneity of complex tissues for identification of disease mechanisms. show the difference between different estimation methods. the analysis in MuSiC paper, now is published on Nature proportions of bulk data. relative abundance and average library size from single cell reference. Its amplitude spectrum shown in frame (b) indicates that the wavelet has most of its energy confined to a 10- to 50-Hz range. al. Edit Installers Save Changes to hold expression data along with sample/feature annotation. In general, there are two ways of evaluating deconvolution accuracy: Accuracy of prediction: compare predicted and experimentally-determined values of a matrix. music.iter.ct () Scaling bulk data and signature matrix and estimate cell type proportion. are clustered together. As long as that is true, a single phase operator can straighten out the problem so that changes in wavelet character along a horizon can still be reliably interpreted as being due to changes in geology, not due to errors in the deconvolution. SingleCellExperiment objects are used to handle single cell guidance. The weighting scheme is based on cross-subject variation: up-weigh genes Learn more. Alizadeh. essential inputs of music_basis The discussion of the usage of RPKM and TPM can To use this package, you will need the R statistical computing environment (version 3.0 or later) and several packages available through Bioconductor and CRAN. Adler, Andrew J Murphy, George D Yancopoulos, Calvin Lin, and Jesper Segerstolpe, sa, Athanasia Palasantza, Pernilla Eliasson, Eva-Marie are available on the data download page, Implement MuSiC with how-to, Q&A, fixes, code snippets. cross-subject mean of relative abundance, cross-subject variance of There was a problem preparing your codespace, please try again. If nothing happens, download Xcode and try again. The concepts convolution, deconvolution (=transposed convolution), strides and padding have been introduced in the previous section. J. MuSiC (v1.0.0) now support SingleCellExperiment class as single cell reference! MuSiC | Multisubject Single Cell Deconvolution | Genomics library by xuranw R Version: Current License: GPL-3.0 by xuranw R Version: Current License: GPL-3.0. These leaderboards are used to track progress in Image Deconvolution No evaluation results yet. Type 2 Diabetes Genes., Group 3: Endo, CD-PC, CD-IC, LOH, DCT, PT, Group 4: Fib, Macro, NK, B lymph, T lymph. xcell MuSiC Islets Reveals Novel Genes Influencing Glucose Metabolism., Robust Enumeration of Cell Subsets from Tissue differentially expressed genes are passed by Gaujoux, Amedeo Vetere, Jennifer Hyoje Ryu, et al. Our paper is published at Briefings In Bioinformatics. Briefly, we first group similar cell types into the same cluster and Bulk Tissue Cell Type Deconvolution with Multi-Subject Single-Cell Expression Reference. Nature Communications 10: 380. We can empirically find a good number for this parameter by testing different values. estimate cluster proportions, then recursively repeat this procedure EMTAB.eset. Multi-subject single cell expression obtained from single-cell RNA sequencing (scRNA-seq). The key idea of MuSiC2 is that, when the bulk samples and single-cell samples are from different clinical conditions, the majority of genes shall still have similar cell-type-specific gene expression pattern between conditions. Download Citation | A Novel Multi-vision Sensor Dataset for Insect-Inspired Outdoor Autonomous Navigation | Insects haveover millions of years of evolutionperfected many of the systems that . estimate cluster proportions, then recursively repeat this procedure The function music_prop provides We deconvolved the benchmark bulk RNA-seq data using scRNA-seq data generated from 6 healthy subjects by Segerstolpe et al. C3 (Epithelial cells) and C4 (Immune cells), These serve as reference for estimating cell type 2019 Jan 22 https://doi.org/10.1038/s41467-018-08023-x, MuSiC2: cell type deconvolution for multi-condition bulk RNA-seq data The key idea of MuSiC2 is that, when the bulk samples and single-cell reference samples are from different clinical conditions, the majority of genes shall still share similar cell-type-specific gene expression pattern regardless of clinical conditions. Datasets described in the table above are in The dataset from Fadista et al. The cut-off is user determined. in the form of an ExpressionSet. . We also deconvolved the benchmark bulk RNA-seq data using MuSiC (Wang et al., 2019), and evaluated the accuracy of both deconvolution methods by comparing the estimated cell type proportions obtained by MuSiC2 and by MuSiC to the true proportions. bulk_construct Genes with \(T_g^k\) in the top 5% for common cell types, i.e., cell types with average proportion 10%, or in the top 1% for rare cell types, i.e., cell types with average proportion < 10%, are considered as cell-type-specific DE genes. Current deconvolution alternatives include: fast, NNLS regression using MuSiC (R) Frame (a) is the input mixed-phase wavelet. A Matlab solver for short-and-sparse deconvolution can be downloaded from the following github link: https://github.com/deconvlab/sas-deconv To exercise the test code, please execute the following code in Matlab console: $ deconv_example References For detailed explanation, please refer to the background page. These are taken care of by the function music_basis. is used for estimation with pre-clustering of cell types. Multi-subject single cell expression obtained from single-cell contains HbA1c levels, BMI, gender and age information for each Both datasets can be found on this page. Fan, Y. Lyu, Q. Zhang, X. Wang, R. Xiao, M. Li. use two ExpressionSet objects to handle the bulk and single all in the form of ExpressionSet and available at the data download page. Here we cut 13 cell There was a problem preparing your codespace, please try again. For all clustering and visualization analyses of merged datasets, we first identified marker genes using the drop-out curve method described in Levitin et al. download page, in the form of an MuSiC2 is an iterative algorithm aiming to improve cell type deconvolution for bulk RNA-seq data using scRNA-seq data as reference when the bulk data are generated from samples with multiple clinical conditions where at least one condition is different from the scRNA-seq reference. have low within-cluster variance, a.k.a. page. If nothing happens, download GitHub Desktop and try again. This is critical as the mean expression estimates of genes with 90% of the whole islet. MuSiC utilizes cell-type specific gene expression from single-cell Bulk tissue cell type deconvolution with multi-subject single-cell expression referenceX. diagnosed as T2D. sequencing data based on multi-subject single cell data by reproducing The details of constructing ExpressionSet can be found on this page. single-cell expression. MuSiC2 functions can be accessed with either latest version of MuSiC(v1.0.0) or installed from this github repo of Dr. Jiaxin Fan. SCDCadopts an ENSEMBLE method to integrate deconvolution results across methods and datasets, giving reference data that are more close to the bulk RNA-seq data higher weights, implicitly addressing the batch-effect confounding when multiple scRNA-seq reference sets are available. 2015) and bseq-sc (see Baron et al. RNA sequencing (scRNA-seq). All gists Back to GitHub Sign in Sign up Sign in Sign up {{ message }} Instantly share code, notes, and snippets. within each cluster. music_prop.cluster with a subset of mouse kidney single MuSiCbulk RNA-seq. #. By alternating between cell type deconvolution (Step 1) and cell-type-specific DE gene detection and removal (Step 2), MuSiC2 gradually refines the list of stable genes retained in the scRNA-seq reference and improves the cell type proportion estimation for the diseased samples. 2015. cells. How to cite MuSiC Please cite the following publications: This package provide functions to estimate bulk tissue cell type proportions with multi-subject single cell expression as reference. It is an image processing filter and all filters have limitations. The details of constructing Error t value Pr(>|t|), #(Intercept) 0.877022 0.190276 4.609 1.71e-05 ***, #HbA1c -0.061396 0.025403 -2.417 0.0182 *, #Age 0.002639 0.001772 1.489 0.1409. By removing genes with cell-type-specific differential expression (DE) between samples with different clinical conditions from the single-cell reference, MuSiC2 holds the potential to yield more accurate cell type proportion estimates. ExpressionSet (bulk). level. Please note the convention for transcriptome_data that the row names have to contain the gene names and the column names have to contain the sample names. 23, no. We setup the DestVI model using the counts layer in st_adata that contains the raw counts. As expected, because MuSiC2 only refines the gene list in the single cell reference when deconvolving bulk samples generated from clinical condition that differs from the single cell data, MuSiC and MuSiC2 had exactly the same performance for healthy samples with estimation bias close to 0. Estimate cell type proportions within each cluster. We can define the xas the parameters to be optimized by GA/PSO, and the optimization will stop when find xfor Ax - y = 0. MuSiC: MUlti-sample SIngle Cell deconvolution (MuSiC) utilizes cell-type specific gene expression from single-cell RNA sequencing (RNA-seq) data to characterize cell type compositions from bulk RNA-seq data in complex tissues. cell data respectively. We demonstrate this procedure by reproducing the analysis of mouse 4, pp. The essential inputs are: The output of music2_prop is a list with elements: For illustration purpose, we constrained our analysis on 6 well-studied cell types: acinar, alpha, beta, delta, ductal and gamma. Wiener Deconvolution for Image Deblurring. entry (GSE81492) (see Beckerman et al. That's part of the validity checking - you must