Name Last modified Size
Parent Directory 30-Jul-2018 20:24 -
About_EvidentialGene.html 11-Jun-2015 14:16 2k
Annotate.info 30-Mar-2011 14:10 1k
Configure.info 30-Mar-2011 14:23 1k
Evaluate.info 30-Mar-2011 14:18 1k
EviGene-results-summary.txt 08-Jun-2011 11:42 10k
config/ 27-Dec-2011 12:11 -
docs/ 10-Sep-2018 16:25 -
eugenes_proteome_y2k_sizes.txt 03-Dec-2012 01:37 6k
evigene_rnaseq_2012_stats.txt 06-Dec-2012 12:00 9k
evigene_rnaseq_denovo2012.txt 07-Jun-2012 21:29 3k
lib/ 03-Nov-2017 14:10 -
scripts/ 09-May-2018 13:35 -
EvidentialGene Gene Set Reconstruction Software
Don Gilbert, gilbertd At indiana edu, 2018
About Evigene-R : genes assembled from RNA pieces
evgpipe_sra2genes is an ombnibus pipeline to reconstruct genes, using
several EvidentialGene methods, starting at Public SRA database of RNA-Seq
(but can start w/ own RNA), and finishing with publication quality,
annotated gene sequences. The basic steps are outlined below.
See also evigene/docs/evgpipe_sra2genes.help.txt
See also http://eugenes.org/EvidentialGene/about/EvidentialGene_trassembly_pipe.html
Too many transcript assemblies are much better than too few. It allows one
then to apply biological criteria to pick out the best ones. Don't be
misled by the "right number" of transcripts that one or other transcript
assembler may produce. It is the "right sequence" you want, and now the
only way to get it is to produce too many assemblies (an "over-assembly")
from RNA data, with several methods and several parameter settings.
EvidentialGene tr2aacds.pl is a pipeline script for
processing large piles of transcript assemblies, from several methods
such as Velvet/Oases, Trinity, Soap, etc, into the most biologically useful
"best" set of mRNA, classified into primary and alternate transcripts.
It takes as input the transcript fasta produced by any/all of the
transcript assemblers. These are classified (not clustered) into valid,
non-redundant coding sequence transcripts ("okay"), and and redundant,
fragment or non-coding transcripts ("drop"). The okay set is close to a
biologically real set regardless of how many millions of input assemblies
you start with.
It solves major problems in gene set reconstruction found in other methods:
1. Others do not not classify alternate transcripts of same locus properly,
Alternates may differ in protein quite a bit, but share identical exon parts.
2. Others remove paralogs with high identity in protein sequence, which
is a problem for some very interesting gene families.
3. Others may select artifacts with insertion errors by selecting longest transcripts.
Evigene works first with coding sequences.
Quality assessment of this Transcript Assembly Software is
described in http://eugenes.org/EvidentialGene/about/EvidentialGene_quality.html
About Evigene-G : traditional genes modeled-on-genome
This works on gene locations on chromosome assembly, in GFF v3 format tables.
Gene models with overlapping CDS exons are "the same locus", each model has
some form of evidence score, and the method picks out those models with
highest evidence score. The trick or trouble is mainly in applying
various evidence scores, and their combination, to return the best models
that a human expert would pick.
See also evigene/docs/evg_overbestgenes.help.txt for details
About Evigene-N : non-coding gene reconstruction
See also evigene/docs/evigene_goals2015b.txt
About Evigene-H : gene reconstruction with hybrid of methods
See also evigene/docs/evigene_goals2015b.txt
=item HOW TO GET SOFTWARE
EvidentialGene software packaged as tar files are what you want, from here
EvidentialGene software in unpackaged form (lots of files) is here
is same as evigene/docs/ in your copy of this package.
=item WHO USES IT?
=item IS IT ANY GOOD?
=item HOW TO INSTALL
Extract the tar archive this way, into current folder, preserving run permission.
tar -xf evigene.tar
Run the Perl ".pl" scripts from extracted evigene folder, as they are a package.
export evigene=`pwd`/evigene; # Unix bash shell, or
set evigene=`pwd`/evigene; # Unix csh/tcsh shell
$evigene/scripts/prot/tr2aacds.pl [options] ..;
$evigene/scripts/evgpipe_sra2genes.pl [options] .. ;
$evigene/scripts/evgmrna2tsa.pl [options] .. ;
Required additional software
You need these additional software for tr2aacds, installed in Unix PATH or via
* fastanrdb of exonerate package, quickly reduces perfect duplicate sequences
* cd-hit, cd-hit-est, clusters protein or nucleotide sequences.
http://cd-hit.org/ OR https://github.com/weizhongli/cdhit/
* blastn and makeblastdb of NCBI BLAST, https://blast.ncbi.nlm.nih.gov/
Basic Local Alignment Search Tool, finds regions of local similarity between sequences.
Most of the shell ".sh" scripts require editing for your cluster; consider
them examples. Perl scripts have brief -help, but most of their
documentation is perl POD. This is a complex package, including my working
scripts for several genome projects, some are obsolete now. One needs a
cheat-sheet to make sense of what is good and I am working on such.
=item TEST DRIVE
Please first try this test case with small input data (TAIR10 mRNA) and tr2aacds outputs,
It is worth your time to set up and run this with same input data to see
that you get same results.
See also evigene/scripts/prot/tr2aacds_test.sh
env trset=arath_TAIR10_20101214up.cdna.gz datad=`pwd` ./tr2aacds_test.sh
env trset=arath_TAIR10_20101214up.cdna.gz datad=`pwd` qsub -q normal tr2aacds_test.sh
You should be able to get same result from same Arabidopsis transcripts
input data file, and where problems appear, please consult a local computer
expert familiar with your cluster computer to resolve. After you get that
test set working ok, running on your data set should be simpler.
