GLF

GLF is a format for storing marginal likelihoods for next-generation sequence data, conditional on a set of possible genotypes.

Generating GLF Files

GLF files can be generated using samtools. To generate a GLF file, use the samtools pileup -g command, which requires a sorted SAM file and a FASTA file with the human genome reference sequence.

  samtools pileup -g -f human_b36_male.fa.gz NA19240.SLX.maq.bam > NA19240.SLX.maq.glf

Generating GLF Files for a Specific Region

Sometimes, you want to generate a GLF file for a specific chromosome region. This can be accomplished by first using samtools view to extract reads for a specific region from a SAM or BAM file and then using samtools pileup -g to generate the GLF file using the selected reads as input. Here is an example:

  samtools view -u NA19240.SLX.maq.bam chr20:10000000-20000000 | samtools pileup -g -f human_b36_male.fa.gz - > NA19240.SLX.chr20_region.glf

File Format

The GLF-file format is defined in an Appendix to the SAM-file format specification.

The current specification (GLF version 3) follows. All integers in are stored in the little-endian byte order. Most GLF files are compressed in a GZIP compatible format; SAMTOOLS will only read GLF files that are compressed with the BGZF library.

Header

Each GLF file starts with a header that identifies the files.

 char[4]              magicNumber = "GLF\3";         // This identifies the file format
 int32_t              headerLength;                  // This is typically zero
 char[headerLength]   headerText;                    // This is typically unused

Chromosome Header

The main header is followed by a series of blocks, summarizing likelihoods along each chromosome. Each of these blocks starts with a header, that records the chromosome label and length.

  int32_t             labelLength;                   // Including the terminating null character
  char[labelLength]   label;                         // Printable string identified the chromosome label; typically, "1", "2", "3"... "22", "X", "Y", "MT" are used as labels for human chromosomes.
  uint32_t            chromosomeLength;              // Length of the original reference sequence. This is a useful sanity check, but samtools can generate GLF entries that go past the end of the sequence.

Likelihood Record Header

Each chromosome header is followed by a series of likelihood records, terminating with an end record of type 0.

   char               refBase:4;                     // 0..15 => XACMGRSVTWYHKDBN, so that 0x01 = A, 0x02 = C, 0x04 = G, 0x08 = T
   char               recordType:4;                  // 0 for the last record in this chromosome, 1 for regular records, 2 for indels

Simple Likelihood Record

These are records with recordType = 1.

   uint32_t           offset;                       // Offset from the previous record
   uint32_t:24        depth;                        // Depth of coverage for the current record 
   uint32_t:8         maxLLK;                       // Maximum log-likelihood, multiplied by -10 log 10
   uint8              mappingQuality;               // Root mean squared mapping quality
   uint8_t            llk[10];                      // Log-likelihood for each genotype, in the order AA..AT..CC..CT..GG..TT 

Indel Likelihood Record

These are records with recordType = 2.

   uint32_t           offset;                       // Offset from the previous record
   uint32_t:24        depth;                        // Depth of coverage for the current record 
   uint32_t:8         maxLLK;                       // Maximum log-likelihood, multiplied by -10 log 10
   uint8              mappingQuality;               // Root mean squared mapping quality
   uint8_t            llkHomozygous11;              // Log-likelihood for an allele 1 homozygote
   uint8_t            llkHomozygous22;              // Log-likelihood for an allele 2 homozygote
   uint8_t            llkHomozygous12;              // Log-likelihood for an allele 1/2 heterozygote
   int16_t            signedAllele1length;          // Length of the first indel allele (positive=ins; negative=del; zero=no-indel)
   int16_t            signedAllele2length;          // Length of the first indel allele (positive=ins; negative=del; zero=no-indel) 
   char               indelSequence1[signedAllele1Length];      // Sequence of the first indel allele
   char               indelSequence2[signedAllele2Length];      // Sequence of the first indel allele

Last Record

Records with recordType = 0 are empty.

Tools That Use GLF Files

Variant Callers

glfSingle

glfTrio

glfMultiples

Utilities

• glfMerge - Combines GLF multiple glfFiles generated for the same individual. Useful for combining data across platforms.