From: Estimating genome-wide DNA methylation heterogeneity with methylation patterns
Method | Formula | Approach | Applicable to non-CG sites | Consideration of pattern similarity | Linearity of the score | Independent of methylation level1 | Genome-wide screening | |
---|---|---|---|---|---|---|---|---|
Model-based (MeH) | ||||||||
Abundance based | \({\left(\sum_{i=1}^{R}{a}_{i}^{2}\right)}^{-1}\in \left\{1,\dots ,{2}^{w}\right\}\) | \(a:\) methylation patterns | Counting distinct methylation patterns | ● | ● | ● | ● | |
Pairwise-similarity based | \({\left(\sum_{i=1}^{R}\sum_{j=1}^{R}{{d}_{ij}p}_{ij}^{2}\right)}^{-1/2}>0\) | \(p:\) methylation patterns | Considering pairwise similarity between patterns | ● | ● | ● | ● | ● |
Phylogenetic-tree based | \({\left(\sum_{i=1}^{B}{L}_{i}{a}_{i}^{2}\right)}^{-1}>0\).5 | \(a:\) methylation patterns | Considering the total similarity among all patterns | ● | ● | ● | ● | ● |
Other methods | ||||||||
\(\frac{\sum_{c=1}^{C}{\omega }_{c}}{\sum_{c=1}^{C}{\omega }_{c}+ \sum_{m=1}^{M}{\omega }_{m}+ \sum_{u=1}^{U}{\omega }_{u}}\in [\mathrm{0,1})\) | \(\omega :\) reads covering CG sites | Measuring the methylation concurrence between patterns | ● | |||||
\(\frac{{\sum }_{r\in {R}_{c}}I(\exists i,j \epsilon r \mathrm{s}.\mathrm{t}. {x}_{j,r}\ne {x}_{i,r})}{|{R}_{c}|}\in [\mathrm{0,1}]\) | \(r:\) reads covering CG sites | Counting distinct methylation patterns among reads | ● | ● | ||||
\(\frac{1}{w}\sum_{k}-{a}_{k}{log}_{2}{a}_{k}\in [\mathrm{0,1}]\) | \(a:\) methylation patterns | Measuring the chaos among the reads of different methylation patterns | ● | ● | ||||
\(1-\sum_{k}{a}_{k}^{2}\in [\mathrm{0,1})\) | \(a:\) methylation patterns | Estimating the probability of observing two different patterns at random | ● | ● | ||||
Fraction of Discordant Read Pairs (FDRP) [18] | \(\frac{{\sum }_{{r}_{s}\in {R}_{c}}{\sum }_{{r}_{t}\in {R}_{c},t>s}I(\exists i\in \left\{{r}_{s}\cap {r}_{t}\right\} \mathrm{s}.\mathrm{t}. {x}_{i,{r}_{s}}\ne {x}_{i,{r}_{t}})}{(\genfrac{}{}{0pt}{}{|{R}_{c}|}{2})}\in [\mathrm{0,1}]\) | \(r:\) reads covering CG sites | Calculating pairwise disagreement of between any two reads | ● | ● | |||
Quantitative FDRP (qFDRP)[18] | \(\frac{{\sum }_{{r}_{s}\in {R}_{c}}{\sum }_{{r}_{t}\in {R}_{c},t>s}\frac{{\sum }_{i\in {\{r}_{s}\cap {r}_{t}\}}I({x}_{i,{r}_{s}}\ne {x}_{i,{r}_{t}})}{|\{{r}_{s}\cap {r}_{t}\}|}}{(\genfrac{}{}{0pt}{}{|{R}_{c}|}{2})}\in [\mathrm{0,1}]\) | \(r:\) reads covering CG sites | Quantifying the similarity of paired-methyl reads by Hamming distance | ● | ● | |||
\(\frac{{\sum }_{l=0}^{L}(l+1)\frac{{\sum }_{r\in {R}_{c}}{\sum }_{i=1}^{\left|r\right|-l}I({x}_{i,r}=1\wedge \dots \wedge {x}_{i+l.r}=1)}{{\sum }_{r\in {R}_{c}}\left|r\right|-l}}{{\sum }_{l=0}^{L}l+1}\in [\mathrm{0,1}]\) | \(r:\) reads covering CG sites | Estimating the fraction of strings that are fully methylated for all possible lengths | ● | ● |