=item BASIC USAGE of Evigene-R
See steps in evigene/scripts/evgpipe_sra2genes.pl
trset=$myspecies.cdna # 1 fasta input file with many transcript sequences, assembled or otherwise
evigene=/your/path/to/evigene # where you un-tarred evigene.tar
ncpu=1 # or 2 or 8 # 8 cpu probably enough, each uses 2+ GB memory
maxmem=32000 # in megabytes, expect 2+GB per cpu, maybe more for complex large over-assemblies
STEP 1. get RNA-Seq data
STEP 4. run assemblers of RNA-seq, with kmer size options, other opts
4a. velvet/oases, ~10 kmer steps
4b. idba_tran, ~10 kmer steps
4c. soap_trans, ~10 kmer steps
4d. trinity / other / user choices
STEP 5. trformat.pl, post process assembly sets
subd=veloset # velvet run directory with several velvet kmer subfolders
$evigene/scripts/rnaseq/trformat.pl -pre $myspecies -out trsets/$subd.tr -log -in $subd/vel*/transcripts.fa
subd=idbaset # idba run directory, several transcripts-kmer.fa outputs
$evigene/scripts/rnaseq/trformat.pl -pre $myspecies -out trsets/$subd.tr -log -in $subd/transcripts-*.fa
Catenate all transcript sets to one file:
cat trsets/*.tr > $myspecies.cdna
STEP 7. tr2aacds, reduce over-assembly to draft gene set
$evigene/prot/tr2aacds.pl -tidy -NCPU $ncpu -MAXMEM $maxmem -log -cdna $myspecies.cdna
STEP 10. evgmrna2tsa, produce public gene sequences
$myspecies.trclass is a result from STEP 7, tr2aacds
$evigene/scripts/evgmrna2tsa2.pl -onlypubset -idprefix $myspeciesEVm -class $myspecies.trclass
=item TR2AACDS PIPELINE ALGORITHM
Prerequisite is that you create transcript assemblies (with any/all
useful methods). This software reads all the transcripts.fasta you have,
chews on them and puts them into good and bad piles (with extras).
0. collect input transcripts.tr,
You supply input transcript sequences in .fasta, an over-assembly with
redundant and variable quality transcripts, as one file.
1. perfect redundant removal: exonerate/fastanrdb input.cds > input_nr.cds,
and protein qualities are used for choosing among cds identicals.
2. perfect fragment removal: cd-hit-est -c 1.0 -l $MINCDS ..
Cluster *idential coding sequences*, short and long, keep the longest
3. blastn, basic local align hi-ident subsequences for alternate tr.,
with -perc_identity CDSBLAST_IDENT (98%), to find high-identity
4. classify main/alternate cds, okay & drop subsets, using
merges alignment table, protein-quality and identity, to score
okay-main, ok-alt, and drop sets.
5. final output files from outclass: okay-main, okay-alts, drops
okayset is for public consumption. The drop set of redundant, fragment,
non-coding sequences, may contain valid coding sequences (more details).
=item OTHER EVIDENTIALGENE COMPONENTS
See STEP 5 of evgpipe_sra2genes.pl
Use BEFORE tr2aacds to regularize IDs in fasta of
Velvet,Soap,Trinity, ensure unique IDs, add prefixes for parameter sets.
See STEPs 8-9 of evgpipe_sra2genes.pl
Use AFTER tr2aacds on okayset, add gene function names from
UniProt-Ref and CDD blastp.
blastp -db refprots -query okay_all.aa -outfmt 7 -out $name.blastp
namegenes.pl -refnames $refdb.names -blast $name.blastp
See STEPs 8-9 of evgpipe_sra2genes.pl
UniVec vector screening and NNN-end trimming, per NCBI or INSDC desires
See STEPs 10 of evgpipe_sra2genes.pl, See evigene/docs/evgmrna2tsa_help.txt
make public mRNA gene set, with pubIDs,
main/alternates, names and annotation, and Genbank TSA format for
=item HELP AND METHOD DOCUMENTS
How To get Best mRNA Transcript assemblies
Please read this brief How-To document that summarizes my tests on best
transcript assembly methods. It points out tips for assembly parameters,
such as using scaffolding and multi-kmer settings (for Velvet, Soap,
others that allow; not Trinity), that will improve your transcript
Best Assembler methods
Best assembly methods compared for mosquito genes
has recent comparison of gene assembler accuracy,
EvidentialGene tr2aacds mRNA classifier description
Classification is based mainly on CDS-dna local alignment identity.
Perfect fragment CDS are dropped, those with some CDS base differences
are kept, with longest CDS as primary transcript. UTR identity is
ignored (for now) because many of the mis-assemblies are from
joined/mangled genes in UTR region.
Error of selecting longest transcripts, as with CD-HIT-EST
Selecting genes by longest transcripts is a mistake.
EvidentialGene ORF/protein computation
EvidentialGene computes ORFs (proteins and coding sequences of those), and
its method is drawn on Brian Haas's ORF computations, which also form the
Transdecoder package now. I've recently looked at results from
Transdecoder versus Evigene, and I don't think Transdecoder is giving you
improvements, it may well be reducing the number of best orthology proteins.
Validating with RNA-Seq map-back
RNA-seq mapping methods are influenced strongly by presence of duplicated
sequence spans, as with alternate transcripts and high identity paralogs.
An accurate statistic of proper paired fragment mapping to a
given transcript should give the same value regardless of whether
alternates to that transcript exist.
TransRate, and some other RNA map-back validation methods, produce
inaccurate statistics, that are influenced by presence/absence of other
biological alternates and paralogs.
Developed at the
Genome Informatics Lab
of Indiana University Biology