10x Genomics Support/Cell Ranger/Analysis/

Building a Custom Reference for 5' Single Cell Immune Profiling

Choose a product below to filter the page content to your needs:

Cell Ranger offers downloadable prebuilt human and mouse reference packages, designed for use with its pipelines. These 10x Genomics reference packages are based on the T cell receptor (TRA, TRB) and B cell immunoglobulin (IGH, IGL, IGK) gene annotations in Ensembl version 94 for the human and mouse references. They also include multiple corrections to various V, D, J, and C genes based on empirical observations and to correct clear errors such as frameshifts, leader peptide truncations, and nucleotides that are never observed in rearrangements. These changes are documented in the release notes of each version of Cell Ranger. See Prebuilt References for details on how these references were created.

If you would like to use your own genome FASTA or gene GTF annotations, Cell Ranger supports the use of customer-generated Ensembl-based references. Cell Ranger also includes support for generating a V(D)J reference from the IMGT database.

There are two ways to generate a V(D)J reference:

  • Making a Genome-based Reference Package (e.g., using Ensembl) * Making a V(D)J Segment-based Reference Package (e.g., using IMGT)

The cellranger mkvdjref tool can be used to generate a custom reference package from a genome sequence FASTA file and a gene annotation GTF.

$ cellranger mkvdjref --genome=my_vdj_ref \ --fasta=GRCh38_ensembl.fasta \ --genes=GRCh38_ensembl.gtf

A Cell Ranger V(D)J reference consists of germline gene segment sequences. It assumes that these sequences are contained within a genome reference FASTA, and that an Ensembl-formatted gene annotation GTF points to the relevant gene segments.

The cellranger mkvdjref tool expects a FASTA file (supplied by the --fasta argument) containing genomic reference sequences whose names are consistent with the names used in the GTF file.

The cellranger vdj pipeline expects a GTF file (supplied by the --genes argument) in an Ensembl-like format that contains information about V(D)J gene segments.

GTF columns

ColumnNameDescription
1ChromosomeMust refer to a chromosome/contig in the genome fasta.
2SourceUnused.
3FeatureCell Ranger vdj only uses rows where this line is equal to one of CDS or five_prime_utr.
4StartStart position on the reference (1-based inclusive).
5EndEnd position on the reference (1-based inclusive).
6ScoreUnused.
7StrandStrandedness of this feature on the reference: + or -.
8FrameUnused.
9AttributesA semicolon-delimited list of key-value pairs of the form key "value". The attribute keys used by Cell Ranger V(D)J are detailed below.

GTF attributes

GTF AttributeDescription
transcript_idBecomes the record_id in the Cell Ranger V(D)J reference entry format.
transcript_biotypeThe value is used to infer the V(D)J segment type. Either transcript_biotype or gene_biotype must be a value in the "Accepted Biotypes" list below. If transcript_biotype is not on the accepted list, then gene_biotype is used.
gene_biotypeSee transcript_biotype.
gene_nameMust be specified. Becomes the gene_name in the Cell Ranger V(D)J reference entry format.

Accepted biotypes

  • TR_C_gene
  • TR_D_gene
  • TR_J_gene
  • TR_V_gene
  • IG_C_gene
  • IG_D_gene
  • IG_J_gene
  • IG_V_gene

Example minimal GTF row used by cellranger vdj

14 havana CDS 21621904 21621946 . + 0 transcript_id "ENST00000542354"; gene_name "TRAV1-1"; transcript_biotype "TR_V_gene";

A successful cellranger mkvdjref run creates a directory whose name is specified by the --genome argument.

$ tree my_vdj_ref my_vdj_ref ├── fasta │ └── regions.fa └── reference.json

The regions.fa is the V(D)J segment FASTA file. The header (first) line contains V(D)J-specific metadata. The feature_id and display_name are feature-specific identifier fields (see table). All other fields describe the V(D)J feature.

>feature_id|display_name record_id|gene_name|region_type|chain_type|chain|isotype|allele_name SEQUENCE
FieldDescription
feature_idUnique integer ID for this feature.
display_nameName used when displaying the segment in, e.g., Loupe V(D)J Browser.
record_idDescribes the accession ID of the source molecule. Unused.
gene_nameThe name of the V(D)J gene, e.g. TRBV2-1.
region_typeSupported values are 5'UTR, L-REGION+V-REGION, D-REGION, J-REGION, and C-REGION. Variations are not accepted.
chain_typeSpecifies whether this is a T or B cell receptor chain. Supported values are TR and IG.
chainThe nucleotide sequence of the V(D)J gene specified under gene_name (e.g., if gene_name is TRBV2-1, chain shows TRBV2-1 sequence).
isotypeSpecifies the class of heavy chain constant region; set to None if not applicable.
allele_nameThe identifier for the allele, e.g. 01 for TRBV2-1*01, or None if no allele is to be specified.

Example V(D)J segment FASTA files

>1|TRAV1*01 AF259072|TRAV1|L-REGION+V-REGION|TR|TRA|None|01 ACGTACGTACGTACGTACGTACGTACGTACGTACGTACGTACGTACGTACGTACGTACGT >979|IGHA*01 J00475|IGHA|C-REGION|IG|IGH|A|01 ACGTACGTACGTACGTACGTACGTACGTACGTACGTACGTACGTACGTACGTACGTACGT

You can directly generate a FASTA file in the segment format and create a V(D)J reference by passing this file to cellranger mkvdjref with the --seqs argument, for example:

cellranger mkvdjref --genome=my_vdj_ref \ --seq=custom_segments.fasta

Cell Ranger comes with the fetch-imgt script that downloads the relevant sequences from IMGT and generates a V(D)J segment FASTA file. The V(D)J segment FASTA file can subsequently be used to construct a V(D)J reference package.

The instructions for creating IMGT references have been tested only for human and mouse species. If you are working with other species, it is advisable to validate your IMGT references. To get started, see the Knowledge Base article titled Building a custom reference for V(D)J using IMGT tool.

This example generates a mouse V(D)J reference based on IMGT.

# source the environment of Cell Ranger 7.1.0 for your shell (bash/csh) # (for bash shell) source path/to/cellranger-7.1.0/sourceme.bash # OR (for C shell) source path/to/cellranger-7.1.0/sourceme.csh echo "version checks" python --version # Using a script that comes with Cell Ranger, get data from IMGT and create a FASTA suitable for use by mkvdjref # The option --species is the name of the species for which the data is to be downloaded. # The option --genome provides the prefix used to name the 2 output files. Only the file with the suffix -mkvdjref-input.fasta is used as input to the mkvdjref utility. path/to/cellranger-7.1.0/lib/bin/fetch-imgt --genome vdj_IMGT_mouse --species "Mus musculus" # Build the CR reference. could also include Cell Ranger on your PATH to avoid specifying the full path for cellranger. # The option --genome is a single identifier with no special symbols aside from hyphen or underscore. The reference will be placed in a directory created with that name. # The option --seqs is the mkvdjref-input.fasta file generated by the fetch-imgt command. path/to/cellranger-7.1.0/cellranger mkvdjref --genome=vdj_IMGT_mouse --seqs=vdj_IMGT_mouse-mkvdjref-input.fasta