Note: if you use MAGeCKFlute in published research, please cite: Binbin Wang, Mei Wang, Wubing Zhang. “Integrative analysis of pooled CRISPR genetic screens using MAGeCKFlute.” Nature Protocols (2019), doi: 10.1038/s41596-018-0113-7.

Input data - results from MAGeCK

MAGeCK (Wei Li and Liu. 2014) and MAGeCK-VISPR (Wei Li and Liu. 2015) are developed by our lab previously, to analyze CRISPR/Cas9 screen data in different scenarios(Tim Wang 2014, Hiroko Koike-Yusa (2014), Ophir Shalem1 (2014), Luke A.Gilbert (2014), Silvana Konermann (2015)). Both algorithms use negative binomial models to model the variances of sgRNAs, and use Robust Rank Aggregation (for MAGeCK) or maximum likelihood framework (for MAGeCK-VISPR) for a robust identification of selected genes.

The command mageck mle computes beta scores and the associated statistics for all genes in multiple conditions. The beta score describes how the gene is selected: a positive beta score indicates a positive selection, and a negative beta score indicates a negative selection.

The command mageck test uses Robust Rank Aggregation (RRA) for robust identification of CRISPR-screen hits, and outputs the summary results at both sgRNA and gene level.

Quick start

Here we show the most basic steps for integrative analysis pipeline. MAGeCKFlute package includes the MAGeCK results from one intrinsic dataset, including countsummary, rra.gene_summary, rra.sgrna_summary, and mle.gene_summary. We will work with them in this document.

Downstream analysis pipeline for MAGeCK RRA

All pipeline results are written into local directory “./MAGeCKFlute_PLX/”, and all figures are integrated into file “PLX_Flute.rra_summary.pdf”.

Downstream analysis pipeline for MAGeCK MLE CRISPR screens can be conducted using three different cell populations: the day 0 population, a drug-treated population (treatment) and a control population (mockdrug control, typically treated with a vehicle such as DMSO). The CRISPR screens are required to analyze with MAGeCK MLE, which generates a gene summary file like mle.gene_summary below. FluteMLE is useful in this scenario to perform downstream analysis.

If CRISPR screens don’t include a control cell population, FluteMLE supports users to use Depmap screens as control.

All pipeline results are written into local directory “./MAGeCKFlute_PLX/”, and all figures are integrated into file “PLX_Flute.mle_summary.pdf”.

Section I: Quality control

** Count summary ** MAGeCK Count in MAGeCK/MAGeCK-VISPR generates a count summary file, which summarizes some basic QC scores at raw count level, including map ratio, Gini index, and NegSelQC. Use function ‘data’ to load the dataset, and have a look at the file with a text editor to see how it is formatted.

##                                   File    Label    Reads   Mapped Percentage
## 1 ../data/GSC_0131_Day23_Rep1.fastq.gz day23_r1 62818064 39992777     0.6366
## 2  ../data/GSC_0131_Day0_Rep2.fastq.gz  day0_r2 47289074 31709075     0.6705
## 3  ../data/GSC_0131_Day0_Rep1.fastq.gz  day0_r1 51190401 34729858     0.6784
## 4 ../data/GSC_0131_Day23_Rep2.fastq.gz day23_r2 58686580 37836392     0.6447
##   TotalsgRNAs Zerocounts GiniIndex NegSelQC NegSelQCPval
## 1       64076         57   0.08510        0            1
## 2       64076         17   0.07496        0            1
## 3       64076         14   0.07335        0            1
## 4       64076         51   0.08587        0            1
##   NegSelQCPvalPermutation NegSelQCPvalPermutationFDR NegSelQCGene
## 1                       1                          1            0
## 2                       1                          1            0
## 3                       1                          1            0
## 4                       1                          1            0

## Warning in data.table::melt(countsummary[, c("Label", "Mapped", "Unmapped")], :
## The melt generic in data.table has been passed a data.frame and will attempt
## to redirect to the relevant reshape2 method; please note that reshape2 is
## deprecated, and this redirection is now deprecated as well. To continue using
## melt methods from reshape2 while both libraries are attached, e.g. melt.list,
## you can prepend the namespace like reshape2::melt(countsummary[, c("Label",
## "Mapped", "Unmapped")]). In the next version, this warning will become an error.

Section II: Downstream analysis of MAGeCK RRA

For experiments with two experimental conditions, we recommend MAGeCK-RRA for identification of essential genes from CRISPR/Cas9 knockout screens. Gene summary file in MAGeCK-RRA results summarizes the statistical significance of positive selections and negative selections.