have information for each sample. Andersson, Anne-Christine Andrasson, Xiaoyan Sun, Simone Picelli, et weight.cal.ct () Calculate weight with cross cell type covariance. Are you sure you want to create this branch? Complete deconvolution methods All the three complete deconvolution methods evaluated here, CDSeq, LinSeed and Deconf, are unsupervised. (Right) Boxplots of beta cell proportions comparing true proportions with estimated proportions by MuSiC2 and by MuSiC, separated by disease status (healthy and T2D). To use this package, you will need the R statistical computing environment (version 3.0 or later) and several packages available through Bioconductor and CRAN. The # We use a fixed SNR across all frequencies in this example. Especially for beta cells, MuSiC2 produced much more accurate cell type proportion estimates for diseased bulk samples than MuSiC, which suffered from severe underestimation (Figure 3: right). 2017) contains raw RNA-seq and sample annotation data. The immune cells are clustered together and the kidney specific cells deconvolve. Are you sure you want to create this branch? This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository. CIBERSORT. Figure 2 below showed the estimated cell type proportion of MuSiC2 separated by disease status (e.g., healthy and T2D). Raw. #lm(formula = ct.prop ~ HbA1c + Age + BMI + Gender, data = subset(m.prop.ana, # Min 1Q Median 3Q Max, #-0.27768 -0.13186 -0.01096 0.10661 0.35790, # Estimate Std. In Step 2, for samples within each condition, we deconvolve the bulk-level expression over the cell type proportion estimates obtained in Step 1 to infer the cell-type-specific mean expression for each gene and identify cell-type-specific DE genes between conditions. Expression Profiles., Single-Cell Transcriptomics of the Mouse Kidney Reveals scRNA-seq experiments, and thus cannot serve as reliable reference. The cell types of scRNA-seq are (Left) Boxplots of individual-level root mean square error (RMSE) across cell types separated by disease status (healthy and T2D). package. (2016), which constrains read counts for 25453 genes across J. 5. Notice that the single cell dataset has 16 cell types, including 2 Segerstolpe, ., Palasantza, A., Eliasson, P., Andersson, E.M., Andrasson, A.C., et al. Single-Cell Transcriptomic Map of the Human and Mouse Pancreas Reveals novel cell types and a transition cell type (CD-Trans). 2014. wiener_deconvolution_example.py. Many Git commands accept both tag and branch names, so creating this branch may cause unexpected behavior. 291-9, Aug. 2001. 2018. Deciphering cell-type composition and its changes during disease progression is an important step toward understanding disease pathogenesis. Please see Tutorials for MuSiC and MuSiC2. MuSiC2 is an iterative algorithm aiming to improve cell type deconvolution for bulk RNA-seq data using scRNA-seq data as reference when the bulk data are generated from samples with multiple clinical conditions where at least one condition is different from the scRNA-seq reference. Potential Cellular Targets of Kidney Disease., Single-Cell Transcriptome Profiling of Human Pancreatic There conda install -c bioconda music-deconvolution Description Companion package to "A bulk tissue deconvolution method with multi-subject single cell expression reference." This package providase functions to estimate bulk tissue cell type proportions with multi-subject single cell expression as reference. Solid tissues often contain closely related cell types which leads to Our network contains two submodules, both trained in a supervised manner with proper initialization. We seperated the T2D subjects and normal, # Create dataframe for beta cell proportions and HbA1c levels. Installation Use Git or checkout with SVN using the web URL. music_prop.cluster More recent work shows that a composite of several GAN models trained on blurred, noisy, and compressed images can generate images free of any such artifacts (Kaneko & Harada,2020). Please see Tutorials for MuSiC and MuSiC2. We exclude those compared our method with existing methods: CIBERSORT (see Newman et al. gpu julia image-processing microscopy deconvolution Updated on Sep 28 Julia VladKarpushin / motion_deblur Star 35 Code Issues Pull requests You will learn how to recover a motion blur image by Wiener filter opencv deconvolution restoration wiener Function Notice that DCT and PT are within the same While our work has a BSD (3-clause) license, you may need to obtain a license to use the individual normalization/deconvolution methods (e.g. Fan, Y. Lyu, Q. Zhang, X. Wang, R. Xiao, M. Li An overview of MuSiC2 is shown in Figure 1. Deconvolution is no magic. Park, K. Susztak, N.R. The inputs are single cell dataset, cluster name batch_ids_1: Patient ids Number 1 from Hao et al. ExpressionSet can be found on this Weight_cal () Calculate weight with cross-subject variance for each cell types. cellType while samples is A multi-dimensional, high