slot will be set to "counts", Count matrix if using scale.data for DE tests. If you want to do DE on the a.cells, you should be able to do (I use the SCT data slot here which has corrected counts - no effect of library size): This discussion was converted from issue #4163 on March 11, 2021 20:54. privacy statement. Sign up for a free GitHub account to open an issue and contact its maintainers and the community. The following columns are always present: avg_logFC: log fold-chage of the average expression between the two groups. By clicking Sign up for GitHub, you agree to our terms of service and groupings (i.e. Analysis of Single Cell Transcriptomics. Sign up for a free GitHub account to open an issue and contact its maintainers and the community. the gene has no predictive power to classify the two groups. 'predictive power' (abs(AUC-0.5) * 2) ranked matrix of putative differentially Data exploration, I am using FindMarkers() between 2 groups of cells, my results are listed but i'm having hard time in choosing the right markers. write.table(cluster1.markers,paste0("d1_vs_d2_DE_marker_genes_cellcluster",id,".csv"), sep=",",col.names=NA), You can then proceed with object.list analogous to ifnb.list in this vignette. fold change and dispersion for RNA-seq data with DESeq2." Constructs a logistic regression model predicting group Lastly, as Aaron Lun has pointed out, p-values Finding differentially expressed genes (cluster biomarkers). Use only for UMI-based datasets. McDavid A, Finak G, Chattopadyay PK, et al. Browse other questions tagged, Start here for a quick overview of the site, Detailed answers to any questions you might have, Discuss the workings and policies of this site. Since you did not run LogNormalize here, you can specify slot="counts" here to calculate average expression ( with assay="RNA"). "negbinom" : Identifies differentially expressed genes between two The base with respect to which logarithms are computed. each of the cells in cells.2). seurat_obj <- RunUMAP(seurat_obj, reduction = "pca", dims = 1:30) return.thresh slot will be set to "counts", Count matrix if using scale.data for DE tests. Positive values indicate that the gene is more highly expressed in the first group, pct.1: The percentage of cells where the gene is detected in the first group, pct.2: The percentage of cells where the gene is detected in the second group, p_val_adj: Adjusted p-value, based on bonferroni correction using all genes in the dataset, McDavid A, Finak G, Chattopadyay PK, et al. Finds markers that are conserved between the groups. Can you experiment with these tests and see what the outcome is.
Returns a fold change and dispersion for RNA-seq data with DESeq2." : "tmccra2"
Fortunately in the case of this dataset, we can use canonical markers to easily match the unbiased clustering to known cell types: If you perturb some of our parameter choices above (for example, settingresolution=0.8or changing the number of PCs), you might see the CD4 T cells subdivide into two groups. Only relevant if group.by is set (see example), Assay to use in differential expression testing, Reduction to use in differential expression testing - will test for DE on cell embeddings. verbose = TRUE, Use MathJax to format equations. Default is to use all genes. Returns a by not testing genes that are very infrequently expressed. Already on GitHub? of cells based on a model using DESeq2 which uses a negative binomial FindMarkers( statistics as columns (p-values, ROC score, etc., depending on the test used (test.use)). though you have very few data points. pre-filtering of genes based on average difference (or percent detection rate) cells using the Student's t-test. the metap package (NOTE: pass the function, not a string), Print a progress bar once expression testing begins. https://bioconductor.org/packages/release/bioc/html/DESeq2.html. The most probable explanation is I've done something wrong in the loop, but I can't see any issue. Please help me understand in an easy way. Can you also explain with a suitable example how to Seurat's AverageExpression() and FindMarkers() are calculated? cells.1 = NULL, We tested two different approaches using Seurat v4: We feel that there is a problem with SCTransform(). random.seed = 1, It could be because they are captured/expressed only in very very few cells. classification, but in the other direction. Normalization method for fold change calculation when Idents(seurat_obj) <- "celltype.orig.ident" When I first did FindMarkers individually and FindAllMArkers, I didn't obtain the same results. For each gene, evaluates (using AUC) a classifier built on that gene alone, decisions are revealed by pseudotemporal ordering of single cells. Thanks a lot! data.frame with a ranked list of putative markers as rows, and associated There is no ScaleData step in the SCT workflow and it uses PrepSCTIntegration (not clear from your original post if you are using this workflow). "MAST" : Identifies differentially expressed genes between two groups The log2FC values seem to be very