From the UCSC genome browser, is there any way to download knownCanonical set from the NCBI Refseq track as it is possible for the UCSC track (see screenshot below)
knownCanonical is not available in the dropdown in table option when we select NCBI Refseq
knownCanonical is a UCSC term so it won't be available for RefSeq. Specifically, what is it that you are looking for from NCBI RefSeq? Are you only interested in the 'Known RefSeqs' (aka RefSeqs with the NM/NR prefix)?
I am trying to run DepthOfCoverage from GATK3 (it's an old - no more supported version) which requires RefSeq file , however, that files contains all transcripts and not just canonical transcript. I was wondering how can I generate that file.
In that case, RefSeq Select is your best option. Note, RefSeq Select is only available for protein-coding loci; so of the ~54k unique GeneIDs annotated currently, 19k are protein-coding and have a RefSeq Select. Are you interested in getting these data in GFF3 format? If so, you can either filter the latest RefSeq GFF3 or download GFF3 for just the RefSeq Select transcripts from the NCBI Nucleotide portal. Go to NCBI Nucleotide and search for the term
RefSeq_Select[Filter]
; then use the 'Send To' link at the top right corner to download 'File' in 'GFF3' format. The latter approach returns a GFF3 file that does not include all of the information normally included in the GFF3 files on FTP but that may be sufficient for your needs.You can probably use MANE instead. There is also RefSeq RNA fasta file for GRCh38 available from UCSC.
MANE is still in progress. There are still several genes that are not part of MANE. For example, only protein-coding genes are currently in the scope of MANE and that too, not all protein-coding genes are in MANE yet. And MANE picks one representative transcript for every gene. So, alternate splice variants that use a different promoter they are not included in the MANE set. If splice variants are important for your downstream analyses, MANE may not be the best choice. However, if you are interested in just one representative transcript for each gene, RefSeq Select may be a better choice for you. Only protein-coding genes are in scope for RefSeq Select as well but at least all genes have a RefSeq Select and MANE is a subset of RefSeq Select.