methyConcerto

Refined data analyses for single-cell bisulfite-based sequencing in DNA methylation researches, including

• Empirical Bayesian Methylation Caller (EBMC): Determinate discrete methyation status for each cytosine
• BS-SNV-Caller: SNV calling with bisuifite-converted sequencing data
• Biologically consistent ASM CpG (BCA CpG) caller: consistent allele-specific methylation (ASM) among a group of samples (cells), rather than occuring in individual cells occasionally

Empirical Bayesian Methylation Caller (EBMC)

Determinate discrete methyation status for each cytosine from C/T read counts. The methylation state of a cytosine is discrete: methylated (1), unmethylated (0), and hemimethyalted (0.5). EBMC gives the predicted state with corresponding posterior probability.

usage

process CG and nonCG sites seperately

zcat ./data/sample.CGmap.gz | cut -f 7,8 | Rscript ./source/EBMC.R --output ./output/sample.prob --summary ./output/sample.methycall --CG TRUE

parameters

param	description
--CG	boolean, TRUE or FALSE, CG site or not
--prior	prior probabilities, defaults: (0.59, 0.4, 0.01) for CG and (0.01, 0.98, 0.01) for nonCG
--error_rate	prior rate of C/T mis-detction, default: 0.01
--Poisson_approx	threshold of depth the Poisson approximation applied, default: 20
--rounds	rounds of iteration, default: 2

input

.CGmap file for a sample/cell:

./data/sample.CGmap.gz, returned by baseeker2/3 or cgmaptools

output

.prob probability file with same #rows of input: two columns

column	discription
1	predicted methylation state with maximum posterior probability
2	posterior probability of the predicted state

0       0.99911
0       0.990017
0       0.996467
0       0.996467
1       0.99932
0.5     0.800168
0.5     0.800168
0       0.999775
0       0.999775
0       0.996467
0       0.990017
1       0.999995
1       0.999995

.methycall summary file

var	description
Rows	number of input rows/sites
methylated(1)	proportion of methylated cytosines
unmethylated(0)	proportion of unmethylated cytosines
epi-heterozygous(0.5)	proportion of hemimethylated cytosines
error rate of measurement (0->1 or 1->0)	error rate of unexpected reads, errors due to libraray construction, sequencing, alignment etc.

Rows:	10000
methylated(1):	0.188000
unmethylated(0):	0.807900
epi-heterozygous(0.5):	0.004100
error rate of measurement (0->1 or 1->0):	0.002497

BS-SNV-Caller

SNV calling with bisuifite-converted sequencing data

usage

zcat ./data/sample.ATCGmap.gz | awk '$6+$7+$8+$9+$11+$12+$13+$14>=10' | Rscript ./source/BS-SNV-Caller.R | gzip > ./output/sample.bssnv.gz

parameters

tune the parameters in the code

# mutation rate
pm = 1/1000/3

# error rate
# in GV oocyte (2n 4c) samples, error rate is set larger
# pe = 1/100/3
pe = 3/100/3

# total mis rate
p = pm + pe

# methylation rate/proportion 
pr.cg = 0.6        # CG context
pr.ncg = 1/100     # non-CG context

input

a .ATCG file, returned by bsseeker2/3 or cgmaptools

output

13 columns, no header line

column	description
1-5	chromosome, reference base, position, CG/CHG/CHH, dinucleoties. They are same with first 5 columns of .ATCG file
6	p-value testing SNV (different from reference) which means the posterior probability of reference base, the smaller, the more likely to be a SNV, homozygous or heterzygous
7-10	estimated allele frequencies of A,T,C, and G respectively
11-12	coverage depths of Watson and Crick strands
13	probability of homozygote