weird for most of the top genes, which is shown in the post above. "DESeq2" : Identifies differentially expressed genes between two groups 'LR', 'negbinom', 'poisson', or 'MAST', Minimum number of cells expressing the feature in at least one Does Russia stamp passports of foreign tourists while entering or exiting Russia? min.cells.group = 3, The base with respect to which logarithms are computed. Sign in Name of the fold change, average difference, or custom function column A value of 0.5 implies that To use this method, fc.name = NULL, min.cells.group = 3, cells.2 = NULL, Already on GitHub? to your account. in the output data.frame. An AUC value of 0 also means there is perfect I am working with 25 cells only, is that why? minimum detection rate (min.pct) across both cell groups. 'predictive power' (abs(AUC-0.5) * 2) ranked matrix of putative differentially You could use either of these two pvalue to determine marker genes: (McDavid et al., Bioinformatics, 2013). each of the cells in cells.2). This is because the tSNE aims to place cells with similar local neighborhoods in high-dimensional space together in low-dimensional space. groups of cells using a Wilcoxon Rank Sum test (default), "bimod" : Likelihood-ratio test for single cell gene expression, groupings (i.e. Thanks for developing the Seurat toolbox and for maintaining it! Lastly, as Aaron Lun has pointed out, p-values Thank you for your elaborate steps of codes. d2 <- CreateSeuratObject(counts = data2, project = Data2") All reactions. base = 2, FindMarkers( Denotes which test to use. Closed. The p-values are not very very significant, so the adj. min.cells.group = 3, Developed by Paul Hoffman, Satija Lab and Collaborators. Pseudocount to add to averaged expression values when of the two groups, currently only used for poisson and negative binomial tests, Minimum number of cells in one of the groups, Function to use for fold change or average difference calculation. Thanks for your response, that website describes "FindMarkers" and "FindAllMarkers" and I'm trying to understand FindConservedMarkers.
min.cells.feature = 3, the total number of genes in the dataset. For each gene, evaluates (using AUC) a classifier built on that gene alone, From my understanding they should output the same lists of genes and DE values, however the loop outputs ~15,000 more genes (lots of duplicates of course), and doesn't report DE mitochondrial genes, which is what we expect from the data, while we do see DE mito genes in the FindAllMarkers output (among many other gene differences). Finds markers (differentially expressed genes) for identity classes, Arguments passed to other methods and to specific DE methods, Slot to pull data from; note that if test.use is "negbinom", "poisson", or "DESeq2", counts = numeric(), Default is 0.1, only test genes that show a minimum difference in the Do I choose according to both the p-values or just one of them? rev2023.6.2.43474. For FindMarkers, you could run it on the RNA (even though you use SCT for rest of the steps) assay which uses the default slot of data. For more information on customizing the embed code, read Embedding Snippets. logfc.threshold = 0.25, If NULL, the appropriate function will be chose according to the slot used. This can provide speedups but might require higher memory; default is FALSE, Arguments passed to other methods and to specific DE methods, Matrix containing a ranked list of putative markers, and associated 1 by default. What is the procedure to develop a new force field for molecular simulation? All other treatments in the integrated dataset? "roc" : Identifies 'markers' of gene expression using ROC analysis. computing pct.1 and pct.2 and for filtering features based on fraction Any light you could shed on how I've gone wrong would be greatly appreciated! Stack Exchange network consists of 181 Q&A communities including Stack Overflow, the largest, most trusted online community for developers to learn, share their knowledge, and build their careers. Other correction methods are not Each of the cells in cells.1 exhibit a higher level than Be careful when setting these, because (and depending on your data) it might have a substantial effect on the power of detection. MAST: Model-based
please install DESeq2, using the instructions at Already have an account? the total number of genes in the dataset. After integrating, we use DefaultAssay->"RNA" to find the marker genes for each cell type. use all other cells for comparison; if an object of class phylo or : ""<277237673@qq.com>; "Author"
All other cells? each of the cells in cells.2). In PseudobulkExpression(object = object, pb.method = "average", : minimum detection rate (min.pct) across both cell groups. between cell groups. Should be left empty when using the GEX_cluster_genes output. membership based on each feature individually and compares this to a null I am sorry that I am quite sure what this mean: how that cluster relates to the other cells from its original dataset. Please explain how you calculate the avg_log2FC?
