Skip to content

PRISM_deconvolution_analysis.md

Latest commit

 

History

History
83 lines (64 loc) · 3.2 KB

PRISM_deconvolution_analysis.md

File metadata and controls

83 lines (64 loc) · 3.2 KB

PRISM cell-type deconvolution analysis

PRISM infers epigenetically heterogeneous subclones from bisulfite sequencing data saved as BAM file. According to authors, this tool has been verified only with RRBS data. Detailed usage is described in https://github.com/dohlee/prism

Run PRISM

Prism firstly extracts epiloci where a group of reads are mapped in a short genomic region. Default settings for read coverage 20 (d=20) and minimum number of CpGs 4 (c = 4) are given in prism extract command. bulk_1.bam and bulk_1.met file names are given input and ouput files each.

prism extract -i bulk_1.bam -o bulk_1.met -u -v

We corrected errorneous methylation pattern in produced bulk_1.met file through prism preprocess command which is based on hidden Markov model. The result is saved in bulk_1_corrected.met file.

prism preprocess -i bulk_1.met -o bulk_1_corrected.met

prism deconvolute command infers the subclonal composition of the sample. Simply give methylation pattern-corrected epiloci file.

Finally, we conducted deconvolution on bulk_1_corrected.met file using prism deconvolute command. Maximum number of cluster is set as 10 (-m option) and bulk_1.result is the final result file.

prism deconvolute -i bulk_1_corrected.met -o bulk_1.result -m 10

Cacluate cell-type proprtion from the PRSIM result

Since PRISM only infers subclone of each epilocus, we calculated proportions of each subclone as the fianl cell-type composition.

This is bulk_1.result file readed in R as a data frame:

head(dt_res)
##                                                      epilocus cluster subclone
## 1 chr1:240426862;chr1:240426875;chr1:240426912;chr1:240426944       1        1
## 2     chr2:79940406;chr2:79940428;chr2:79940464;chr2:79940486       1        1
## 3     chr7:71041799;chr7:71041804;chr7:71041834;chr7:71041841       1        1
## 4 chr1:242570514;chr1:242570540;chr1:242570545;chr1:242570574       1        1
## 5 chr15:53716758;chr15:53716776;chr15:53716788;chr15:53716811       1        1
## 6     chr8:70602381;chr8:70602390;chr8:70602399;chr8:70602419       1        1
##   depths fingerprint_counts fingerprint_fractions\n    
## 1     19                  9                   0.4736842
## 2     21                  2                   0.0952381
## 3     20                  7                   0.3500000
## 4     18                  8                   0.4444444
## 5     21                  7                   0.3333333
## 6     23                 11                   0.4782609

We calculated ratio of each subclone and assigned the ratio as a proportion of each cell type as below:

n_subclones = length(unique(dt_res$subclone))
composition <- unlist(lapply(unique(dt_res$subclone), function(i){nrow(dt_res[dt_res$subclone == i,])/nrow(dt_res)}))

if(sum(composition) == 0){
  composition <- unlist(lapply(-1:0, function(i){nrow(dt_res[dt_res$subclone == i,])/nrow(dt_res)}))
}

names(composition) <- lapply(1:n_subclones, function(x){paste0("ctype_", as.character(x))})
print(composition)
##   ctype_1   ctype_2 
## 0.8438215 0.1561785