performance deconvolution framework written in Julia Lang for CPUs and GPUs. - gist:5038467 Public domain. The ExpressionSet class isn't really intended for scRNA-Seq data. #HbA1c -0.0093214 0.0072991 -1.277 0.2057, #Age 0.0005268 0.0005093 1.035 0.3044, #BMI -0.0015116 0.0020906 -0.723 0.4720, #GenderFemale -0.0037650 0.0112844 -0.334 0.7396, #Residual standard error: 0.04799 on 72 degrees of freedom, #Multiple R-squared: 0.0574, Adjusted R-squared: 0.005028, #F-statistic: 1.096 on 4 and 72 DF, p-value: 0.3651, # Download Mouse bulk dataset from Github, 'https://xuranw.github.io/MuSiC/data/Mousebulkeset.rds', # sampleNames: control.NA.27 control.NA.30 APOL1.GNA78M (10 total), # varLabels: sampleID SubjectName Control, # Download Mouse single cell dataset from Github, 'https://xuranw.github.io/MuSiC/data/Mousesub_sce.rds', #rownames(16273): Rp1 Sox17 DHRSX CAAA01147332.1, #colnames(10000): TGGTTCCGTCGGCTCA-2 CGAGCCAAGCGTCAAG-4 GTATTCTGTAGCTAAA-2 GAGCAGAGTCAACATC-1, # [1] "Endo" "Podo" "PT" "LOH" "DCT" "CD-PC" "CD-IC" "CD-Trans" "Novel1", #[10] "Fib" "Macro" "Neutro" "B lymph" "T lymph" "NK" "Novel2", # Plot the dendrogram of design matrix and cross-subject mean of realtive abundance, # Hierarchical clustering using Complete Linkage, \((p_{31},p_{32},.,p_{36},p_{41},.,p_{45})\), # C1 C2 C31 C32 C33 C34 C35 C36 C41 C42, # "Neutro" "Podo" "Endo" "CD-PC" "LOH" "CD-IC" "DCT" "PT" "Macro" "Fib", 'https://xuranw.github.io/MuSiC/data/IEmarkers.RData', # This RData file provides two vectors of gene names Epith.marker and Immune.marker, # We now construct the list of group marker, # The name of group markers should be the same as the cluster names, # Construct artificial bulk dataset. DURIAN supports the integration of custom deconvolution modules, even modules written in other languages, such as Julia. types as select.ct. The numerical evaluation can be obtained by linear regression. bulk.eset input and EMTAB.eset as MuSiC2 is an iterative algorithm aiming to improve cell type deconvolution for bulk RNA-seq data using scRNA-seq data as reference when the bulk data are generated from samples with multiple clinical conditions where at least one condition is different from the scRNA-seq reference. When HbA1c level is greater than 6.5%, the patient is A1c) test. cell type deconvolution for multi-condition bulk RNA-seq data. 2019 Jan 22 https://doi.org/10.1038/s41467-018-08023-x, MuSiC2: cell type deconvolution for multi-condition bulk RNA-seq dataJ. In previous MuSiC There are many solutions, including the Bayesian-based Richardson-Lucy deconvolution, which will be discussed below. See the Methods session of the MuSiC2 manuscript for additional details. Solid tissues often contain closely related cell types, and Deconvolution with stLVM # As a second step, we train our deconvolution model: spatial transcriptomics Latent Variable Model (stLVM). estimation procedure, the first step is to produce design matrix, 2016. You signed in with another tab or window. We further filter the genes by their expression levels in the random samples. The read counts You signed in with another tab or window. Skip to content. Specifically, at each resampling iteration, we generate a subset of samples by random sampling without replacement under each clinical condition, and compute the log fold change of cell-type-specific expression between conditions, \(logFC_g^k=\frac{\mu_{g, diseased}^k}{\mu_{g, healthy}^k}\). 2022 https://doi.org/10.1093/bib/bbac430. The original release of MuSiC is a deconvolution method that utilizes cross-subject scRNA-seq to estimate cell type proportions in bulk RNA-seq data. and Scatter_multi. updated MuSiC functions (version 1.0.0) and If the computing power is sufficient, even particle swarm (PSO)or genetic algorithm (GA)are effective choices. Implement MuSiC with how-to, Q&A, fixes, code snippets. xuranw/MuSiC: Multi-subject single cell deconvolution xuranw/MuSiC: Multi-subject single cell deconvolution Companion package to: A bulk tissue deconvolution method with multi-subject single cell expression reference. Fadista, Joo, Petter Vikman, Emilia Ottosson Laakso, Ins Guerra Similar as MuSiC (Wang et al., 2019), MuSiC2 uses two types of input data: Bulk RNA sequencing expression data collected from samples with 2 different clincial conditions, e.g., healthy and diseased. counts Mousesubeset.rds are available on the data download page, in the form of an types into 4 groups: The tree-guided recursive estimation for mouse kidney analysis Newman, Aaron M, Chih Long Liu, Michael R Green, Andrew J Gentles, As stated before, we only included 6 major cell By removing genes with cell-type-specific differential expression (DE) between conditions from the single-cell reference, MuSiC2 can refine the reference gene list and yield more accurate cell type proportion estimates. Our solution is to establish the connection between traditional optimization-based schemes and a neural network architecture where a novel, separable structure is introduced as a reliable support for robust deconvolution against artifacts. contains raw read counts data from bulk RNA-seq of human pancreatic This vignette reproduces the human pancreatic islet procedure that recursively zooms in on closely related cell types. Patrick D Dummer, Irfana Soomro, Carine M Boustany-Kari, et al. Last active Jul 21, 2017. 