This is 2 years old? Why did I just find it?
Meh - maybe it wasn't very good until the version posted 3 months ago that is the one I'm using?
Here's the thing - lots of tools can do the deep learning stuff for you now. We've had PROSIT for years and ElFragmentador is a great web tool.
Why use something else?
This one installs locally (pretty easily, as well) and not only kicks out your predicted peptides and peptide fragments - you also get your 1/k0 values (if you're using one of those screwy TIMS things).
In TIMSTOF data it's pretty hard to just go into your officially licensed copy of Data Analysis and find a peptide you're looking for. But if you have the licensed version (or you can beg your sales rep into multiple free demos) you can find your peptide if you know your 1/k0 value and retention time and mass.
Or you can just predict all of that on your desktop with this easy GUI.
A couple targets is no problem at all. Dump in one protein? It digests it and kicks out your predictions. Super straight-forward.
Got a lot? DON'T DO IT ON YOUR CPU! Unless you want to heat your office or something. PeptDeep can also be configured to use your GPU. I haven't done that part, but I probably will. Why have 8,000 CUDA cores just sitting there doing nothing?
Even cooler? You can easily retrain your prediction models based on your own data. One of the recent updates on Github is an HLA model! But you can take any result file that has real data (formatted properly, of course) and drop it in to improve your models. All the coolest bioinformatics kids know how to do that, but this is the first time I've had something easy enough that I could use.
No comments:
Post a Comment