##       id num  neg.score neg.p.value  neg.fdr neg.rank neg.goodsgrna neg.lfc
## 1    NF2   4 4.1770e-12  2.9738e-07 0.000275        1             4 -1.3580
## 2 SRSF10   4 4.4530e-11  2.9738e-07 0.000275        2             4 -1.8544
## 3 EIF2B4   8 2.8994e-10  2.9738e-07 0.000275        3             8 -1.5325
## 4  LAS1L   6 1.4561e-09  2.9738e-07 0.000275        4             6 -2.2402
## 5   RPL3  15 2.3072e-09  2.9738e-07 0.000275        5            12 -1.0663
## 6 ATP6V0   7 3.8195e-09  2.9738e-07 0.000275        6             7 -1.6380
##   pos.score pos.p.value pos.fdr pos.rank pos.goodsgrna pos.lfc
## 1   1.00000     1.00000       1    16645             0 -1.3580
## 2   1.00000     1.00000       1    16647             0 -1.8544
## 3   1.00000     1.00000       1    16646             0 -1.5325
## 4   0.99999     0.99999       1    16570             0 -2.2402
## 5   0.95519     0.99205       1    15359             2 -1.0663
## 6   1.00000     1.00000       1    16644             0 -1.6380

Visualization of negative selections and positive selections

Read the MAGeCK-RRA results

Two type of scores are optional, lfc (logrithm fold change) and rra (Robust Rank Aggregation score).

##       id   Score      FDR
## 1    NF2 -1.3580 0.000275
## 2 SRSF10 -1.8544 0.000275
## 3 EIF2B4 -1.5325 0.000275
## 4  LAS1L -2.2402 0.000275
## 5   RPL3 -1.0663 0.000275
## 6 ATP6V0 -1.6380 0.000275

Compute the similarity between the CRISPR screen with Depmap screens

##          estimate p.value
## A2780   0.5791483       0
## OVCAR8  0.5789564       0
## R262    0.5755967       0
## HS578T  0.5754257       0
## SF295   0.5751215       0
## KYSE180 0.5749847       0

Omit common essential genes from the data

##          estimate       p.value
## TE11    0.2394741 5.392597e-149
## OVCAR8  0.2390556 6.636024e-154
## YH13    0.2386189 2.470833e-153
## HUPT3   0.2380013 3.869874e-147
## NCIH661 0.2379885 1.640237e-152
## LXF289  0.2360406 5.494129e-150

Dot plot

Visualize negative and positive selected genes separately.

## Warning: Removed 6390 rows containing missing values (geom_point).
## Warning: Removed 6390 rows containing missing values (geom_text_repel).

sgRankView - visualize the rank of sgRNAs targeting top selected genes.

Enrichment analysis

For more information about functional enrichment analysis in MAGeCKFlute, please read the [second document] (https://www.bioconductor.org/packages/3.9/bioc/vignettes/MAGeCKFlute/inst/doc/MAGeCKFlute_enrichment.html), in which we introduce all the available options and methods.

## Warning in fgsea(pathways = geneSets, stats = geneList, nperm = nPerm, minSize = minGSSize, : There are ties in the preranked stats (4.58% of the list).
## The order of those tied genes will be arbitrary, which may produce unexpected results.

Section III: Downstream analysis of MAGeCK MLE

** Gene summary ** The gene summary file in MAGeCK-MLE results includes beta scores of all genes in multiple condition samples.

##     Gene      dmso       plx
## 1   FEZ1 -0.045088 -0.036721
## 2    TNN  0.094325  0.065533
## 3  NAT8L  0.026362  0.044979
## 4   OAS2 -0.271210 -0.289010
## 5 OR10H3 -0.098324 -0.365730
## 6  CCL16 -0.309750 -0.148830

Batch effect removal

Is there batch effects? This is a commonly asked question before perform later analysis. In our package, we provide HeatmapView to ensure whether the batch effect exists in data and use BatchRemove to remove easily if same batch samples cluster together.

