Arabidopsis GFF files

Question

[biolab]

Dear all,

Does anyone who works on Arabidopsis know what's the differences between TAIR10_gff3_genes.gff and Spliced_Junctions_clustered.gff. Both are downloaded from ftp://ftp.arabidopsis.org/home/tair/Genes/TAIR10_genome_release/TAIR10_gff3/

I found the former has 217183 exon records, while the latter has 406034 exon lines. Is the first experimentally confirmed and the latter in silico predicted?

Thanks for your advice!

Answer 1

[zoo741]

Although this question was brought up a while ago, I thought it's worth giving some updates. Both of these 2 GFF3 files are based on experimental evidence yet took different routes. The details are summarized as follows,

Spliced_Junctions_Clustered.gff

We used high-throughput RNA sequencing data (RNA-seq data) from the Ecker and Mockler labs, and used alignment tools called Tophat and Supersplat to align these sequences to the Arabidopsis genome, resulting in 203,000 clustered spliced RNA-seq junctions.

Cited from ref1

TAIR10_gff3_genes.gff

utilized RNA-seq, proteomic datasets, gene models provided by NCBI and manually curated gene models from Swiss-Prot. It went through multiple steps including mapping, assembling, gene model construction etc. ref2

In summary, one could see Spliced_Junctions_Clustered.gff as the direct output of TopHat, likely junctions.bed, for RNA-seq data sets. On the other hand, TAIR10_gff3_genes.gff used the datasets described above to update the gene models. There are more exon records in TAIR10_gff3_genes.gff because it includes the junctions for ALL the transcripts, many of which may not be detected in the tissue used to generate Spliced_Junctions_Clustered.gff. On the other hand, there should be a bunch of junctions unique to Spliced_Junctions_Clustered.gff because those junctions may not go through the pipeline for TAIR10_gff3_genes.gff, i.e. did not assemble into a transcript.

Answer 2

[biogirl]

This webpage tells you how the annotation was done. Therefore, I imagine the TAIR10_gff3_genes.gff contains a mixture of models and experimental data. The Spliced_Junctions_clustered.gff file will contain the splice variants as predicted by gene models.

[biolab]

Thanks a lot for your comments!