How to filter Ensembl cDNA and ncRNA FASTA files by primary assembly?

Question

I'm currently performing differential expression analysis using alignment-free quantification using Kallisto. To do this, I need to create a Kallisto index using Ensembl's cDNA and ncRNA annotations available at the following two links:

ftp://ftp.ensembl.org/pub/current_fasta/mus_musculus/cdna/Mus_musculus.GRCm38.cdna.all.fa.gz ftp://ftp.ensembl.org/pub/current_fasta/mus_musculus/ncrna/Mus_musculus.GRCm38.ncrna.fa.gz

However, later on in my analysis I realized that I had multiple Ensembl gene IDs pointing to the same gene. For example, I was gettign results for both ENSMUSG00000026012 and ENSMUSG00000102412 which both map to Cd28. I realized that the reason for this error was due to the fact that the annotations I was using contained haplotypes.

I would like to remove these haplotypes, and to do this I believe that I need to rerun my analysis using a primary assembly. However, Ensembl doesn't have annotated cDNA and ncRNA FASTA files of the primary assembly. How can I filter these files to contain only genes present in the primary assembly? Is there a better way for me to solve this issue?

Answer 1

The simplest method is to download the GTF file for GRCm38 and filter that. You can then use one of the many tools out there (bedtools getfasta (make sure to enforce strandedness and possibly use the --split option, if you happen to use a BED12 file), gtf_to_fasta from tophat2 (it's the only reason to keep tophat2 installed), etc.) to make a transcriptomic fasta file.

Alternatively, Ensembl BioMart can be used to directly download a transcriptomic fasta file using a filter to select only those transcripts on pre-defined chromosomes. Other filters can be applied as well. If you don't already have a GTF file downloaded and one of the previously mentioned packages installed, then this may be the simplest way to get what you want.