Trinity-home

Likely coding regions can be extracted from Trinity transcripts using a set of utilities included in the Trinity software distribution. The system works as follows:

  • the longest ORF is identified within the Trinity transcript (if strand-specific, this is restricted to the top strand).

  • of all the longest ORFs extracted, a subset corresponding to the very longest ones (and most likely to be genuine) are identified and used to parameterize a Markov model based on hexamers. These likely coding sequences are randomized to provide a sequence composition corresponding to non-coding sequence.

  • all longest ORFs found are scored according to the Markov Model (log likelihood ratio based on coding/noncoding) in each of the six possible reading frames. If the putative ORF proper coding frame scores positive and is highest of the other presumed wrong reading frames, then that ORF is reported.

  • if a high-scoring ORF is eclipsed by (fully contained within the span of) a longer ORF in a different reading frame, it is excluded.

The scoring of ORFs is largely based on that described for GeneID, and the ORF selection process was added here.

Extracting Best ORFs

Extracting likely coding regions from Trinity transcripts can be done using:

% $TRINITY_HOME/trinity-plugins/transdecoder/transcripts_to_best_scoring_ORFs.pl
with options:
############################# Options ###################################################
#
# Required:
#
# -t transcripts.fasta
#
#  Optional:
#
# -m minimum protein length (default: 100)
#
# -G genetic code (default: universal, options: Euplotes, Tetrahymena, Candida, Acetabularia)
#
# -h print this option menu and quit
# -v verbose
#
# -C complete ORFs only ********
# -S strand-specific (only analyzes top strand)
# -T top longest ORFs to train Markov Model (hexamer stats) (default: 500)
#
##############################################################################################

Sample Data

Sample data and example execution are provided at

TRINITY_RNASEQ_ROOT/sample_data/test_Trinity_Coding_Extraction

Just run runMe.sh to run the process.

The process involves

Final output files include:

best_candidates.eclipsed_orfs_removed.cds
best_candidates.eclipsed_orfs_removed.pep
best_candidates.eclipsed_orfs_removed.gff3
best_candidates.eclipsed_orfs_removed.bed

The .bed file can be loaded into IGV for viewing along with the additional Trinity assembly and alignment data.