Download 1,000 Random Human RefSeq Transcripts

[mapauley]

Hi all,

I want to download a specific number (e.g., 1,000) of random--it's essential that they be random, although "representative" is perhaps a better word--RNA reference sequence transcripts, preferably from a specific reference build (e.g., GRCh38), although this isn't super important. Any thoughts on how to do this?

By random I mean that there would be nothing to distinguish one batch of 1,000 sequences from another, e.g., in the number of curated (NM_, NR_) versus model (XM_, XR_) sequences. As alluded to above, my goal is to have a completely representative subset of transcript sequences.

Mark

[vkkodali]

I am assuming that you are interested in fetching sequences in FASTA format.
I suggest you download a file with all transcript sequences in FASTA format for a specific genome assembly's latest annotation using NCBI Datasets and then use a program like seqkit to extract n random sequences from the file.
EntrezDirect is not the most efficient way to download data in bulk. But if you don't have a choice, one way to achieve something like this would be to: (1) get a list of all accessions using a combination of esearch and efetch -format acc, (2) use a Unix tool like sort -R or shuf to extract n random accessions, then (3) use a combination of epost and efetch to download sequences in FASTA format.

[mapauley]

Very helpful. Thank you! I downloaded the GRC38 patch Patch 14 transcripts, extracted just the accession
numbers, randomized the result using shuf, and then downloaded the corresponding records via efetch.