## Standardizing Data across genes
##            s1       s2       s3       s4
## [1,] 7.327202 7.346844 6.229324 8.287386
## [2,] 7.518553 5.095516 5.563952 6.836320
## [3,] 5.695361 5.893885 5.937569 5.700798
## [4,] 8.174489 6.005973 6.874356 7.088165
## [5,] 5.973574 5.508085 6.000529 5.472104
## [6,] 8.025503 8.111390 8.192283 8.051765

Normalization of beta scores

It is difficult to control all samples with a consistent cell cycle in a CRISPR screen experiment with multi conditions. Besides, beta score among different states with an inconsistent cell cycle is incomparable. So it is necessary to do the normalization when comparing the beta scores in different conditions. Essential genes are those genes that are indispensable for its survival. The effect generated by knocking out these genes in different cell types is consistent. Based on this, we developed the cell cycle normalization method to shorten the gap of the cell cycle in different conditions. Besides, a previous normalization method called loess normalization is available in this package.(Laurent Gautier 2004)

##     Gene        dmso         plx
## 1   FEZ1 -0.05869343 -0.05660750
## 2    TNN  0.12278783  0.10102282
## 3  NAT8L  0.03431681  0.06933767
## 4   OAS2 -0.35304838 -0.44552525
## 5 OR10H3 -0.12799354 -0.56379346
## 6  CCL16 -0.40321793 -0.22942986
##     Gene        dmso         plx
## 1   FEZ1 -0.04161108 -0.04019792
## 2    TNN  0.10078924  0.05906876
## 3  NAT8L  0.03144985  0.03989115
## 4   OAS2 -0.26942198 -0.29079802
## 5 OR10H3 -0.09644456 -0.36760944
## 6  CCL16 -0.30785246 -0.15072754

Distribution of all gene beta scores

After normalization, the distribution of beta scores in different conditions should be similar. We can evaluate the distribution of beta scores using the function ‘DensityView’, and ‘ConsistencyView’.

## Warning in data.table::melt(beta, id = NULL): The melt generic in data.table
## has been passed a data.frame and will attempt to redirect to the relevant
## reshape2 method; please note that reshape2 is deprecated, and this redirection
## is now deprecated as well. To continue using melt methods from reshape2 while
## both libraries are attached, e.g. melt.list, you can prepend the namespace like
## reshape2::melt(beta). In the next version, this warning will become an error.

Nine-square scatter plot - identify treatment-associated genes

Session info

## R version 3.6.2 (2019-12-12)
## Platform: x86_64-pc-linux-gnu (64-bit)
## Running under: Ubuntu 18.04.3 LTS
## 
## Matrix products: default
## BLAS:   /home/biocbuild/bbs-3.10-bioc/R/lib/libRblas.so
## LAPACK: /home/biocbuild/bbs-3.10-bioc/R/lib/libRlapack.so
## 
## locale:
##  [1] LC_CTYPE=en_US.UTF-8       LC_NUMERIC=C              
##  [3] LC_TIME=en_US.UTF-8        LC_COLLATE=C              
##  [5] LC_MONETARY=en_US.UTF-8    LC_MESSAGES=en_US.UTF-8   
##  [7] LC_PAPER=en_US.UTF-8       LC_NAME=C                 
##  [9] LC_ADDRESS=C               LC_TELEPHONE=C            
## [11] LC_MEASUREMENT=en_US.UTF-8 LC_IDENTIFICATION=C       
## 
## attached base packages:
## [1] stats     graphics  grDevices utils     datasets  methods   base     
## 
## other attached packages:
## [1] ggplot2_3.2.1     MAGeCKFlute_1.6.2
## 
## loaded via a namespace (and not attached):
##   [1] fgsea_1.12.0           colorspace_1.4-1       ggsignif_0.6.0        
##   [4] ggridges_0.5.1         qvalue_2.18.0          XVector_0.26.0        
##   [7] ggpubr_0.2.4           farver_2.0.1           urltools_1.7.3        
##  [10] graphlayouts_0.5.0     ggrepel_0.8.1          bit64_0.9-7           
##  [13] AnnotationDbi_1.48.0   xml2_1.2.2             codetools_0.2-16      
##  [16] splines_3.6.2          GOSemSim_2.12.0        knitr_1.26            
##  [19] pathview_1.26.0        polyclip_1.10-0        zeallot_0.1.0         
##  [22] jsonlite_1.6           annotate_1.64.0        GO.db_3.10.0          
##  [25] dbplyr_1.4.2           png_0.1-7              pheatmap_1.0.12       
##  [28] graph_1.64.0           ggforce_0.3.1          BiocManager_1.30.10   
##  [31] compiler_3.6.2         httr_1.4.1             rvcheck_0.1.7         
##  [34] backports_1.1.5        assertthat_0.2.1       Matrix_1.2-18         
##  [37] lazyeval_0.2.2         limma_3.42.0           tweenr_1.0.1          
##  [40] htmltools_0.4.0        prettyunits_1.0.2      tools_3.6.2           
##  [43] igraph_1.2.4.2         gtable_0.3.0           glue_1.3.1            
##  [46] reshape2_1.4.3         DO.db_2.9              dplyr_0.8.3           
##  [49] rappdirs_0.3.1         fastmatch_1.1-0        Rcpp_1.0.3            
##  [52] enrichplot_1.6.1       Biobase_2.46.0         vctrs_0.2.1           
##  [55] Biostrings_2.54.0      nlme_3.1-143           ggraph_2.0.0          
##  [58] xfun_0.11              stringr_1.4.0          lifecycle_0.1.0       
##  [61] clusterProfiler_3.14.2 XML_3.98-1.20          DOSE_3.12.0           
##  [64] europepmc_0.3          zlibbioc_1.32.0        MASS_7.3-51.5         
##  [67] scales_1.1.0           tidygraph_1.1.2        hms_0.5.2             
##  [70] parallel_3.6.2         KEGGgraph_1.46.0       RColorBrewer_1.1-2    
##  [73] yaml_2.2.0             curl_4.3               memoise_1.1.0         
##  [76] gridExtra_2.3          biomaRt_2.42.0         triebeard_0.3.0       
##  [79] stringi_1.4.3          RSQLite_2.1.5          genefilter_1.68.0     
##  [82] S4Vectors_0.24.1       BiocGenerics_0.32.0    BiocParallel_1.20.1   
##  [85] matrixStats_0.55.0     rlang_0.4.2            pkgconfig_2.0.3       
##  [88] bitops_1.0-6           evaluate_0.14          lattice_0.20-38       
##  [91] purrr_0.3.3            labeling_0.3           cowplot_1.0.0         
##  [94] bit_1.1-14             tidyselect_0.2.5       ggsci_2.9             
##  [97] plyr_1.8.5             magrittr_1.5           R6_2.4.1              
## [100] IRanges_2.20.1         DBI_1.1.0              withr_2.1.2           
## [103] mgcv_1.8-31            pillar_1.4.3           survival_3.1-8        
## [106] KEGGREST_1.26.1        RCurl_1.95-4.12        tibble_2.1.3          
## [109] crayon_1.3.4           BiocFileCache_1.10.2   rmarkdown_2.0         
## [112] viridis_0.5.1          progress_1.2.2         grid_3.6.2            
## [115] sva_3.34.0             data.table_1.12.8      blob_1.2.0            
## [118] Rgraphviz_2.30.0       digest_0.6.23          xtable_1.8-4          
## [121] tidyr_1.0.0            gridGraphics_0.4-1     openssl_1.4.1         
## [124] stats4_3.6.2           munsell_0.5.0          viridisLite_0.3.0     
## [127] ggplotify_0.0.4        askpass_1.1