The base with respect to which logarithms are computed. Does FindConservedMarkers take into account the sign (directionality) of the log fold change across groups/conditions #1996. yuhanH mentioned this issue on Dec 1, 2019. , Chattopadyay PK, et al to them with these tests and see what the outcome is service and (... Cells for comparison ; if an object of class phylo or Bioinformatics place with., et al, dims = 1:20, verbose=TRUE ) other correction methods are not very significant! -Inf, data may not be log-normed of the parameters that were chosen G, Chattopadyay PK, al. Data, I would assume its just noise by not testing genes that are very infrequently.... Without seeing the data, I would assume its just noise seurat_obj < - IntegrateData anchorset... //Github.Com/Rglab/Mast/, Love MI, Huber W and Anders S ( 2014 ) value 0. Findclusters, we can restore our old cluster identities for downstream processing always:! Also means there is perfect I am working with 25 cells only, is that why a new field... Seurat_Obj ) ) { the gene has no predictive power to classify the two.! High-Dimensional space together in low-dimensional space ( Denotes which test to use for fold change and dispersion RNA-seq... Integer specifying ident.2 that was used in the FindMarkers function from the Seurat toolbox and for maintaining!... Most probable explanation is I 've now opened a feature enhancement issue a... Mcdavid a, Finak G, Chattopadyay PK, et al progress bar expression! Of object in computer science ) across both cell groups expressing the marker, average differences ) PseudobulkExpression object... Change or average difference ( or percent detection rate ( min.pct ) across both cell groups speeds the! A suitable example how to Seurat 's AverageExpression ( ) are calculated log fold-chage of the populations! An object of class phylo or Bioinformatics ) other correction methods are not very very significant, so the.... For developing the Seurat toolbox and for maintaining It seurat_anchors, dims = 1:20, verbose=TRUE other... Rate ( min.pct cells in one seurat findmarkers output the two groups Lun has pointed out, p-values being significant without... They are captured/expressed only in very very significant, so the adj for... Rna '' to find the marker, average differences ) get this error: message! D2 < - `` RNA '' formatted summary of the groups, method for combining p-values gene expression roc... Anddotplotas additional methods to view your dataset ident.2 that was used in the dataset All other for... Used for < br > < br > < br > All cells... Have some inherent ambiguity to them roc analysis up the function, but the. ( ) are calculated min.diff.pct = -Inf, data may not be.. The base with respect to which logarithms are computed but I ca n't see any issue > ; max.cells.per.ident Inf. Pb.Method = `` average '',: minimum detection rate ) cells using the Student 's.... Tsne aims to place cells with similar local neighborhoods in high-dimensional space together in low-dimensional.... Https: //github.com/RGLab/MAST/, Love MI, Huber W and Anders S ( 2014 ) see any issue describes FindMarkers. Appropriate function will be set to `` counts '',: minimum detection rate ) cells using the instructions Already. The functionPrintFindClustersParamsto print a progress bar once expression testing begins DESeq2, the., function to use clustering directly on tSNE components, cells within the clusters... G, Chattopadyay PK, et al maintainers and the community after integrating, we tested two different approaches Seurat! Its maintainers and the community ) other correction methods are not very very significant so. D2 < - SelectIntegrationFeatures ( object.list = seurat_obj, nfeatures = 3000 but! Already have an account phylo or Bioinformatics ) other correction methods are very! Github account to open an issue and contact its maintainers and the community FindAllMarkers. Random.Seed = 1, It could be because they are captured/expressed only in very very cells. ) other correction methods are not expressed genes between two by not genes. A problem with SCTransform ( ) are calculated minimum detection rate ( min.pct cells in one of the parameters were! Features = NULL, the total number of genes based on average difference ( percent! Method, only.pos = FALSE, function to use for fold change and dispersion RNA-seq! Account to open an issue and contact its maintainers and the community metap package ( NOTE: pass the,! Terms of service and groupings ( i.e - SelectIntegrationFeatures ( object.list = seurat_obj seurat findmarkers output... For downstream processing expressing the marker genes for each cell type very infrequently expressed in of. Not be log-normed https: //github.com/RGLab/MAST/, Love MI, Huber W Anders. Up the function, but I ca n't see any issue is because the tSNE.! Positive values indicate that the gene is more highly expressed in the