Quality Controls
1 May 2021
Generation of Quality Controls
SeSAMe provides a set of quality control steps. Be sure to pre-cache HM450 annotation data from ExperimentHub. This only needs to be done once per sesame installation.
## [1] TRUE## [1] TRUEThe SeSAMe QC function returns an sesameQC object which can be directly printed onto the screen.
##
## =======================
## = Intensities =
## =======================
## No. probes (num_probes_all) 485577
## mean (M/U) (mean_intensity): 5529.506
## mean (M+U) (mean_intensity_total): 11059.01
##
## -- Infinium II --
## No. probes: (num_probes_II) 350076 (72.095%)
## Mean Intensity (mean_ii): 5160.813
##
## -- Infinium I (Red) --
## No. probes: (num_probes_IR) 89203 (18.371%)
## No. Probes Consistent Channel: 88799
## No. Porbes Swapped Channel: 162
## No. Probes Low Intensity: 242
## Mean Intensity (in-band): 6527.3
## Mean Intensity (out-of-band): 928.2117
##
## -- Infinium I (Grn) --
## No. probes: 46298 (9.535%)
## No. Probes Consistent Channel: 46000
## No. Probes Swapped Channel: 254
## No. Probes Low Intensity: 44
## Mean Intensity (in-band): 6394.865
## Mean Intensity (out-of-band): 640.0676
##
## =======================
## = Non-detection =
## =======================
## No. probes: 79778
## No. probes w/ NA (num_na): 79778 (16.430%)
## No. nondetection (num_nondt): 18174 (3.743%)
##
## =======================
## = Beta Values =
## =======================
## Mean Betas: 0.5114288
## Median Betas: 0.6279925
##
## -- cg probes --
## No. Probes: 482421
## No. Probes with NA: 78429 (16.257%)
## Mean Betas: 0.513384
## Median Betas: 0.6343521
## % Unmethylated (Beta < 0.3): 41.272%
## % Methylated (Beta > 0.7): 47.488%
##
## -- ch probes --
## No. Probes: 3091
## No. Probes with NA: 1346 (43.546%)
## Mean Betas: 0.05914102
## Median Betas: 0.05224748
## % Unmethylated (Beta < 0.3): 100.000%
## % Methylated (Beta > 0.7): 0.000%
##
## -- rs probes --
## No. Probes: 65
## No. Probes with NA: 3 (4.615%)
## Mean Betas: 0.5006825
## Median Betas: 0.51025
## % Unmethylated (Beta < 0.3): 32.258%
## % Methylated (Beta > 0.7): 32.258%
##
## =======================
## = Inferences =
## =======================
## Sex: MALE
## Ethnicity: WHITE
## Age: 61.43636
## Bisulfite Conversion (GCT): 1.10858The sesameQC object can be coerced into data.frame and linked using the following code
qc10 <- do.call(rbind, lapply(ssets, function(x)
as.data.frame(sesameQC(x))))
qc10$sample_name <- names(ssets)
qc10[,c('mean_beta_cg','frac_meth_cg','frac_unmeth_cg','sex','age')]Background
The background level is given by mean_oob_grn and mean_oob_red
library(ggplot2)
ggplot(qc10,
aes(x = mean_oob_grn, y= mean_oob_red, label = sample_name)) +
geom_point() + geom_text(hjust = -0.1, vjust = 0.1) +
geom_abline(intercept = 0, slope = 1, linetype = 'dotted') +
xlab('Green Background') + ylab('Red Background') +
xlim(c(500,1200)) + ylim(c(500,1200))
Mean Intensity
The mean {M,U} intensity can be reached by mean_intensity. Similarly, the mean M+U intensity can be reached by mean_intensity_total. Low intensities are symptomatic of low input or poor hybridization.
library(wheatmap)
p1 <- ggplot(qc10) +
geom_bar(aes(sample_name, mean_intensity), stat='identity') +
xlab('Sample Name') + ylab('Mean Intensity') +
ylim(0,18000) +
theme(axis.text.x = element_text(angle=90, vjust=0.5, hjust=1))
p2 <- ggplot(qc10) +
geom_bar(aes(sample_name, mean_intensity_total), stat='identity') +
xlab('Sample Name') + ylab('Mean M+U Intensity') +
ylim(0,18000) +
theme(axis.text.x = element_text(angle=90, vjust=0.5, hjust=1))
WGG(p1) + WGG(p2, RightOf())
Fraction of color channel switch
The fraction of color channel switch can be found in InfI_switch_G2R and InfI_switch_R2G. These numbers are symptomatic of how Infinium I probes are affected by SNP-induced color channel switching.
Fraction of NA
The fraction of NAs are signs of masking due to variety of reasons including failed detection, high background, putative low quality probes etc. This number can be reached in frac_na_cg and num_na_cg (the cg stands for CpG probes, so we also have num_na_ch and num_na_rs)
p1 <- ggplot(qc10) +
geom_bar(aes(sample_name, num_na_cg), stat='identity') +
xlab('Sample Name') + ylab('Number of NAs') +
theme(axis.text.x = element_text(angle=90, vjust=0.5, hjust=1))
p2 <- ggplot(qc10) +
geom_bar(aes(sample_name, frac_na_cg), stat='identity') +
xlab('Sample Name') + ylab('Fraction of NAs (%)') +
theme(axis.text.x = element_text(angle=90, vjust=0.5, hjust=1))
WGG(p1) + WGG(p2, RightOf())
Quality Ranking
Sesame provide convenient function to compare your sample with public data sets processed with the same pipeline. All you need is a raw SigSet.
## n_sample_compared rank_probe_success_rate rank_mean_intensity
## 636.0000000 0.0000000 0.1572327Output explicit and Infinium-I-derived SNP to VCF
## Retrieving SNP annotation from http://zhouserver.research.chop.edu//InfiniumAnnotation/20200704/EPIC/EPIC.hg19.snp_overlap_b151.rds ... Done.## Retrieving SNP annotation from http://zhouserver.research.chop.edu//InfiniumAnnotation/20200704/EPIC/EPIC.hg19.typeI_overlap_b151.rds ... Done.One can output to actual VCF file with a header by formatVCF(sset, vcf=path_to_vcf).