References

Hiroko Koike-Yusa, E-Pien Tan, Yilong Li. 2014. “Genome-wide recessive genetic screening in mammalian cells with a lentiviral CRISPR-guide RNA library.” http://science.sciencemag.org/content/343/6166/80.long.

Laurent Gautier, Benjamin M. Bolstad, Leslie Cope. 2004. “affy—analysis of Affymetrix GeneChip data at the probe level.” https://academic.oup.com/bioinformatics/article/20/3/307/185980.

Luke A.Gilbert, BrittAdamson, Max A.Horlbeck. 2014. “Genome-Scale CRISPR-Mediated Control of Gene Repression and Activation.” https://linkinghub.elsevier.com/retrieve/pii/S0092-8674(14)01178-7.

Ophir Shalem1, *, 2. 2014. “Genome-scale CRISPR-Cas9 knockout screening in human cells.” http://science.sciencemag.org/content/343/6166/84.long.

Silvana Konermann, Alexandro E. Trevino, Mark D. Brigham. 2015. “Genome-scale transcriptional activation by an engineered CRISPR-Cas9 complex.” https://www.nature.com/nature/journal/vnfv/ncurrent/full/nature14136.html.

Tim Wang, David M. Sabatini, Jenny J. Wei1. 2014. “Genetic Screens in Human Cells Using the CRISPR-Cas9 System.” http://science.sciencemag.org/content/343/6166/80.long.

Wei Li, Han Xu, Johannes Köster, and X. Shirley Liu. 2015. “Quality control, modeling, and visualization of CRISPR screens with MAGeCK-VISPR.” https://genomebiology.biomedcentral.com/articles/10.1186/s13059-015-0843-6.

Wei Li, Tengfei Xiao, Han Xu, and X Shirley Liu. 2014. “MAGeCK enables robust identification of essential genes from genome-scale CRISPR/Cas9 knockout screens.” https://genomebiology.biomedcentral.com/articles/10.1186/s13059-014-0554-4.