Skip to content

Latest commit

 

History

History
48 lines (35 loc) · 2.26 KB

pat_format.md

File metadata and controls

48 lines (35 loc) · 2.26 KB
Raw

pat.gz file

A bgzipped tab-separated text file storing the read-level methylation information.

Example:

chr1    46       CT    1
chr1    47       CC..TC  1
chr1    47       T       13
chr2    2300000  C       4
chr10   14633440 TTC     1

The first read covers 2 CpG sites: CpG46 (methylated) and CpG47 (unmethylated).

Specification

A pat files contains ≥4 columns:

  • chrom: a string (E.g., chr1, chr2, ..., chrX, chrY, chrM).

  • CpG index: integer in range [1,NR_SITES]. The index of the first site occurring on the current read. The pat file is sorted by this column.
    In hg19, NR_SITES == 28,217,448. Run wgbstools convert -s CpG_i-CpG_j to translate the CpG indexes to genomic loci.

  • methylation pattern: a string representing the methylation pattern of all consecutive CpG sites in the current read. Each site is represented by a single character:

symbol meaning
C Methylated CpG
T Unmethylated CpG
. CpG with an unknown status
  • count: The number of times a read with these exact values (chrom, CpG index and methylation pattern) occurred.
  • Additional columns are optional.

From bam to pat

A pat file can be thought of as a compact methylation-oriented representation of a bam file. Most of the fields are omitted, as well as non-CpG nucleotides. Reads not covering any CpG sites are discarded. In paired-end data, the two mates are merged into a single read in the pat file. When 2+ reads cover the exact same CpG sites and present identical methylation pattern, they are merged to a single line in the pat file, and the count field is increased. The rest of the nucleotides are omitted.

❗Note: The pat file is sorted by CpG index column (sort -k2,2n -k3,3. E.g., chr1 is followed by chr2, not by chr10), which is different from the UCSC order (sort -k1,1 -k2,2n)

compressing and indexing

The pat file is compressed using bgzip and indexed using tabix (with a *.csi file). This compression is compatible with gzip, and with the indexing allows for a fast random access (try wgbstools view).