Genes in gtf or gff format

From genomewiki
Revision as of 22:40, 15 June 2017 by Chmalee (talk | contribs) (→‎Using kent commands with the public database server: adding note about matching gene_id and transcript_id when using knownGene table as input)
Jump to navigationJump to search

Convert genePred to GTF with the genePredToGtf command line utility

GenePred is a table format commonly used for gene prediction tracks in the UCSC Genome Browser. The genePredToGtf command-line utility can be used to convert genePred to GTF.

While the Table Browser does contain an option to output query results in GTF, the output is limited, and in some cases, may contain bugs. The best method to convert genePred to GTF is the genePredToGtf command-line utility. The operating-specific utility can be downloaded from the utilities directory.

Once downloaded (and permissions changed to executable), you can run the utility without arguments to see the usage statement:

$ genePredToGtf
genePredToGtf - Convert genePred table or file to gtf.
usage:
   genePredToGtf database genePredTable output.gtf
If database is 'file' then track is interpreted as a file
rather than a table in database.
options:
   -utr - Add 5UTR and 3UTR features
   -honorCdsStat - use cdsStartStat/cdsEndStat when defining start/end
    codon records
   -source=src set source name to use
   -addComments - Add comments before each set of transcript records.
    allows for easier visual inspection
Note: use a refFlat table or extended genePred table or file to include
the gene_name attribute in the output.  This will not work with a refFlat
table dump file. If you are using a genePred file that starts with a numeric
bin column, drop it using the UNIX cut command:
    cut -f 2- in.gp | genePredToGtf file stdin out.gp

Using Table Browser output as input for the genePredToGtf

You can use Table Browser output as input for the genePredToGtf utility, but you will need to check that the Table Browser output is indeed in the correct GenPred format. In some cases, you may have trailing columns that need to be removed.

For example,

  1. From the UCSC Genome Browser, click on "Genome Browser" at the top menu bar, then select "Reset All User Settings" to refresh to the default hg38 assembly and its default position.
  2. Go to the Table Browser, and keeping all options as default, change only 1 setting: region should be set to "position" instead of genome.
  3. Accept the default drop-down option for "output format" as "all fields from selected table" and
  4. Type in a name for "output file" to download your file (e.g., "knownGeneABO.txt").
  5. Click "get output."

Note that you will have 12 columns; and you will need to remove the last two columns to get genePred format:

cat knownGeneABO.txt | cut -f1-10 > knownGeneABO.genePred

Now convert to GTF, using the "file" argument for genePredToGTF:

genePredToGtf file knownGeneABO.genePred knownGeneABO.gtf

Use genePredToGtf with a downloaded genePred table

You can directly download a table (for example, the knownGene table), which will be in genePred format. You can then use this local file as input for the genePredToGtf conversion.

ftp://hgdownload.cse.ucsc.edu/goldenPath/hg19/database/knownGene.txt.gz

The SQL structure:

ftp://hgdownload.cse.ucsc.edu/goldenPath/hg19/database/knownGene.sql

As noted in the usage message, the file can be used with the command in place of the database table specification. In this case, beware of files that are only partially genePred format. For example, the knownGene.txt.gz file has extra columns after the exonEnds column. Therefore, use cut to extract just the columns for genePred:

$ zcat knownGene.txt.gz | cut -f1-10 | genePredToGtf file stdin knownGene.gtf

Example with downloaded refGene.txt.gz

Here are detailed steps for converting a local hg19 refGene table (in genePred format) to GTF.

