What process and input data is required for a cellranger reference transcriptome?

Question

I'm analysing single-cell RNA-Seq data using the 10X Genomics cellranger platform. While they provide reference data for Mouse and Humans, other species require a reference "transcriptome" to be generated with the cellranger mkref tool. However, I've found it to be very temperamental giving errors such as:

gffread genome_GFF3_genes.gff -T -o genome_genes.gtf

cellranger mkgtf genome_genes.gtf genome_genes.filtered.gtf \
                   --attribute=gene_biotype:protein_coding \
                   --attribute=gene_biotype:lincRNA \
                   --attribute=gene_biotype:antisense \
                   --attribute=gene_biotype:IG_LV_gene \
                   --attribute=gene_biotype:IG_V_gene \
                   --attribute=gene_biotype:IG_V_pseudogene \
                   --attribute=gene_biotype:IG_D_gene \
                   --attribute=gene_biotype:IG_J_gene \
                   --attribute=gene_biotype:IG_J_pseudogene \
                   --attribute=gene_biotype:IG_C_gene \
                   --attribute=gene_biotype:IG_C_pseudogene \
                   --attribute=gene_biotype:TR_V_gene \
                   --attribute=gene_biotype:TR_V_pseudogene \
                   --attribute=gene_biotype:TR_D_gene \
                   --attribute=gene_biotype:TR_J_gene \
                   --attribute=gene_biotype:TR_J_pseudogene \
                   --attribute=gene_biotype:TR_C_gene

cellranger mkref --genome="reference-name" \
                 --fasta="transcriptome.fasta" \
                 --genes="genome_genes.filtered.gtf"

Fatal INPUT FILE error, no valid exon lines in the GTF file: reference_name/genes/genes.gtf Solution: check the formatting of the GTF file. Most likely cause is the difference in chromosome naming between GTF and FASTA file.

subprocess.CalledProcessError: Command '['STAR', '--runMode', 'genomeGenerate', '--genomeDir', 'reference_name/star', '--runThreadN', '1', '--genomeFastaFiles', 'reference_name/fasta/genome.fa', '--sjdbGTFfile', 'reference_name/genes/genes.gtf', '--limitGenomeGenerateRAM', '17179869184', '--genomeSAsparseD', '1', '--genomeSAindexNbases', '12', '--genomeChrBinNbits', '10']' returned non-zero exit status 104

subprocess.CalledProcessError: Command '['STAR', '--runMode', 'genomeGenerate', '--genomeDir', 'reference_name/star', '--runThreadN', '1', '--genomeFastaFiles', 'reference_name/fasta/genome.fa', '--sjdbGTFfile', 'reference_name/genes/genes.gtf', '--limitGenomeGenerateRAM', '17179869184', '--genomeSAsparseD', '1', '--genomeSAindexNbases', '12', '--genomeChrBinNbits', '10']' returned non-zero exit status -11

What input data is required for the cellranger to construct a reference? What format do the fasta and gtf files need be in? Is it possible to use a "transcriptome" rather than a "genome" as a reference here? For example, I have a transcriptome reference with fasta names >name for each transcript rather than each chromosome.

Answer 1

Your problem is caused by using the transcriptome fasta file rather than the genome fasta file. You've already given it transcriptome information with genome_genes.filtered.gtf, it needs the genomic sequence to go along with that.

As an aside, your GTF filtering command is unlikely to be useful unless you have a Gencode or Ensembl GTF for human. Have a look at the gene_biotypes in your file and decide for yourself what (if anything) needs to be filtered. For example, I made a cellranger reference last night for dm3 (the old fruit fly genome) with:

cellranger mkgtf annotation.gtf genes.gtf \
    --attribute=gene_biotype:protein_coding \
    --attribute=gene_biotype:ncRNA \
    --attribute=gene_biotype:snRNA \
    --attribute=gene_biotype:snoRNA \
    --attribute=gene_biotype:pre_miRNA

That should be customized for relevance. It's quite possible that your GTF doesn't even have gene_biotype attributes, that's specific to Ensembl I think.