Difference between revisions of "OSCtable"

From ZENBU documentation wiki
Jump to: navigation, search
(ZENBU interpretation of OSCtable files)
Line 76: Line 76:
  
 
= ZENBU interpretation of OSCtable files =
 
= ZENBU interpretation of OSCtable files =
OSCtable is one of the main interchange formats for ZENBU. It allows all possible mapping of data into the [[DataModel|ZENBU data model]].  
+
OSCtable is one of the main interchange formats for ZENBU. It allows all possible mapping of data into the [[DataModel|ZENBU data model]]. Since the OSCtable specification is highly flexible, it was possible for the ZENBU OSCtable parser to have an extended vocabulary of metadata directives and column name spaces.
 +
 
 +
 
 +
== Metadata ==
 +
While the official OSCtable specification include mandatory metadata elements, the ZENBU OSCtable parser relaxes this requirement. All ## metadata lines are parsed as optional metadata.
 +
The only requirement for a valid OSCtable for ZENBU is a column header line.
 +
 
 +
=== Experiment metadata===
 +
The original OSCtable specification works very well with single experiment data files, but does not provide enough fine control of assigning metadata in a multi-experiment data file. To alleviate this, ZENBU added an additional metadata directive to the OSCtable specification
 +
 
 +
* '''##ExperimentMetadata[key][experiment-name] = value'''
 +
 
  
 
== Column name spaces ==
 
== Column name spaces ==
Line 100: Line 111:
 
** if ''bed:thickEnd'' is  not equal to ''end'' then the region from ''bed:thickEnd'' to ''end'' is interpreted into a SubFeature of category '''3utr'''
 
** if ''bed:thickEnd'' is  not equal to ''end'' then the region from ''bed:thickEnd'' to ''end'' is interpreted into a SubFeature of category '''3utr'''
  
== ZENBU extensions to OSCtable ==
+
==== ignoring columns ====
The OSCtable specification is flexible enough to allow the option for extension. ZENBU has added two additional features to the specification which allows for more flexible data import.
 
 
 
=== ignoring columns ===
 
 
The OSCtable allows for easy ''wrapping'' of any tab-texted file into an OSCtable by simply pre-pending a header with the appropriate column names. But sometimes these original files contain columns which one might not really need. To simplify the process of loading ZENBU added a special column-name
 
The OSCtable allows for easy ''wrapping'' of any tab-texted file into an OSCtable by simply pre-pending a header with the appropriate column names. But sometimes these original files contain columns which one might not really need. To simplify the process of loading ZENBU added a special column-name
 
* ''ignore.xxxx'' -- ignore this column
 
* ''ignore.xxxx'' -- ignore this column
 
If a column is label as such, on loading this column will be stripped from the data file and thrown away. This can simplify the process for bioinformatician and avoid un-needed data-scripting to [http://en.wikipedia.org/wiki/Data_munging ''munge'' data] prior to loading.
 
If a column is label as such, on loading this column will be stripped from the data file and thrown away. This can simplify the process for bioinformatician and avoid un-needed data-scripting to [http://en.wikipedia.org/wiki/Data_munging ''munge'' data] prior to loading.

Revision as of 08:58, 8 May 2012

OSCtable1.1

Basic structure

  • A simple tabdelimited text file.
  • Lines starting with '#' are comments.
  • Lines starting with '##' are attributes or metadata of the table. (See 'Metadata' section below)
  • The first line after the comments/metadata (see below) is a header line, which indicate column names of the table.
  • Column order is flexible
  • The first column should describe a 'key' (unique in many cases, but not necessarily) of the data, and the column name should be 'id'
  • If a cell need to include multiple values, comma(',') is recommended to be used as a separator.
  • All the comment and attribute lines should appear above the header line
  • All the columns should be described in Metadata (See 'Metadata' section below)

Metadata ('##' lines)

The basic structure is '##qualifier = value'.
Required metadata: FileFormat, Date, ProtocolREF, ColumnVariable, ContactName, ContactEmail

'genome_assembly' for parameter value is required, when using

##NameSpace=genomic_coordinate


Required (mandatory) metadata

FileFormat

describes file format of this file.
example:

##FileFormat = OSCtable1.1

Date

describes the date when the data file is generated

##Date = 20090602

ProtocolREF

describes the protocol used to generate the data file.

##ProtocolREF = CAGEmappingv1.0

ColumnVariable

describes the ALL columns used in the data file

##ColumnVariable[start] = this is a start position of the genomic coordinate
##ColumnVariable[end] = this is a stop position of the genomic coordinate
##ColumnVariable[norm.THP10h] = this is TPM normalized value with 10h
##ColumnVariable[entrez_gene_id] = Entrez gene ID, which is assigned to the cluster

ContactName

describes the contact name about the data file.

