russbainer / gcrisprtools Goto Github PK

Hi,

I know you would prefer me using the dev package, but... would be nice to be able to do directly from here.

library(githubinstall)
R.3.4.4.>githubinstall("gCrisprTools")
Suggestion:

• OscarBrock/gCrisprTools An R/Bioconductor package for the integrated analysis of pooled high-throughput CRISPR/Cas9 or shRNA screening experiments.
  Do you want to install the package (Y/n)?
  Installing "gCrisprTools" package from GitHub, but already installed from unknown repository.
  This package is already installed. Are you sure you want to overwrite it (Y/n)? Y
  Downloading GitHub repo OscarBrock/gCrisprTools@master
  from URL https://api.github.com/repos/OscarBrock/gCrisprTools/zipball/master
  Installation failed: Does not appear to be an R package (no DESCRIPTION)

currently sample keys are checked against objects via ct.checkInput(), which returns a logical. Update to a better-named function that invisibly returns the compliant object, similar to ct.prepareAnnotation().

Hi,

Nice code, very useful. Now here is my bug:

In my case

resultsDF <-
ct.generateResults(
fit,
annotation = ann,
RRAalphaCutoff = 0.1,
permutations = 1000,
scoring = "combined"
)

produces data which fails ct.resultCheck

Error in ct.resultCheck(resultsDF) :
Some of the columns in the supplied result object seem to be of the wrong type. Please supply a summaryDF object generated from ct.generateResults() in the gCrisprTools package.

This fixes the problem - but should it be done within ct.generateResults?
resultsDF$geneID <- as.character(resultsDF$geneID)

Alt / useful:

ct.resultCheck could be more verbose in what column fails the check, rather than needing to look it up in the code here.

Thanks
/Alistair

ct.UpSet does not properly assign labels to the combination sets. Requires code rework.

Hi,

In the below example, I am making a small dataset, and showing the median normalization in the ct.normalizeMedians() code versus the ~correct formula (from limma's https://rdrr.io/bioc/limma/src/R/norm.R ). In the corrected formula, the summary shows that the medians for the three groups are the same post-normalization, whereas the medians after ct.normalizeMedians() are different. Do you think this is correct?

g1 <- data.frame(sample.id="g1",guide.id=1:1000,counts=rnorm(1000,mean=9,sd=1))
g2 <- data.frame(sample.id="g2",guide.id=1:1000,counts=rnorm(1000,mean=10,sd=1))
g3 <- data.frame(sample.id="g3",guide.id=1:1000,counts=rnorm(1000,mean=11,sd=2))
df <- rbind(g1,g2,g3)


counts <- pivot_wider(df,names_from=sample.id,values_from=counts)
counts <- counts[,-1]
lib.size <- colSums(counts)  
y <- t(log2(t(counts + 0.5)/(lib.size + 1) * 1e+06))
  cmed <- apply(y, 2, median, na.rm = TRUE)
  cmed <- cmed - mean(cmed)
  correctedCounts <- 2^t(t(y) - cmed)
correctedCounts1 <- (t(t(correctedCounts) * ((lib.size + 1) / 1e+06)) - 0.5)

normalizeMedianValues <- function(x) 
{
narrays <- NCOL(x)
if(narrays==1) return(x)
cmed <- log(apply(x, 2, median, na.rm=TRUE))
cmed <- exp(cmed - mean(cmed))
t(t(x)/cmed)
}
gs.counts <- normalizeMedianValues(counts)
## compare
summary(counts)
summary(gs.counts)
summary(correctedCounts1)

Thanks,
Rumen

From https://www.bioconductor.org/packages/devel/bioc/vignettes/gCrisprTools/inst/doc/gCrisprTools_Vignette.html

Code for best practise generation of the aln object would be helpful, commenting on what script/tool usually generates the source matrix.

may => must (in documentation)

2.5 Alignment Statistics Users may provide a matrix of alignment statistics to enhance some applications, including QC reporting. These should be provided as a numeric matrix in which rows correspond to targets (reads containing a target cassette), nomatch (reads containing a cassette sequence but not a known target sequence), rejections (reads not containg a cassette sequence), and double_match (reads derived from multiple cassettes). The column names should exactly match the colnames() of the ExpressionSet object. Simple charting functionality is also provided to inspect the alignment rates of each sample.

ct.makeQCReport(es,
trim = 1000,
log2.ratio = 0.05,
sampleKey = sk,
annotation = ann,
# aln = aln,
identifier = 'Crispr_QC_Report',
lib.size = NULL,
outdir = file_path_report
)

Error in ct.makeQCReport(es, trim = 1000, log2.ratio = 0.05, sampleKey = sk, :
argument "aln" is missing, with no default