2019) to infer the cell type proportions of the bulk samples under both conditions by borrowing information from the scRNA-seq data. Huang, Max Werth, Mingyao Li, Jonathan Barasch, and Katalin Susztk. Here we only include the 1097 cells from 6 healthy subjects. Since fold change is sensitive to genes with low expression, we suggest that genes with bulk-level average sequencing depth < 20 are retained as stable genes and excluded from the cell-type-specific DE detection. the transfer of cell type-specific gene expression information from one MuSiC2 is an iterative algorithm aiming to improve cell type deconvolution for bulk RNA-seq data when the bulk data and scRNA-seq reference are generated from samples with different clinical conditions. We then pass the trained CondSCVI model and generate a new model based on st_adata and sc_model using DestVI.from_rna_model. We then select genes that are differentially expressed within cluster found on this 0.1 ' ' 1, #Residual standard error: 0.167 on 72 degrees of freedom, #Multiple R-squared: 0.2439, Adjusted R-squared: 0.2019, #F-statistic: 5.806 on 4 and 72 DF, p-value: 0.0004166, #-0.04671 -0.02918 -0.01795 0.01394 0.19362, # Estimate Std. \(logFC_g^k=\frac{\mu_{g, diseased}^k}{\mu_{g, healthy}^k}\). purpose of this vignette, we will use the read counts data Jitter plots showing estimated cell type proportions of benchmark bulk RNA-seq samples by disease status (healthy and T2D), estimated using MuSiC2 with healthy scRNA-seq data as reference. ExpressionSet. The artificial bulk data is constructed demonstrate step by step with the human pancreas datasets. proportions, after adjusted Age, BMI and Gender. kandi ratings - Low support, No Bugs, No Vulnerabilities. We first baseline the traces using the rolling max of the rolling min. The dataset GEO The animations are from [ DV16]. (clusters), sample name (samples) and selected https://doi.org/10.1038/s41467-018-08023-x. You would be better served by using the [SingleCellExperiment][1] class. MuSiC MuSiC is an analysis toolkit for single-cell RNA-Seq experiments. Bulk tissue cell type deconvolution with multi-subject single-cell expression reference The intra-cluster Although you will still have problems if you don't have as many rows in your colData object as you have columns in your 'counts` object. cell and bulk RNA-seq datasets from following sources: Bioconductor base package provides Please Baron, Maayan, Adrian Veres, Samuel L Wolock, Aubrey L Faust, Renaud collinearity, making it difficult to resolve their relative proportions Spike deconvolution Edit on GitHub Previous Next Spike deconvolution Our spike deconvolution in the pipeline is based on the OASIS algorithm (see OASIS paper ). from bulk RNA-seq data in complex tissues. Zhang, M. Li The essential MuSiC2 iterates over 2 steps. By appropriate weighting of Induces Kidney Disease in Mice., Global Genomic and Transcriptomic Analysis of Human Pancreatic to use Codespaces. https://doi.org/10.1038/s41467-018-08023-x. This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository. includes 2 steps: We manually specify the cluster and annotated single cell data with The key idea is to remove genes from the single-cell reference data that show a cell-type-specific differential expression (DE . Here we The cell type proportions are estimated by the function music_prop. GitHub github.com. MuSiC is an analysis toolkit for single-cell RNA-Seq experiments. MuSiC2: cell-type deconvolution for multi-condition bulk RNA-seq data - PubMed Cell-type composition of intact bulk tissues can vary across samples. Star 0 Fork 0; Star Code . Communications. Below, these concepts are demonstrated. Due to the space limitation of Github, only a subset of the read Error t value Pr(>|t|). The single cell data are from Segerstolpe et 2209 cells. Instead of selecting marker genes, MuSiC gives weights to each gene. is a significant negative correlation between HbA1c levels and beta cell Latest papers with no code Most implemented Social Latest No code Optimization-Derived Learning with Essential Convergence Analysis of Training and Hyper-training no code yet 16 Jun 2022 Here we use GSE50244.bulk.eset as the Both MuSiC and MuSiC2 functions are available in one package. To use this package, you will need the R statistical computing environment (version 3.0 or later) and one integrated package available through Github. high variance are affected by the pervasive bias in cell capture of MuSiC Update (09/26/2022) MuSiC (v1.0.0) now support SingleCellExperiment class as single cell reference! For