BISuilfite-seq CUI Toolkit
Clone or download
Latest commit 7215bc8 Jul 17, 2018
Failed to load latest commit information.
lib fix to mac compilcation memmem May 15, 2018

BISCUIT Travis-CI Build Status DOI

BISulfite-seq CUI Toolkit (BISCUIT) is a utility suite for analyzing sodium bisulfite conversion-based DNA methylation/modification data. It was written to perform alignment, DNA methylation and mutation calling, allele specific methylation from bisulfite sequencing data.


Source code

All releases are available here. Note after v0.2.0, if you choose to use Git, make sure use git clone --recursive to get the submodules.

You can also download with Curl using

$ curl -O $(curl -s | grep browser_download_url | grep | cut -d '"' -f 4)

The precompiled

$ # macOS
$ curl -O $(curl -s | grep browser_download_url | grep macOS | cut -d '"' -f 4)
$ # Linux
$ curl -O $(curl -s | grep browser_download_url | grep x86_64 | cut -d '"' -f 4)


If you download source code, BISCUIT can be compiled through,

$ unzip
$ cd biscuit-release
$ make

The created biscuit binary is the main entry point.


Index reference for alignment

biscuit index GRCh38.fa

The index of BISCUIT composed of the 2-bit packed reference (.bis.pac, .bis.amb, .bis.ann). The suffix array and FM-index of the parent strand (.par.bwt and and the daughter strand (.dau.bwt and

Read alignment

The following snippet shows how BISCUIT can be used in conjunction with samtools to produce indexed alignment BAM file.

$ biscuit align -t 10 GRCh38.fa fastq1.fq.gz fastq2.fq.gz | samtools sort -T . -O bam -o output.bam
$ samtools index output.bam
$ samtools flagstat output.bam >output.bam.flagstat

See here for more information.

Visualize alignment

The tview subroutine colors the alignments in bisulfite mode. Here is a screenshot.

$ biscuit tview -g chr19:7525080 input.bam ref.fa

Unlike samtools, in this subroutine, a reference fasta file is mandatory so that bisulfite conversion can be identified.

Mark duplicate reads

This step is optional. The mark duplicate of BISCUIT is bisulfite strand aware.

$ biscuit markdup input.bam output.bam


Like samtools, BISCUIT extract DNA methylation as well as genetic information. The following shows how to produce a tabix-indexed VCF file.

$ biscuit pileup GRCh38.fa input.bam -o output.vcf -q 20
$ bgzip output.vcf
$ tabix -p vcf output.vcf.gz

Make bed files

The following extract CpG beta values from the VCF file.

$ biscuit vcf2bed -k 10 -t cg input.vcf.gz

-t can also take

  • snp - SNP information
  • c - all cytosines
  • hcg - HCG for NOMe-seq
  • gch - GCH for NOMe-seq

-e output sequence contexts.

QC your run

A bash script is provided to simplify QC procedure.

$ ./scripts/ -v input.vcf setup_file input.bam

generates QC information that can be picked up by MultiQC. Setup files for hg19, hg38 and mm10 are provided. Other genome builds can be built following the same format.

EPI-reads and allele-specific methylation

Following illustrates how to produce epiread which carries the information of epi-haplotype.

$ biscuit epiread -r GRCh38.fa -i input.bam -B snp.bed

To test all SNP-CpG pair,

$ biscuit epiread -r GRCh38.fa -P -i input.bam -B snp.bed

Details can be found here.

sort -k1,1 -k2,2n -k3,3n in.epiread >out.epiread
biscuit asm out.epiread >out.asm

Validate bisulfite conversion label

Sometimes, the bisulfite conversion labels in a given alignment are inaccurate, conflicting or ambiguous. The bsstrand command summarizes these labels given the number of C>T, G>A substitutions. It can correct inaccurate labels as an option.

$ biscuit bsstrand -g "chr1:1000000-1050000" GRCh37.fa input.bam 

gives something like

Mapped reads: 16688
Unmapped reads: 29
Corrected reads: 0 (0.00%)

Confusion counts:
orig\infer     BSW (f)      BSC (r)      Conflict (c) Unknown (u)
     BSW (f):   8426         42           4            12
     BSC (r):   15           8167         3            19
Conflict (c):   0            0            0            0
 Unknown (u):   0            0            0            0

The inferred YD tag gives the following

  • f: foward/Waston strand
  • r: reverse/Crick strand
  • c: conflicting strand information
  • u: unintelligible strand source (unknown)

YD is inferred based on the count of C>T (nCT) and G>A (nGA) observations in each read according to the following rule: If both nCT and nGA are zero, YD = u. s = min(nG2A,nC2T)/max(nG2A,nC2T). if nC2T > nG2A && (nG2A == 0 || s <= 0.5), then YD = f. if nC2T < nG2A && (nC2T == 0 || s <= 0.5), then YD = r. All other scenarios gives YD = c. -y append nCT(YC tag) and nGA(YG tag) in the output bam.

Summarize and filter reads by bisulfite conversion

$ biscuit bsconv -g "chr1:1000000-1050000" GRCh37.fa input.bam

For some library preparation, incomplete conversion are enriched in a subset of reads that needs to be filtered. This command transforms bam into one that contains file ZN tag e.g., ZN:Z:CA_R0C11,CC_R1C14,CG_R0C2,CT_R1C5. This tag summarizes counts of retention and conversion for four different cytosine contexts CpA, CpC, CpG and CpT. It contains a minimum threshold of CpA, CpC, CpT or CpH in general. The -b option outputs the summary in tables instead of as tags in the BAM file.


  • lib/aln was adapted from Heng Li's BWA-mem code.
  • lib/htslib was subtree-ed from the htslib library.
  • lib/klib was subtree-ed from Heng Li's klib.