Efficient Gene Prediction on Large Eukaryotic Genomes

Question

[Shuo]

I want to run gene prediction tools on a large eukaryotic genome dataset. I started with AUGUSTUS because it provides species-specific parameters that I need. However, I found that it runs quite slowly, often taking up to two hours to process a single FASTA file.

On the other hand, GlimmerHMM runs faster but lacks comprehensive species parameters. I am looking for gene prediction tools that can run efficiently while maintaining good accuracy.

Are there any faster gene prediction tools that still offer species-specific models? Besides using a large-scale computing cluster, what other methods can be used to accelerate gene prediction?

Thanks in advance for your suggestions!

[Mensur Dlakic]

Not sure why we are discussing this. Is it more important to you to get these gene predictions right, or to save 20-50 hours? Unless you have to submit a paper or defend a thesis / dissertation in two weeks, it shouldn't matter if you pick a higher quality predictor that takes a bit longer.

[GenoMax]

on a large eukaryotic genome dataset

and

I found that it runs quite slowly, often taking up to two hours to process a single FASTA file.

How many genomes are being analyzed? Unless it is hundreds (which would have taken significantly longer to assemble) how come something taking two hours considered "running slowly".

[colindaven]

As others have said, 2 hours is nothing. Wait till you get to assembly or long read alignment. The quality of results is far more important.

If you want to optimize

• learn a little patience <- start here
• divide up the fasta by chromosome or contig and run augustus in parallel, then recombine results