Integrative RNA-seq and methylation analysis

Thank you for your questions. Please find below the detailed methods as requested.

Estimating methylation levels

Methylation levels were calculated as percentages as cited in our methods from Sabbagh et al. In Sabbagh et al,. they used “call_methylated_sites from Methylpy package (https://github.com/yupenghe/methylpy/)” to determine methylation level at individual cytosines. 

Pearson correlations in figure 2 were calculated with the cor function in R and run as below:

Fig2E. cor((vector of wild type mCH values in gene body), (vector of wild type mCG values in gene body), method = c("pearson"))

Fig2F. cor(((vector of wild type mCH values in gene body) - (vector of DKO mCH values in gene body)), ((vector of wild type mCG values in gene body) - (vector of DKO mCH values in gene body)), method = c("pearson"))

Integrative RNA-seq and methylation analysis

In Figure 4C, methylation versus gene expression plots were made using running average binning on significantly differentially expressed genes (DEGs, padj <0.01). Below are the steps taken to generate these plots in more detail.

1. Methylation levels were determined as described above
2. Calculate change in % methylation as ((% mCH in cKO) - (% mCH in WT)) or as ((% mCG in cKO) - (% mCG in WT)). Here ‘cKO’ is either Dnmt3a or MeCP2 conditional knockout (cKO) and ‘WT’ is wildtype mice with endogenous levels of either protein.
3. Order genes based on ascending methylation value. 
4. Bin so that each bin contains the same number of genes, with 80 percent gene overlap between consecutive bins. 
   1. For the plots of genes significantly misregulated only in the Dnmt3a cKO model, each bin has 25 genes and the window moves by five genes per bin. This results in 126 bins.
   2. For plots of genes significantly misregulated only in the Mecp2 cKO model, the bin size is 10 genes and the window slides by two each time. This results in 62 bins. 
   3. For plots of genes significantly misregulated in both models, the bin size is 10 genes and the window slides by two each time. This results in 42 bins.
5. Plot one point per bin. 
6. After binning, a univariate linear model was fit to the data with the lm function in R, and the R² (percentage variance in log2 fold change explained by methylation) was calculated.
7. Compare these R² values to R² values from 1000 random gene repetitions, with each repetition containing the same number of non-differentially expressed genes (padj >0.01).
8. Plot with gray points represent each iteration of this process, with the original DEG (padj <0.01) values for R²shown in larger orange, green, or blue points. 
9. Compute P value as (r+1)/(n+1) as cited in our methods from North et al.; r is number of repetitions where R² is greater than that in the DEGs, and n is total number of repetitions.

Category