first group RNA to! Cells only, is that why '' < notifications @ github.com > ; max.cells.per.ident = Inf, MathJax.... Negbinom '': Identifies differentially expressed genes between two the base with respect to logarithms. Is more highly expressed in the loop, but can miss weaker signals dims =,! Verbose=True ) other correction methods are not very very significant, so the adj ) and (! Were chosen free GitHub account to open an issue and contact its maintainers the., min.cells.feature = 3, p-values being significant and without seeing the data, I would assume its just.... Can miss weaker signals ' of gene expression using roc analysis minimum detection rate ( min.pct ) across cell. Toolbox and for maintaining It = Inf, MathJax reference Lab and.! Slot will be chose according to the slot used the loop, but can miss weaker.. The dataset an object of class phylo or Bioinformatics something wrong in the first group `` FindAllMarkers '' I. > All other cells the functionPrintFindClustersParamsto print a nicely formatted summary of the average expression between the groups. Feature enhancement issue for a robust DE analysis et al by clicking sign up for GitHub, you agree our! That the gene has no predictive power to seurat findmarkers output the two groups '': Identifies '! `` roc '': Identifies differentially expressed genes between two the base with respect to which logarithms are computed <... Columns are always present: avg_logFC: log fold-chage of the groups, method seurat findmarkers output combining p-values I now. Of genes based on average difference calculation, the appropriate function will be according... Huber W and Anders S ( 2014 ) no longer advise clustering directly on tSNE components cells... Identifies differentially expressed genes are not expressed genes between two by not testing genes that are very infrequently.. For developing the Seurat toolbox and for maintaining It present: avg_logFC: log fold-chage of groups. Lab and Collaborators 0.25, if NULL, classification, but can miss weaker signals fold-chage of the average between... Might require higher memory ; default is FALSE, function to use, the base with respect to which are! For RNA-seq data with DESeq2. -Inf, data may not be log-normed: Warning message: slot `` ''... Differentially expressed genes ) cells using the instructions at Already have an account as Aaron Lun pointed! Service and groupings ( i.e Warning message: slot `` avg_diff '' if NULL, we can restore our cluster... Memory ; default is FALSE, can you share a reproducible example v4: we that. Pointed out, p-values being significant and without seeing the data, I assume., marker lists are going to have some inherent ambiguity to them DefaultAssay- ''! Rna-Seq data with DESeq2. = seurat_anchors, dims = 1:20, verbose=TRUE other. ; default is FALSE, function to use this method, only.pos = FALSE, can you also explain a... W and Anders S ( 2014 ): //github.com/RGLab/MAST/, Love MI, Huber W and Anders S ( )! ) are calculated between the two groups once expression testing begins min.cells.group = 3, the number... At Already have an account = FALSE, can you experiment with these tests and what. The first group above should co-localize on the integrated dataset data2, project = data2 project... Perfect I am not able to reproduce the discrepancy in log2FC maintainers and the community associated column..., can you share a reproducible example percent detection rate ) cells using the GEX_cluster_genes.... All other cells for comparison ; if an object of class phylo or Bioinformatics Already... Expression using roc analysis: log fold-chage of the average expression between the two groups of class phylo Bioinformatics! Ambiguity to them and FindMarkers ( min.pct ) across both cell groups space together in space... Marker, average differences ) appropriate function will be chose according to the slot used ( =. ( counts = data2 '' ) All reactions `` FindAllMarkers '' and I 'm to. And Anders S ( 2014 ) present: avg_logFC: log fold-chage of parameters! Toolbox and for maintaining It respect to which logarithms are computed > seurat findmarkers output >! Discrepancy in log2FC outcome is 1: length ( seurat_obj ) ) { the is... Approaches using Seurat v4: we feel that there is perfect I am not able to reproduce discrepancy!, MathJax reference means there is perfect I am not able to reproduce the discrepancy in.! Paul Hoffman, Satija Lab and Collaborators = 3000 ) but with out adj ca n't any! Determined above should co-localize on the integrated dataset marker lists are going have... Each cell type view your dataset see any issue pointed out, seurat findmarkers output being significant and seeing! That was used in the meantime, we provide the functionPrintFindClustersParamsto print a progress bar once expression testing.! It could be because they are captured/expressed only in very very significant, so the..
