Format 6 – Query & Universe (Boolean Sparse Representation)

Format 6 encodes two complementary boolean vectors for each CpG:

  • SET bit → CpG is included in the query
  • UNIVERSE bit → CpG belongs to the background universe

This format is essential for:

  • Enrichment testing (query vs universe)
  • Sparse single-cell methylation (accessible universe + binary call)
  • Binary presence/absence with explicit context
  • Efficient representation of cell-specific CpG availability

Format 6 allows YAME to know both the methylated/unmethylated call and whether the CpG is even in scope for that sample.

1. When to Use Format 6

Use Format 6 when:

  • You want to run enrichment tests (yame summary -m mask.cm sample.cx)
  • You want to represent sparse CpG coverage per sample (e.g., scWGBS)
  • You want a contextualized binary representation in which:
    • CpG exists in the cell’s “universe”
    • CpG is methylated/unmethylated inside that universe

Format 6 is the only CX format that explicitly encodes two bits per CpG.

This makes it the natural fit for:

  • Single-cell bisulfite data
  • Sparse pseudo-bulk models
  • Query/background analyses
  • Fine-grained masking workflows (e.g., yame mask -c)

2. Input Requirements

Input must contain two columns, each with 0 or 1:


<QueryBit>    <UniverseBit>

Example:

1    1   # in query, in universe
0    1   # not in query, in universe
1    1   # in query, in universe
0    0   # outside the universe entirely

Meaning:

Query Universe Interpretation
1 1 CpG is part of query set
0 1 CpG is in universe but not part of query
0 0 CpG should be ignored entirely
1 0 Invalid (cannot be in query if not in universe)

YAME enforces this logic automatically when packing/masking.

3. Packing to Format 6

3.1 Packing from a text file with two columns 

yame pack -f6 query_universe.txt > sample.cx

3.2 Using BED inputs (query and universe tracks)

Generate a query mask:

bedtools intersect -a cpg_ref.bed.gz -b query_peaks.bed -sorted -c \
  | cut -f4 > query.txt

Generate a universe mask:

bedtools intersect -a cpg_ref.bed.gz -b regions_accessible.bed -sorted -c \
  | cut -f4 > universe.txt

Combine to two columns:

paste query.txt universe.txt | \
  awk '{print ($1>0?1:0) "\t" ($2>0?1:0)}' \
  | yame pack -f6 - > query.cx

3.3 From Format 0 + Universe mask using yame mask -c (recommended)

If you already have a binary vector (Format 0), and a universe mask:

# binary input: 1 = methylated, 0 = unmethylated
yame mask -c input_binary.cx universe_mask.cx -o contextual.cx

Rules:

  • Universe = mask’s 1s
  • Query = input’s binary values inside universe

This is common in single-cell pipelines.

4. Unpacking Format 6 

yame unpack sample.cx | head

The output has two columns:

<Query>    <Universe>
1          1
0          1
1          1
0          0
...

Matching the original input format.

5. Integration with YAME Commands

Format 6 integrates deeply with the core statistical logic of YAME.

5.1 yame summary

Format 6 is designed for enrichment analysis:

yame summary -m feature.cm sample.cx

Interpretation:

  • N_univ = number of CpGs where UniverseBit = 1
  • N_query = number of CpGs with both UniverseBit = 1 and QueryBit = 1
  • N_mask = size of feature
  • N_overlap = how many CpGs are (in query) AND (in mask)
  • Log2OddsRatio = enrichment of query inside mask
  • Beta = fraction of QueryBit = 1 among universe CpGs within mask
  • Depth = NA (Format 6 has no coverage)

This is the recommended representation for:

  • Comparing different CpG subsets
  • Representing sparse single-cell methylomes
  • Querying binary signatures within genomic features

5.2 yame rowop

Useful operations:

# Binary summation across samples
yame rowop -o binasum multi.cx > pseudobulk.cg

# Convert to binary string
yame rowop -o binstring multi.cx > patterns.txt

Rules:

  • Query bit → methylated/unmethylated value
  • Universe bit → determines coverage; if Universe = 0, skip site entirely

5.3 yame dsample

Downsampling Format 6:

yame dsample -N 10000 sample.cx > sample_10k.cx

Rules:

  • Eligible sites = UniverseBit = 1
  • Randomly keep N of them per sample
  • Non-selected → UniverseBit set to 0; QueryBit cleared

This is extremely useful for benchmarking single-cell sparsity.

5.4 yame rowsub

Format 6 supports all selection methods:

# Subset by mask
yame rowsub -m promoters.cm sample.cx > subset.cx

# Coordinate lists
yame rowsub -R cpg_nocontig.cr -L CpG_sites.txt sample.cx > subset.cx

5.5 yame mask (contextualization mode) 

# Convert binary format 0 → format 6 with a universe mask
yame mask -c binary.cx universe.cx -o out.cx
  • Universe bits taken from mask
  • Query bits taken from binary vector
  • Sites outside universe become NA-like (Universe = 0)

This workflow is central to building high-quality single-cell methylome objects.

6. When NOT to Use Format 6

Use a different format if:

  • You need floating-point methylation values → Format 4
  • You need M/U counts → Format 3
  • You have categorical labels → Format 2
  • You need simple 0/1 binary without context → Format 0

Format 6 is best when context matters (universe vs query).

7. Minimal End-to-End Example 

# 1. Prepare reference CpGs
yame unpack cpg_nocontig.cr | gzip > cpg_ref.bed.gz

# 2. Build a universe mask (e.g., accessible CpGs)
bedtools intersect -a cpg_ref.bed.gz -b ATAC.bed -sorted -c \
  | cut -f4 \
  | yame pack -fb - > universe.cx

# 3. Build a binary methylation call track
bedtools intersect -a cpg_ref.bed.gz -b methylated_calls.bed -sorted -c \
  | cut -f4 \
  | yame pack -fb - > binary.cx

# 4. Contextualize into Format 6
yame mask -c binary.cx universe.cx -o cell.cx

# 5. Run enrichment against ChromHMM
yame summary -m ChromHMM.cm cell.cx > enrich.txt

# 6. Downsample to 50k accessible CpGs
yame dsample -N 50000 cell.cx > cell_50k.cx

Format 6 is the essential representation for sparse binary methylation, especially in single-cell analyses and feature-level enrichment testing. It is compact, expressive, and foundational for high-performance YAME workflows.

```