1. Go to NCBI Entrez Protein search
2. Search with all [filter] query. This will give you all protein entries
3. Locate "Taxonomic Groups" box on the right. Display tree and locate "Green plants". Click and wait a moment. You should now see proteins from green plants. The query in the Search box should change to (all [filter]) AND "green plants"[porgn:__txid33090].
4. Download everything as GI list. This is your "Plant GI list"

Now you can:

1. Download full nr database in FASTA format

2. Using a custom script select from the nr database only those entries that have a GI from the "Plant GI list"

3. Create the final plant_nr using formatdb

OR:

use new blast where apparently you can filter the nr database based on gi list using '-gilist' option of blast itself! (http://www.ncbi.nlm.nih.gov/books/NBK1763/). But I haven't used that yet.

Alternatively, plantgdb maintains a uniprot-curated list of plant protein sequences.

It should be the UniProt_Protein.tar.bz2 file in ftp://ftp.plantgdb.org/download/FASTA/

This will extract out A LOT of sequence files that I wanted to join into one big file. I had to use xargs to get around the 'argument list too long' error that cat was giving me.

$ find . -type f | xargs cat > out.txt