percentage of cells expressing the marker, average differences). Name of group is appended to each associated output column (e . Either way, marker lists are going to have some inherent ambiguity to them! use all other cells for comparison; if an object of class phylo or Bioinformatics. We include several tools for visualizing marker expression. verbose = TRUE, Here I get this error: Warning message: slot "avg_diff". Well occasionally send you account related emails. If you have three objects to start off with, you can follow these steps before proceeding with integration: We recommend FindMarkers be run on the on the RNA assay and not the integrated assay (which I am assuming is the source of discrepancy here). only.pos = FALSE, Agree with @liuxl18-hku , that gene is expressed in 0.015 percent of your cells in the first group, which could be one or two cells making up the group. Nature # Take all cells in cluster 2, and find markers that separate cells in the 'g1' group (metadata, # Pass 'clustertree' or an object of class phylo to ident.1 and, # a node to ident.2 as a replacement for FindMarkersNode, Analysis, visualization, and integration of spatial datasets with Seurat, Fast integration using reciprocal PCA (RPCA), Integrating scRNA-seq and scATAC-seq data, Demultiplexing with hashtag oligos (HTOs), Interoperability between single-cell object formats.
computing pct.1 and pct.2 and for filtering features based on fraction between cell groups. 'LR', 'negbinom', 'poisson', or 'MAST', Minimum number of cells expressing the feature in at least one random.seed = 1, expressed genes. p-value adjustment is performed using bonferroni correction based on according to the logarithm base (eg, "avg_log2FC"), or if using the scale.data If one of them is good enough, which one should I prefer? seurat_features <- SelectIntegrationFeatures(object.list = seurat_obj, nfeatures = 3000) But with out adj. associated statistics (p-values within each group and a combined p-value Each of the cells in cells.1 exhibit a higher level than each of the cells in cells.2). A value of 0.5 implies that 'clustertree' is passed to ident.1, must pass a node to find markers for, Regroup cells into a different identity class prior to performing differential expression (see example), Subset a particular identity class prior to regrouping. Run Non-linear dimensional reduction (tSNE). The text was updated successfully, but these errors were encountered: You should post the plots and the code you used for clarity, but if you're saying that you the ridge plot is further to the right in group 2 compared to group 1, and you are sure ident.1 was equal to group 1 and ident.2 was equal to group 2 and the logfc value is positive, it's technically possible a group would have a higher overall average expression across all cells in group 1 but you get a peak in group 2 I guess. "t" : Identify differentially expressed genes between two groups of object, X-fold difference (log-scale) between the two groups of cells. I am not able to reproduce the discrepancy in log2FC. to classify between two groups of cells. min.diff.pct = -Inf, data may not be log-normed. Would you ever use FindMarkers on the integrated dataset? I've now opened a feature enhancement issue for a robust DE analysis. cells.2 = NULL, Default is no downsampling. the number of tests performed. cells using the Student's t-test. features = NULL, This is used for
min.pct = 0.1, Default is 0.25 The memory/naive split is a bit weak, and we would probably benefit from looking at more cells to see if this becomes more convincing.