##ContactName = Hideya Kawaji

ContactEmail

describes the contact address about the data file

##ContactEmai = kawaji@gsc.riken.jp


Optional metadata

InputFile

describes the file(s) used to generate the data file

##InputFile = lane1.fa
##InputFile = lane2.fa

ParameterValue

describes the parameter(s) used to generate the data file in the protocol
the parameter(s) should be consistent with the protocol description

##ParameterValue[alignment_program] = BWA
##ParamterValue[aligment_program_version] = 1.3.5
##ParameterValue[UCSC_gene_tracks] = RefSeq
##ParameterValue[UCSC_gene_tracks] = ENSEMBL transcript

NameSpace

describes the name space for the column names. See below (NameSpace)

##NameSpace=genomic_coordinate
##NameSpace=expression

Column Name Spaces

  • A set of column names (and parameters) to be used for a specific purpose or context.
  • The same column names with the same name space are recognized as the same (equivalent) meaning.
  • Supported name space: genomic_coordinate, expression

genomic_coordinate

  • column names are: chrom, start.0base, start.1base, end, strand
  • parameter value: genome_assembly
    • chrom: chromosome name used in the genome assembly. For example, chr1, chr2, chr3, ... chrM for the UCSC hg18 genome assembly.
    • start.0base: start position (bp) on the chromosome in 0start coordinate system (BED, PSL, BLAT, exonerate, and nexAlign style)
    • start.1base: start position (bp) on the chromosome in 1start coordinate system (conventional coordinate system; adopted in GFF as well)
    • end: end position (bp) on the chromosome
    • strand: strand on the chromosome; optional
    • Note:
      • All of the above columns are not necessarily required. For example, start.0base would not be required if you have start.1base, and strand would not be required if the annotation do not have strand distinction
      • 'genome_assembly' for parameter value is required.

expression

generic expression tags
  • column names are: raw.ZZZ, norm.ZZZ, or exp.YYY.ZZZ
    • raw.ZZZ : for raw value of expression such as raw_counts and raw signal intensities.
    • norm.ZZZ : for normalized value.
    • ZZZ indicates the experiment (cell conditions, RNAs, etc) of the expressions
    • YYY should not include dot (.), and indicates the type of expression.
'mapcount' tag

To specify that alignments may be be mapped on more than one location, you can either use...

ZENBU interpretation of OSCtable files

OSCtable is one of the main interchange formats for ZENBU. It allows all possible mapping of data into the ZENBU data model. Since the OSCtable specification is highly flexible, it was possible for the ZENBU OSCtable parser to have an extended vocabulary of metadata directives and column name spaces.


Metadata

While the official OSCtable specification include mandatory metadata elements, the ZENBU OSCtable parser relaxes this requirement. All ## metadata lines are parsed as optional metadata. The only requirement for a valid OSCtable for ZENBU is a column header line.

Experiment metadata

The original OSCtable specification works very well with single experiment data files, but does not provide enough fine control of assigning metadata in a multi-experiment data file. To alleviate this, ZENBU added an additional metadata directive to the OSCtable specification

  • ##ExperimentMetadata[key][experiment-name] = value


Column name spaces

official OSCtable column names

The official OSCtable specification has very few predefined column names

  • chrom -- chromosome name
  • start.0base -- chromosome start in a 0base coordinate system
  • start.1base -- chromosome start in a 1base coordinate system
  • end -- chromosome end location
  • strand -- chromosome strand
  • ID -- interpreted as the name of the Feature

While ZENBU uses a 1-based-inclusive coordinate space internally, it can automatically handles the conversion between coordinate spaces.

ZENBU additional column namespaces

  • name -- the name of the Feature
  • eedb::name -- same as name
  • score -- is stored in the Feature significance.
  • eedb::score -- same as score
  • bed::blockCount, bed::blockSizes, bed::blockStarts -- are taken from the BED file specification. These three columns work together and are interpreted into SubFeatures on the primary Feature. Each of these SubFeatures are created with a FeatureSource category of block.
  • bed:thickStart and bed::thickEnd -- are taken from the BED file specification.
    • if bed::thickStart is not equal to start then the region from start to bed:thickStart is interpreted into a SubFeature of category 5utr
    • if bed:thickEnd is not equal to end then the region from bed:thickEnd to end is interpreted into a SubFeature of category 3utr

ignoring columns

The OSCtable allows for easy wrapping of any tab-texted file into an OSCtable by simply pre-pending a header with the appropriate column names. But sometimes these original files contain columns which one might not really need. To simplify the process of loading ZENBU added a special column-name

  • ignore.xxxx -- ignore this column

If a column is label as such, on loading this column will be stripped from the data file and thrown away. This can simplify the process for bioinformatician and avoid un-needed data-scripting to munge data prior to loading.