Format 0 – Binary Presence/Absence Data

Format 0 stores binary data, one bit per CpG, representing presence/absence or a simple methylation state.

  • Typical extension: .cg
  • Input: a vector of 0/1 values, one per CpG
  • Best for:
    • Binary methylation calls
    • Presence/absence features
    • DMR indicators
    • Peak overlap masks
    • Cell-level binary CpG accessibility

Format 0 is the smallest and fastest CX format: 8 CpGs per byte (~32× compression over text).

1. Input Requirements

Your input must be:

  • Aligned to a CpG reference file (.cr, format 7)
  • Exactly one line per CpG
  • Each line containing 0 or 1

Example input (binary_data.txt):

1
0
1
1
0

Interpretation:

  • 1 → presence, methylated, peak overlap, or “TRUE”
  • 0 → absence, unmethylated, no peak, or “FALSE”

There is no NA value in format 0. Use Format 4 or 6 if you need NA or universe masking.

2. Packing to Format 0

2.1 From a raw 0/1 text vector 

yame pack -fb binary_data.txt > binary_output.cg

-fb selects Format 0 (b).

2.2 From BED features or presence/absence annotations

This is the canonical way to create a .cm/.cg mask from BED:

bedtools sort -i dmr_sites.bed \
  | bedtools intersect -a cpg_ref.bed.gz -b - -sorted -c \
  | cut -f4 \
  | yame pack -fb - > dmr_binary.cg

Explanation:

  • bedtools intersect -c counts how many times each CpG overlaps the BED input
  • cut -f4 extracts that count (0 or >0)
  • yame pack -fb converts the result into a compressed CX file

If a CpG overlaps ≥ 1 region, the value becomes 1; otherwise 0.

3. Unpacking Format 0

To get the 0/1 vector back:

yame unpack binary_output.cg | head

Example output:

1
0
1
1
0

Format 0 does not include sample-level metadata unless packed from multiple samples; YAME will unpack each sample sequentially if multiple samples exist.

4. Integration with Other YAME Commands

Format 0 integrates well with many downstream YAME tools.

4.1 yame summary

Format 0 is ideal for:

  • Feature masks
  • Peak overlaps
  • Region presence indicators

Example:

yame summary -m promoter.cm sample.cg

Outputs for each mask:

  • N_mask
  • N_overlap
  • log2OddsRatio
  • Binary fraction (Beta)
  • Universe counts depending on the query

4.2 yame rowop

Useful operations on binary data:

# Sum across samples (pseudobulk for binary)
yame rowop -o binasum multi_sample.cg > pseudobulk.cg

# Convert to binary string representation
yame rowop -o binstring multi_sample.cg > patterns.txt

binasum → format 3 output, with M = #1 votes and U = #0 votes.

binstring → one binary string per row, e.g.:

01011001
11100011
...

4.3 yame dsample

Downsample a binary file to N “present” sites:

yame dsample -N 10000 -s 1 binary.cg > dsampled.cg

For Format 0:

  • Eligible sites are those with value 1
  • Non-selected sites become 0

4.4 yame rowsub

Subset rows from Format 0 the same as other formats:

# By row IDs
yame rowsub -l row_ids.txt binary.cg > subset.cg

# By coordinate list and reference
yame rowsub -R cpg_nocontig.cr -L CpG_sites.txt binary.cg > subset.cg

# By mask
yame rowsub -m promoter.cm binary.cg > subset.promoters.cg

4.5 yame mask

Format 0 plays especially well with yame mask:

# Mask out positions where mask == 1
yame mask binary.cg lowquality.cm -o masked_binary.cg

If you use -c, Format 0 can be contextualized into Format 6 to define a universe for sparse annotations.

5. When to Use Format 0 vs Other Formats

Use Format 0 when:

  • Your data is intrinsically binary (e.g., peak/no-peak)
  • You want minimal storage footprint
  • You want maximum speed for summarization / enrichment
  • You are constructing feature files (promoters, enhancers, windows)

Consider alternatives if:

  • You need M/U counts → Format 3
  • You need NA handling or fractions → Format 4
  • You need structured query + universe semantics → Format 6

6. Minimal End-to-End Example 

# 1. Prepare CpG reference (only once)
yame unpack cpg_nocontig.cr | gzip > cpg_ref.bed.gz

# 2. Create binary mask from BED peaks
bedtools intersect -a cpg_ref.bed.gz -b H3K27ac.bed -sorted -c \
  | cut -f4 \
  | yame pack -fb - > H3K27ac.cm

# 3. Summarize enrichment of a methylation sample over H3K27ac peaks
yame summary -m H3K27ac.cm sample.cg > enrich.txt

# 4. Subset sample to only CpGs in promoter regions
yame rowsub -m promoters.cm sample.cg > sample.promoters.cg

# 5. Downsample the promoter mask to 5000 sites
yame dsample -N 5000 promoters.cm > promoters_5k.cm