I wrote a script awhile back that takes the longest isoform per gene and writes it to a new fasta. I've adapted it to solve your problem. Rename your collections.fasta to collections.txt, if it is not in fasta format. This script will take the following inputs: your trinity fasta containing all transcripts, your gene ids file (one ids per line), and the name of your output fasta.

#!/usr/bin/env python

import sys
from collections import defaultdict
from Bio import SeqIO

fasta_file = sys.argv[1]
gene_ids_file = sys.argv[2]
output_file = sys.argv[3]

with open(fasta_file, "rU") as file:
        records = list(SeqIO.parse(file, "fasta"))

with open(gene_ids_file, 'r') as f:
        gene_ids = [line.strip() for line in f]

with open(output_file, 'w') as out:
        for i in gene_ids:
                for n in range(len(records)):
                        if i in records[n].id:
                                print >> out, '>' + records[n].id, '\n', records[n].seq

Save this to a file, and run with python script.py Trinity.fasta collections.txt outputfile