1	C	1550	CHG	CT	9.79052e-01	7.46622e-07	1.05278e-02	9.89471e-01	7.46622e-07	14	0	0.97916
1	T	1551	--	--	9.78959e-01	7.46623e-07	9.89423e-01	1.05750e-02	7.46623e-07	14	0	0.97907
1	G	1552	CHG	CA	9.99995e-01	7.55074e-07	7.55074e-07	7.55074e-07	9.99998e-01	14	0	1.00000
1	C	1553	CHG	CT	9.79052e-01	7.46622e-07	1.05278e-02	9.89471e-01	7.46622e-07	14	0	0.97916
1	T	1554	--	--	2.13408e-24	3.49345e-05	6.55191e-03	3.49345e-05	9.93378e-01	14	0	0.98676
1	G	1555	CHG	CA	9.99995e-01	7.55074e-07	7.55074e-07	7.55074e-07	9.99998e-01	14	0	1.00000
1	C	1556	CHH	CT	9.79052e-01	7.46622e-07	1.05278e-02	9.89471e-01	7.46622e-07	14	0	0.97916
1	T	1557	--	--	9.78959e-01	7.46623e-07	9.89423e-01	1.05750e-02	7.46623e-07	14	0	0.97907
1	C	1558	CHH	CT	9.79052e-01	7.46622e-07	1.05278e-02	9.89471e-01	7.46622e-07	14	0	0.97916
1	T	1559	--	--	9.78959e-01	7.46623e-07	9.89423e-01	1.05750e-02	7.46623e-07	14	0	0.97907

Biologically consistent ASM CpG (BCA CpG) caller

Call BCA allele-specific methylation (ASM) among a group of cells upon the ASM analysis in each single cells. The BCA ASM is ASM consistent among a group of samples (cells). The p-value is evaluated under the null hypothesis that ASM CpGs are distributed randomly in the whole genome and independently among cells.

usage

source ./source/BCA-ASM-Caller.R

input files

SNV file: ./data/snv-10K-merge.simple.gz

a merged file containing detected SNV sites in each samples, four columns of chromosome, reference base, postion, and sample. no header line

1       T       1554    AF1_12
1       C       15023   AF1_12
1       G       15088   AF1_12
1       G       15091   AF1_12
1       T       15103   AF1_12

ASM file: ./data/methpipe-10K-merge.asm.gz

a merged file containing detected ASM CpGs in each cells, 12 columns of 'chr', 'pos', 'strand', 'CpG', 'p_value', 'dp', 'MM', 'MU', 'UM', 'UU', and 'sample'. no header line. The first 11 clomuns are returned by methpipe::allelicmeth

1       23736   +       CpG     1       11      11      0       0       0       AF1_12
1       50772   +       CpG     1       13      0       0       0       13      AF1_12
1       50809   +       CpG     1       13      0       0       0       13      AF1_12
1       69991   +       CpG     1       12      0       0       0       12      AF1_12
1       99975   +       CpG     1       16      16      0       0       0       AF1_12
1       100868  +       CpG     1       16      16      0       0       0       AF1_12

output file

./output/asm.consensus.tsv

variable	description
chr	chromosome
pos	position
count.cov	count of cells with this CpG effectively covered
count.asm	count of cells indicating ASM in this CpG
chi.sq	Chi-square ($\chi^2$) statistic
p.chisq	p-value of $\chi^2$ test, minor method
lambda	lambda ($\lambda$) statistic
p.pois	p-value of Poisson test, major method

chr	pos	count.cov	count.asm	count.snv	chi.sq	p.chisq	lambda	p.pois
16	8868	48	47	2	NA	NA	26.758615249594396	4.508674530477948e-4
16	9276	49	47	4	Inf	0	26.067905464137255	8.393246565554962e-4
16	9284	50	47	0	Inf	0	29.110167311050276	8.200382170259079e-4
16	9507	50	47	1	Inf	0	28.493378989882395	9.142007577857896e-4
8	1661	48	47	0	Inf	0	28.361876688204955	4.803623038606877e-4

Citations

• Yuan X, Chen N, Feng Y, et al. Single-cell multi-omics profiling reveals key regulatory mechanisms that poise germinal vesicle oocytes for maturation in pigs. Cell Mol Life Sci. 2023;80(8):222. Published 2023 Jul 22. doi:10.1007/s00018-023-04873-x