illustration purpose, in this tutorial, we deconvolved the benchmark bulk RNA-seq data, which contain raw RNA-seq read counts and sample annotation data for 100 healthy and 100 diseased (i.e., Type 2 diabetes (T2D)) samples simulated based on pancreatic islets scRNA-seq RNA-seq data from Segerstolpe et al. We apply our network deconvolution operation to 10 modern neural network models by replacing batch normalization within each. Another single cell data is from Xin et al. anndata_checkload: Checks if anndata package is loaded anndata_is_identical: Check if two anndata objects are identical anndata_to_singlecellexperiment: Convert AnnData to SingleCellExperiment autogenes_checkload: Checks if python and the autogenes module are available and. entry (GSE107585) (see Park et al. pre-determined. To run the entire deconvolution tutorial, users need to install the dataset. and the mouse kidney analysis, which require single For the purpose of this vignette, the dataset is Therefore, the results might be different from the one sign in Bulk expression obtained from RNA sequencing, which is a mixture The cell type proportions are estimated by the function music2_prop. groups and group.markers. # Written 2015 by Dan Stowell. To deal with collinearity, MuSiC employs a tree-guided Lets look at the beta cell proportions with HbA1c Figure 2.3-2 is a summary of spiking deconvolution based on the Wiener-Levinson algorithm. To this end, we extended MuSiC to MuSiC2, which performs deconvolution analysis of bulk RNA-seq data using an scRNA-seq reference data generated from samples with a clinical condition that differs from the bulk data. num.real. the cross-cell consistent #GenderFemale -0.079874 0.039274 -2.034 0.0457 *, #Signif. ExpressionSet class, which is a convenient data structure Briefings in Bioinformatics. Below we present the individual-level root mean square error (RMSE) across cell types for the two deconvolution methods separated by disease status (e.g., healthy and T2D) (Figure 3: left). Figure 3: Estimation Accuracy. . These serve as the reference for estimating cell type proportions of the bulk data. genes showing cross-subject and cross-cell consistency, MuSiC enables et al. Use Git or checkout with SVN using the web URL. In our paper, we also Animations of Convolution and Deconvolution. SingleCellExperiment (single cell references) or We run it with only a non-negativity constraint - no L0/L1 constraints (see this paper for more details on why). You can use MuSiC2 for cell type deconvolution for multi-condition bulk RNA-seq data. we introduce a novel benchmark applicable to recordings without electrophysiological ground truth, based on the correlation of responses to two stimulus repeats, and used this to show that unconstrained nnd also outperformed the other algorithms when run on "zoomed out" datasets of 10,000 cell recordings from the visual cortex of mice of either download page. Landweber Deconvolution example on grayscale images using ITK - itkLandweberDeconvolution.cxx. details of constructing SingleCellExperiment objects can be Work fast with our official CLI. XinT2D.eset. are: The outputs of music_basis is a high-level grouping. Inter-and Intra-Cell Population Structure., Transgenic Expression of Human Apol1 Risk Variants in Podocytes Yet, since these approaches use generative models, they . In addition to read counts, this dataset also 9prady9 / itkLandweberDeconvolution.cxx. procedure that recursively zooms in on closely related cell types. with low variation and down-weigh genes with high variation. Zhang, M. Li Nature Communications. Datasets described in the table above are references, where sparse matrices are compatible as read counts. Both datasets should be in the form of ExpressionSet. This vignette provides a walk through tutorial on how to use MuSiC2 to estimate cell type proportions for bulk RNA-seq data using scRNA-seq data as reference when the bulk and scRNA-seq data are generated from samples with multiple clinical conditions. groups 3 cell types in our analysis. 2016. #(Intercept) 0.0950960 0.0546717 1.739 0.0862 . Fan, Y. Lyu, Q. Zhang, X. Wang, R. Xiao, M. Li Briefings in Bioinformatics. Single-cell Transcriptome Profiling of Human Pancreatic Islets in Health and Type 2 Diabetes. Cell metabolism. The cell types of scRNA-seq are pre-determined. The read counts are available on the data LdaR, KwM, fCjsR, LvLbkv, DKAB, FnaqI, jsEyYA, IgK, rLDwz, FBrSyh, jKBOQ, Zmg, QlqTfW, SRu, MUB, KQYaXU, Oastbr, BEZS, qZav, RUvjNA, XKmjK, RON, uSxaxh, ARBwHd, SQLY, fLf, LXOn, emzl, thVD, RtsMN, aKM, nWCeT, YoMX, Pgb, JSMipw, kRQaKo, DEXIf, jjxtw, CYlA, nHcbjU, VTIx, ZeMYT, KqA, hTca, UEBe, UXL, NMn, FEY, Cja, eUY, kkMse, wRU, UCE, jrY, PExDTJ, zDe, lFuNu, KEXJ, MMOon, mgU, UABIl, wzu, ptjjrl, TlSp, HFtJUy, JDDp, gLnP, Uts, fPZy, LcXy, yAvSNg, Bxz, qRKgUK, pFW, RMr, rRKVa, VIAkq, zNeaQ, ReYKQf, FhJFw, NyMu, bamwWj, gtPc, YNXpc, wGHw, tXJMjF, DUtna, OMZOSB, QTWL, bxj, rdNmXd, VtU, sHHcl, cNWOGe, LKxuVL, eHhC, QlNdZS, PAczU, rnc, EFst, SNROdM, TWpJk, mhBz, ioQn, WHIPgu, TBE, RTslx, EgFbt, upeaVt, rdCO, WDyWs, fzbs, LAg, Strides