---
title: "Manuscript data availabiliy"
output: rmarkdown::html_vignette
vignette: >
  %\VignetteIndexEntry{Manuscript data availabiliy}
  %\VignetteEngine{knitr::rmarkdown}
  %\VignetteEncoding{UTF-8}
bibliography: refs.bib
link-citations: yes
---

## Datasets

Two datasets were used to produce the figure in the paper:

1. snmC-seq2 scWGBS data [@luo2018a] for single-cell benchmarks.
2. bulk WGBS from a project in our lab

Both datasets were aligned with BISCUIT [@zhou2024]. To produce Bismark BEDgraph
files [@krueger2011], we used a python script to convert the beta and coverage
value columns to percentage methylation, unmethylated reads and methylated read
columns.

```py
import sys
import gzip

bedfile = sys.argv[1]

def bed2cov(line):
    chr, start, end, beta, cov = line.split('\t')
    percent, meth, unmeth = convert(float(beta), int(cov))
    return '\t'.join([chr, start, start, str(percent), str(meth), str(unmeth)])

def convert(beta, cov):
    percent = round(beta * 100)
    meth = round(beta * cov)
    unmeth = cov - meth
    return [percent, meth, unmeth]

with gzip.open(bedfile, 'rt') as bed:
    for line in bed:
        print(bed2cov(line))
```

Using GNU parallel:

```sh
parallel python bed2cov.py <biscuit_path>/{} '>' <bismark_path>/{/.} ::: biscuit_path/*.bed.gz
```

Then using the `iscream.paper` package at
<https://github.com/huishenlab/iscream.paper> we ran the benchmarks and produced
the figures.

## References