search

simaataei / TooT-BERT_ICAT

0 stars 1 forks source link

How to use ToT-Bert_ICAT on a protein FASTA file #2

Open Rohit-Satyam opened 7 months ago

Rohit-Satyam commented 7 months ago

Hi @ramou @simaataei

I was able to run Toot-BertT easily given its straightforward installation but how can I run ToT-Bert_ICAT? Are you still writing executables like run.py or something that we can use

ramou commented 7 months ago

Hi Rohit, I'd wait for Sima's response, but in the meantime... I recall discussions of this tool, but it's not yet passed through me. After a quick glance at the code, I suspect the Train.py is an example of how to use it given a specified dataset (Uniprot in that example). Presumably you'd write your own test set after and feed it in to get the results given that said method outputs to a results.txt file. Definitely not the more polished interface we've been pushing on the other bioinformatics-group tools.

Absent a response from Sima, that's where I'd start.

simaataei commented 7 months ago

Hi Rohit,

This project has been updated with a better as the "Inorganic ion predictor". You can find the run file and the instructions here . Thank you for bringing it up I forgot to ask @ramou to update the link. @ramou could you please update the link for the TooT-BERT-ICAT project to the inorganic ion predictor one?

ramou commented 7 months ago

https://github.com/bioinformatics-group/TooT-BERT_ICAT ... it looks like I didn't manage to maintain the history I'd been trying to do and it might be easier for me to nuke and re-fork... presumably it can still be called TooT-BERT_ICAT to go with the naming convention? I think @simaataei, you may also be able to muck with the group repo. Anyway, this should settle things for now.

Rohit-Satyam commented 7 months ago

. it looks like I didn't manage to maintain the history I'd been trying to do and it might be easier for me to nuke and re-fork... presumably it can still be called TooT-BERT_ICAT

Is this a stable release @simaataei ? Besides, the readme says that this tool classifies the transmembrane transport proteins so do I need to run something like Toot-BERT-T first to find the transmembrane transport protein and then feed the ones being predicted as transporter to Inorganic_Ion_Predictor or can I feed all the proteins irrespective of knowledge that they are transporter or not?

ramou commented 7 months ago

Apologies for the delay in response. While we wait for Sima's input, I thought I'd share my insights based on my involvement with the software's backend over the years. From what I've observed, pre-classifying as transporters doesn't significantly alter the outcomes when using any subsequent (more specific) classification tool in general (from the ones we've done that I poke around with). That said, identifying something as a transmembrane protein is indeed swift, and applying any TC tool on proteins already identified as transporters can certainly speed things along. Hope this helps until Sima can provide a more detailed answer!