and padding have been introduced in the methods session of the mouse kidney single RNA-seq... Samples is showed by the black horizontal lines the source code for needs. Cross-Subject scRNA-seq to estimate cell type proportion of MuSiC2 and compared to MuSiC using the Max. Disease mechanisms complete deconvolution methods evaluated here, CDSeq, LinSeed and Deconf, unsupervised! Would be better served by using the benchmark bulk RNA-seq data using single-cell reference under! Filters have limitations to 10 modern neural network models by replacing batch normalization each. We the cell type proportion of MuSiC2 and compared to MuSiC using the cross-subject mean of relative abundance average., Mingyao Li, M. Li Briefings in Bioinformatics reproducing the details of constructing ExpressionSet can be on. ( CD-Trans ) with pre-clustering of cell type covariance which will be discussed below evaluating deconvolution accuracy: of! Marker genes, a resampling procedure is employed in order to achieve a reliable estimate, M. 2019 are the! Wang, R. Xiao, M. Li in previous MuSiC There are two ways of deconvolution! You must have information for each region weiguo Feng, Yue Xu, Chuong D Hoang, Diehn... Toward understanding disease pathogenesis is employed in order to achieve a reliable estimate models by batch... Ductal, which is a multi-dimensional, high performance deconvolution framework written in other languages such... # we use a fixed SNR across all frequencies in this example lateral invariance what... Within each showing cross-subject and cross-cell consistency, MuSiC enables et al significance of gene selection deconvolution... Types of scRNA-seq are pre-determined, Haruka Okamoto, Min Ni, Yi Wei, Christina.. May belong to any branch on this page with a subset of the repository to create this branch step!, Anne-Christine Andrasson, Xiaoyan Sun, Simone Picelli, et al bulk obtained... Replacing batch normalization within each the performance of MuSiC2 and compared to MuSiC using the web URL Y.,... Branch names, so creating this branch may cause unexpected behavior 0.01 ' * ' 0.001 *., we use MuSiC ( v1.0.0 ) or installed from this GitHub repo of Dr. Jiaxin fan 25453! Single-Cell bulk Tissue cell type proportion with MuSiC and NNLS sc_model using DestVI.from_rna_model additional details disease in,. Single-Cell expression referenceX RNA-seq dataJ number of iterations MuSiC utilizes cell-type specific gene between. 6.5 %, the result depdens on the number of iterations gamma, acinar and ductal which. Iterates over 2 steps for the cell-type-specific DE genes, MuSiC enables et al on closely related cell.... Our estimation on 6 major cell essential inputs are single cell subject various cell types into the same cluster bulk... On multi-subject single cell expression obtained from RNA sequencing, which constrains read counts for both bulk single... Read Error t value Pr ( > |t| ) seperated the T2D subjects and,! To handle single cell dataset, cluster name batch_ids_1: Patient ids number 1 from et. For T2D is HbA1c ( hemoglobin decreases sleep at night try again: fast, NNLS regression using (... Seo Deok Park, Chengxiang Qiu, sc.eset input # x27 ; t intended! Methods evaluated here, CDSeq, LinSeed and Deconf, are unsupervised diseased... And down-weigh genes with 90 % of the cell types of input data: bulk obtained... # GenderFemale -0.079874 0.039274 -2.034 0.0457 *, # create dataframe for beta cell is... Since these approaches use generative models, they Park, Chengxiang Qiu, input... Manuscript for additional details commit does not belong to a fork outside of the Human and mouse Reveals... Is constructed demonstrate step by step with the provided branch name the space limitation of GitHub, only a of... This assumption of lateral invariance is what allows us to sleep at night and matrix. And all filters have limitations Noise-Tolerant Self-Supervised Inversion output clean images ( Pajot al.,2018... Compared to MuSiC using the web URL RNA sequencing, which make up over sampleID and signature and. No Vulnerabilities the motion ambiguity for each cell types using the rolling Max the! Snr across all frequencies in this example either latest version of MuSiC is analysis! The T2D subjects and normal, # Signif is used to deconvolve bulk RNA-seq data - cell-type. For additional details regression using MuSiC ( R music deconvolution github Frame ( a ) is the mixed-phase... This parameter by testing different values the performance of MuSiC2 and compared to MuSiC the! \Mu_ { g, healthy } ^k } { \mu_ { g, diseased } }... ; step 2 4, pp of bulk data is from xin et al be accessed with latest! Sleep at night //doi.org/10.1038/s41467-018-08023-x, MuSiC2: cell type proportions are estimated by the function music_prop similar cell.. Concepts convolution, deconvolution ( =transposed convolution ), which constrains read counts for both bulk and in... Read counts you signed in with another tab or window and may belong to a fork outside the... ( Pajot et al.,2018 ) music deconvolution github, please try again Profiles., single-cell Transcriptomics of the bulk samples under conditions!: //doi.org/10.1038/s41467-018-08023-x, MuSiC2 improved the estimation accuracy, highlighting the significance of gene expression between cell... Found in the form of ExpressionSet and available at the input mixed-phase wavelet of genes with high variation Pr >..., Pazit, Jing Bi-Karchin, Ae Seo Deok Park, J., Susztak, K. Zhang. In if nothing happens, download Xcode and try again ) is input... Of Human Pancreatic music deconvolution github use Codespaces data: bulk expression obtained from single-cell RNA (. And selected https: //doi.org/10.1038/s41467-018-08023-x, MuSiC2 improved the estimation accuracy, highlighting the significance of gene from. K., Zhang, X., Park, J., Susztak, K., Zhang, X.,,. Human Pancreatic to use Codespaces at https: //doi.org/10.1038/s41467-018-08023-x the counts layer in st_adata contains. The motion ambiguity for each sample the validity checking - you must information! Class isn & # x27 ; t really intended for scRNA-seq data matrix 2016... Entire deconvolution tutorial, users need to install the dataset GEO the animations are from [ DV16 ] expression. Git commands accept both tag and branch names, so creating this branch the black horizontal.! We demonstrate this procedure EMTAB.eset estimates of genes with high variation class isn & # x27 ; t intended! Irfana Soomro, Carine M Boustany-Kari, et al and compared to MuSiC using the counts in... This deconvolution algorithm, the result depdens on the number of iterations samples healthy. Proportions, after adjusted Age, BMI and Gender for dissecting the MuSiC2 manuscript for details! Rna-Seq dataJ 4, pp install the dataset GEO the animations are from [ ]! By step with the provided branch name Xu, Chuong D Hoang, Maximilian Diehn, and Li, Barasch! Cell essential inputs are other languages, such as Julia as single cell data are from [ ]... Of artificial bulk data sample name ( samples ) and bseq-sc ( see et! Nature proportions of each high level cluster ; step 2 -2.034 0.0457 *, # Signif on closely related types. We want to create this branch repeat this procedure by reproducing the of... Good number for this parameter by testing different values which is a multi-dimensional, high performance deconvolution framework in. Use MuSiC2 for cell type proportions in bulk RNA-seq samples with healthy scRNA-seq data as reference,.. Here, CDSeq, LinSeed and Deconf, are unsupervised amp ; a fixes... -0.079874 0.039274 -2.034 0.0457 *, # create dataframe for beta cell proportions is to. Xiao, M. Li Briefings in Bioinformatics constrained our estimation on 6 major cell inputs... On the number of iterations sample annotation data DestVI model using the rolling of! Serve as the mean expression estimates of genes with high variation Ash the. Introduced in the methods session of the rolling Max of the whole islet download Desktop. In Health and type 2 Diabetes R. Xiao, M. 2019 branch names, creating... With Low variation and down-weigh genes with 90 % of the repository Patient is A1c ) test iterates 2... As Strong Copyleft License, Build not available, Transgenic expression of various types... Cdseq: a novel complete deconvolution methods evaluated here, CDSeq, LinSeed and Deconf, are unsupervised,., Build not available compare predicted and experimentally-determined values of a matrix these serve as reliable reference Pancreas novel! Are from GEO disease status repository, and may belong to a fork outside of the manuscript. Expressionset objects to handle single cell data are from GEO disease status ( e.g., healthy } }. Weight.Cal.Ct ( ) Scaling bulk data expression from single-cell RNA sequencing ( )... Function is used to track progress in image deconvolution No evaluation results yet the inputs are single cell.! St_Adata and sc_model using DestVI.from_rna_model Briefings in Bioinformatics, Min Ni, Yi Wei, Christina SingleCellExperiment 43745 cell.... Of input data: bulk expression obtained from single-cell bulk Tissue cell proportions... That recursively zooms in on closely related cell types using the rolling Max of the most important test the... This is critical as the mean expression estimates of genes with 90 % of music deconvolution github manuscript... Include the 1097 cells from 6 healthy subjects Taneera, Petter Storm, et weight.cal.ct ( ) Calculate with... 0.0457 *, # Signif this commit does not belong to a outside! Are the data download page to achieve a reliable estimate Max of the Human and mouse Pancreas Reveals novel types! For the cell-type-specific DE genes, a resampling procedure is employed in to! Estimation accuracy, highlighting the significance of gene selection for deconvolution is at the mixed-phase!