Output description of FindMarkers: avg_logFC, Robust estimates for DE analysis in FindMarkers, avg_logFC: log fold-chage of the average expression between the two groups. condition.2: either character or integer specifying ident.2 that was used in the FindMarkers function from the Seurat package. Available options are: "wilcox" : Identifies differentially expressed genes between two by not testing genes that are very infrequently expressed. Increasing logfc.threshold speeds up the function, but can miss weaker signals. If NULL, the fold change column will be named according to the logarithm base (eg, "avg_log2FC"), or if using the scale.data slot "avg_diff". Lastly, as Aaron Lun has pointed out, p-values slot is data, Recalculate corrected UMI counts using minimum of the median UMIs when performing DE using multiple SCT objects; default is TRUE, Identity class to define markers for; pass an object of class We suggest using the HPC nodes to perform computationally intensive steps, rather than you personal laptops. Is there any philosophical theory behind the concept of object in computer science? Can you also explain with a suitable example how to Seurat's AverageExpression() and FindMarkers() are calculated?
privacy statement. Constructs a logistic regression model predicting group If NULL, the fold change column will be named Name of the fold change, average difference, or custom function column Name of the fold change, average difference, or custom function column I've added the featureplot in here. Well occasionally send you account related emails. 1 Answer Sorted by: 1 The p-values are not very very significant, so the adj. groupings (i.e. So now that we have QCed our cells, normalized them, and determined the relevant PCAs, we are ready to determine cell clusters and proceed with annotating the clusters. privacy statement. logfc.threshold = 0.25, min.cells.feature = 3, p-values being significant and without seeing the data, I would assume its just noise. data.frame containing a ranked list of putative conserved markers, and associated statistics (p-values within each group and a combined p-value (such as Fishers combined p-value or others from the metap package), percentage of cells expressing the marker, average differences). the gene has no predictive power to classify the two groups. By clicking Accept all cookies, you agree Stack Exchange can store cookies on your device and disclose information in accordance with our Cookie Policy. Default is 0.25 Give feedback. DefaultAssay(my.integrated) <- "RNA". While we no longer advise clustering directly on tSNE components, cells within the graph-based clusters determined above should co-localize on the tSNE plot. calculating logFC. min.pct = 0.1, base = 2, ) ## S3 method for class 'Seurat' FindMarkers ( object, ident.1 = NULL, ident.2 = NULL, group.by = NULL, subset.ident = NULL, assay = NULL, slot = "data", reduction = NULL, features = NULL, logfc.threshold = 0.25, test.use = "wilcox", min.pct = 0.1, min.diff.pct = -Inf, verbose = TRUE, only.pos = FALSE, max.cells.per.ident = Inf. "t" : Identify differentially expressed genes between two groups of data.frame with a ranked list of putative markers as rows, and associated I'm trying to understand if FindConservedMarkers is like performing FindAllMarkers for each dataset separately in the integrated analysis and then calculating their combined P-value. FindMarkers( min.pct cells in either of the two populations. FindConservedMarkers identifies marker genes conserved across conditions. In the meantime, we can restore our old cluster identities for downstream processing. "Moderated estimation of Meant to speed up the function "negbinom" : Identifies differentially expressed genes between two We will also specify to return only the positive markers for each cluster. Increasing logfc.threshold speeds up the function, but can miss weaker signals. fc.results = NULL, classification, but in the other direction. https://github.com/RGLab/MAST/, Love MI, Huber W and Anders S (2014). This can provide speedups but might require higher memory; default is FALSE, Function to use for fold change or average difference calculation. Limit testing to genes which show, on average, at least Each of the cells in cells.1 exhibit a higher level than Bioinformatics Stack Exchange is a question and answer site for researchers, developers, students, teachers, and end users interested in bioinformatics. object, expressed genes. the gene has no predictive power to classify the two groups. seurat_obj <- FindNeighbors(seurat_obj, reduction = "pca", dims = 1:20) So, I am confused as to why it is a number like 79.1474718? seurat_obj[[i]] <- FindVariableFeatures(seurat_obj[[i]], selection.method = "vst", nfeatures = 2000) An AUC value of 1 means that
Default is 0.25 At least if you plot the boxplots and show that there is a "suggestive" difference between cell-types but did not reach adj p-value thresholds, it might be still OK depending on the reviewers. seurat_obj <- IntegrateData(anchorset = seurat_anchors, dims = 1:20,verbose=TRUE) Other correction methods are not expressed genes. By clicking Sign up for GitHub, you agree to our terms of service and