Skip to content

Tutorial

Jump to bottom Edit New page
mheuer-nmdp edited this page Nov 17, 2014 · 51 revisions

This tutorial assumes you have an Amazon Web Services account registered with the NMDP. This will grant you access to a public machine image with all the data, tools, and compute infrastructure you need to proceed. If you do not have these things, go here first.

Get the code

If you have a GitHub account with a public key and would like to contribute your local changes, we encourage you to do so but ask that you kindly follow our instructions, otherwise you may clone and use the pipeline anonymously:

$ git clone https://github.com/nmdp-bioinformatics/pipeline.git

Will create a local clone (working copy) of the GitHub repository, which contains several shell scripts for parallel execution of pipeline components.

Get the sample data

Public sample data from the sequence read archive are provided here:

/opt/data/DRP000941/

Each file (73 total) contains phased NGS data for 6-locus HLA published by Hosomichi et al, 2013. The files must be decompressed from SRA format to FASTQ before processing. SRA provides tools for this purpose. The decompressed data are also provided in the fastq/ directory.

Run the pipeline

Be Careful, please do not process all of the SRA data on the NMDP-provided instance, use the following command to process a single sample and then scale up as you wish using your own Amazon account or other high-performance compute environment.

From within your cloned pipeline directory:

$ ./tutorial/prep_data_dir.bash ~/workdir
$ ./splitter.bash ~/workdir

Processing the data will take several minutes, depending on your hardware instance and available resources. In the meantime, you may proceed using preprocessed results.

Interpret and validate the results

Clinical interpretation of HLA DNA sequence for transplantation is typically confined to the antigen recognition sites (ARS), which correspond to exons 2 and 3 or exon 2 of class I and class II HLA genes, respectively. The NMDP's interpretation service currently requires consensus sequences that are trimmed of other structural elements (non-ARS exons, introns, promoters and other untranslated regions). We will use HLA-A results for a single homozygous sample DRR003809 (DKB) for illustrative purposes. The first step is to filter contigs by region, in this case exons 2 and 3 of HLA-A:

$ ngs-filter-consensus \
  -i /opt/data/DRP000941/results/DRR003809_1.fastq.contigs.bwa.sorted.bam \
  -x hla-a.txt \
  -g HLA-A \
  -m -c -b 0.5 > DKB.fasta

The parameters are as follows:

   -i, --input-bam-file [class java.io.File]  input BAM file [required]
   -x, --input-genomic-range-file [class java.io.File]  input file of genomic ranges, space-delimited exon chrom:start-end [required]
   -o, --output-file [class java.io.File]  output FASTA file, default stdout [optional]
   -g, --gene [class java.lang.String]  gene name, written to the FASTA headers [required]
   -c, --cdna  output cDNA from the same contig (phased consensus sequence) in FASTA format [optional]
   -r, --remove-gaps  remove alignment gaps in the filtered consensus sequence [optional]
   -b, --minimum-breadth-of-coverage [class java.lang.Double]  filter contigs less than minimum, default 0.5 [optional]
   -p, --expected-ploidy [class java.lang.Integer]  filter contigs more than expected ploidy, default 2 [optional]

You may validate that the consensus sequences were correctly filtered by uploading to the UCSC genome browser BLAT tool:



Finally, use the NMDP's interpretation service to assign nomenclature on the command-line. A simple web-interface is also available (not shown):

curl -T DKB.fasta http://interp.b12x.org/hla/api/rs/interp | awk '{split($0,a,"|"); print a[1]}'

The awk command limits output to the first assigned allele, resulting in:

  % Total    % Received % Xferd  Average Speed   Time    Time     Time  Current
                                 Dload  Upload   Total   Spent    Left  Speed
100 26011    0 25455  100   556   9232    201  0:00:02  0:00:02 --:--:--  9232
HLA-A*24:02:01:01+HLA-A*24:02:01:01

Consistent with the known typing for cell line DKB.

Create an HML message

With this tutorial completed you have some of the basic elements to construct an HML message. The following will demonstrate some additional tools and process that should facilitate creation of fully-compliant HML documents appropriate for submitting genetic data to the NMDP.

