Plant gene set reconstructions with EvidentialGene are accurate
 

Recent Animal and Plant gene set reconstructions with EvidentialGene:
comparisons to other popular, recent gene reconstructions.

Comparison to Pac-Bio RNA sequencing, Trinity-Illumina assembly, and genome
gene models from NCBI and MAKER pipelines, indicate EvidentialGene methods are
more accurate than commonly used methods. Evigene sets for Arabidopsis model
plant, Zea mays corn, and pine trees, animals of Bemisia white fly, Daphnia
water fleas, Aedes and Anopheles mosquitoes, and others are available at
http://eugenes.org/EvidentialGene/

Not only are the easy, well known ortholog genes reconstructed well, but
harder gene problems of alternate transcripts, paralogs, and complex
structured genes are usually more complete with Evigene methods.

Who should use EvidentialGene for animal and plant gene reconstruction?

* genomicists desiring accurate, complete and objectively reconstructed genes
including those of you who may not believe my claims, but will look at
objective results supporting them.

* new species genome projects
- use as primary gene set, with most alternate transcripts,
add the 10% un-expressed genes with modeling.
- assess genome gene models for accuracy and completeness.
- assess fragmentation, mis-assembly of chromosome assembly,
and use to join chromosome fragments

* model and well-supported genome projects
curators can use evigene reconstructions to improve precision of
high value gene information.

* gene/genome improvement projects
add missing alternate transcripts, un-discovered and fragmented gene models,
improve complex genes

* transcriptome and expression study projects
use for more accurate gene information as the base for expression comparisons

One of my goals with this work is to reconstruct many high-value (model,
otherwise) animal and plant gene sets in coming years as feasible. I welcome
collaborations, especially from any group who can provide genomics/informatics
expertise. This methodology is highly automatable (think BIG DATA), but still
wants some improvements. Over-assembly of suitable RNA takes a only few days
on compute clusters, and produces all the accurate genes, plus a bigger pile
of less accurate ones. The main time sink is in sensibly classifying and
reducing these to a "perfect" set (not too many, not too few), with use of
additional gene evidence.

Reconstruction from RNA only provides independent gene evidence, free of
errors and biases from chromosome assemblies and other species gene sets.
Evigene gene sets offer an independent assessment of a complete species gene
catalog, rather than the easiest few percent of genes represented in BUSCO
and other orthology reference sets.

There are now a few public Pac-Bio RNA gene sets, and publications suggesting
genes from single-molecule sequencing may be more accurate than genes from
Illumina short reads. My comparison for 3 plant species, Arabidopsis model
plant, Zea mays corn, and pine trees, provides an objective comparison with
different results: fully assembled Illumina RNA produces the more accurate
sets, including for loci where both methods recover some transcripts,
for alternate and paralog transcript reconstruction.

Evigene's RNA-only constructions often surpass accuracy of genome-modeled gene
sets, those derived from many sources of gene evidence (prediction on
chromosomes, RNA, other species proteins). This is likely due to the greater
complexity of combining many evidence sources in modeled genes, with greater
chances of mis-modeling.

These recent works include Arabidopsis model plant, Zea mays corn, and pine
trees, animals of Bemisia white fly, Daphnia water fleas, Aedes and Anopheles
mosquitoes, and others. Species genes built with Evigene by independent
authors include a range of plants, fishes, a mouse, insects, crustaceans, and
several of these papers provide their independent review of evigene versus
other methods.

-- Don Gilbert
gilbertd @ indiana.edu