1. Download your gene set of interest for hg19. For this example, I'll use the refGene table, but you can choose other gene sets, such as the knownGene table from the "UCSC Genes" track.

rsync -a -P rsync://hgdownload.soe.ucsc.edu/goldenPath/hg19/database/refGene.txt.gz ./

2. Unzip

gzip -d refGene.txt.gz

3. Remove the first "bin" column:

cut -f 2- refGene.txt > refGene.input

4. Convert to gtf:

genePredToGtf file refGene.input hg19refGene.gtf

5. Sort output by chromosome and coordinate

cat hg19refGene.gtf  | sort -k1,1 -k4,4 > hg19refGene.gtf.sorted

Example output for hg19refGene.gtf.sorted:

$head hg19refGene.gtf.sorted
chr1 refGene.input exon 10002682 10002840 . - . gene_id "LZIC"; transcript_id "NM_001316973"; exon_number "7"; exon_id "NM_001316973.7"; gene_name "LZIC";
chr1 refGene.input exon 10002682 10002840 . - . gene_id "LZIC"; transcript_id "NM_001316975"; exon_number "7"; exon_id "NM_001316975.7"; gene_name "LZIC";
chr1 refGene.input exon 10002682 10002840 . - . gene_id "LZIC"; transcript_id "NM_001316976"; exon_number "5"; exon_id "NM_001316976.5"; gene_name "LZIC";
chr1 refGene.input exon 10002682 10002840 . - . gene_id "LZIC"; transcript_id "NM_032368"; exon_number "7"; exon_id "NM_032368.7"; gene_name "LZIC";
chr1 refGene.input CDS 10002739 10002793 . - 0 gene_id "LZIC"; transcript_id "NM_001316974"; exon_number "7"; exon_id "NM_001316974.7"; gene_name "LZIC";
chr1 refGene.input exon 10002739 10002840 . - . gene_id "LZIC"; transcript_id "NM_001316974"; exon_number "7"; exon_id "NM_001316974.7"; gene_name "LZIC";
chr1 refGene.input start_codon 10002791 10002793 . - 0 gene_id "LZIC"; transcript_id "NM_001316974"; exon_number "7"; exon_id "NM_001316974.7"; gene_name "LZIC";
chr1 refGene.input exon 10002981 10003083 . + . gene_id "NMNAT1"; transcript_id "NM_001297778"; exon_number "1"; exon_id "NM_001297778.1"; gene_name "NMNAT1";
chr1 refGene.input transcript 10002981 10045556 . + . gene_id "NMNAT1"; transcript_id "NM_001297778";  gene_name "NMNAT1";
chr1 refGene.input exon 10003307 10003485 . - . gene_id "LZIC"; transcript_id "NM_032368"; exon_number "8"; exon_id "NM_032368.8"; gene_name "LZIC";

Using kent commands with the public database server

To use the kent commands with the public database server, add this four line file ".hg.conf" to your home directory:

$ cat $HOME/.hg.conf
db.host=genome-mysql.cse.ucsc.edu
db.user=genomep
db.password=password
central.db=hgcentral

And set the permissions:

$ chmod 600 .hg.conf

Now you can use the command to extract GTF files directly from the UCSC database. For example, fetch the UCSC gene track from hg19 into the local file knownGene.gtf:

$ genePredToGtf hg19 knownGene knownGene.gtf

Note that this still produces incorrect GTF, where the gene_id and transcript_id match. You will either need to write a custom script to fix these fields, or use the "file" option described above on a custom knownGene table download to get correct GTF.

Bed format gene tracks (convert bed > genePred > GTF)

Some gene tracks are in a bed format in the database, perhaps with extra columns past the standard bed format. In this case, extract the standard bed columns, convert it to a genePred and then to a gtf. For example wgRna:

mysql --user=genome --host=genome-mysql.cse.ucsc.edu -A -N \
  -e "select chrom,chromStart,chromEnd,name,score,strand,thickStart,thickEnd from wgRna;" hg19 \
   | bedToGenePred stdin stdout | genePredToGtf file wgRna.gtf

Note that in the above methods, it was necessary to cut columns 1 - 10 to remove the extra trailing columns. With the method detailed here, this cut is not necessary in the case of using the database table since the command can determine from the table structure which columns to use.