Create a template HML from the XSD

There are a few ways you can create a new HML message. Since HML is just XML and is based off of an XML schema, you can generate HML directly from the HML schema. For the Hackathon, the public schema is located at http://schemas.nmdp.org/ and is currently 0.9.5 (beta). You can point a schema-aware text editor at this location and generate HML. Some examples of tools that do this are XMLSpy, jEdit, and most development IDEs like Eclipse. For XMLSpy, here is the tutorial for how to generate an XML message from the HML schema: http://manual.altova.com/xmlspy/spyenterprise/index.html?generatesamplexmlfile.htm.

Use the HMLGenerator tool

Auto-generating a message from a schema has a couple drawbacks. Many times there are conditional or 'choice' elements in a schema where either one OR the other element can be generated, but not both. Some generators just choose the first option which may not match your data needs. Another drawback is that optional or abstract attributes and elements may or may not be generated depending on your text editor. To get around these limitations, there is a command-line tool included with the open-source NGS toolset built for the Hackathon. Based on simple command-line input, a sample HML structure will be created that matches your business needs.

ngs-tools command line tools

$ ngs-hml-generator

Enter HML version [0.9.5] > 0.9.5

Enter NMDP reporting center code like '567' > 678

Do you have typing data that refers to a typing test list? [Y] > y

Enter the 3-digit NMDP center-code to use for this sample. > 999

=== Select GENE-FAMILY ===
 (1) HLA
 (2) KIR
---------- 
1

...

Using this tool, you can create a reference HML message for any combination of typing methods and interpretations.

Create <sequence> tags

The following Groovy code snippet uses BioJava's FASTA reader to validate the consensus sequence generated above and reformat it in HML.

import java.io.BufferedReader

import java.io.File

import org.biojava.bio.BioException;

import org.biojava.bio.seq.io.SeqIOTools
import org.biojava.bio.seq.SequenceIterator

File file = new File("DKB.fasta")
BufferedReader reader = new BufferedReader(new FileReader(file))

for (SequenceIterator sequences = SeqIOTools.readFastaDNA(reader); sequences.hasNext(); ) {
  try {
    println "<sequence alphabet=\"DNA\">${sequences.nextSequence().seqString()}<\\sequence>"
  }
  catch(BioException error) {
    println "invalid sequence: ${error}"
  }
}

For other file formats, such as VCF, the feature parser has corresponding methods to validate DNA fields.

Create <targeted-region> tags

The following Groovy code snippet uses the feature parser to reformat HLA-A clinical exons into proper HML targeted-region tags.

import org.nmdp.ngs.feature.Locus
import org.nmdp.ngs.feature.parser.FeatureParser

def filename = "/opt/nmdp/regions/clinical-exons/hla-a.txt"

new File(filename).each { line ->
  def (index, coordinate) = line.split("\t")
  def locus = FeatureParser.parseLocus(coordinate)
  
  println  "<targeted-region \
             \n assembly=\"GRCh38\" \
             \n contig=\"${locus.getContig()}\" \
             \n start=\"${locus.getMin()}\" \
             \n end=\"${locus.getMax()}\" \
             \n strand=\"1\" \
             \n id=\"file://${filename}\" \
             \n description=\"HLA-A exon ${index}\"/>"
}

Create <glstring> tags

This is as simple as placing the interpreted allele between tags such as:

<glstring>HLA-A*24:02:01:01+HLA-A*24:02:01:01</glstring>

Alternatively (and preferably), the GL String may be registered using the GlClient java class from gl-client module

GlClient client = new JsonGlClient(...);
String identifier = client.registerGenotypeList("HLA-A*24:02:01:01+HLA-A*24:02:01:01");

gl-tools command line tools

$ echo "HLA-A*24:02:01:01+HLA-A*24:02:01:01" |\
  gl-register-genotype-lists -s https://gl.immunogenomics.org/imgt-hla/3.16.0/

http://gl.immunogenomics.org/imgt-hla/3.16.0/genotype-list/2

or directly via HTTP using curl

$ curl --header "content-type: text/plain" \
       --data "HLA-A*24:02:01:01+HLA-A*24:02:01:01" \
       -X POST https://gl.immunogenomics.org/imgt-hla/3.16.0/genotype-list \
       -v 2>&1 | grep Location \

< Location: http://gl.immunogenomics.org/imgt-hla/3.16.0/genotype-list/2

Then include the returned identifier URI in the glstring element

<glstring uri="http://gl.immunogenomics.org/imgt-hla/3.16.0/genotype-list/2"/>

DaSH

Clone this wiki locally