Add Epitope data from TDC

[phalem]

I add Epitope data from tdc here :
https://tdcommons.ai/single_pred_tasks/epitope/

I will need help in validation my approach, need to ensure the indices start with 0 or 1 in epitope active binding one

[kjappelbaum]

@phalem Thanks a lot for your efforts; it is really great to have you take the lead on the datasets; we could only have that much progress thanks to your contributions.
One thing with your PRs is that they currently fail in our automatic checks. This is mostly the case because some lines need to be shorter. Could you break long (>120 character) lines to make the content more readable?

Do you think that makes sense? If you want help with that or to discuss this, I'm happy to do so.

[phalem]

@phalem Thanks a lot for your efforts; it is really great to have you take the lead on the datasets; we could only have that much progress thanks to your contributions. One thing with your PRs is that they currently fail in our automatic checks. This is mostly the case because some lines need to be shorter. Could you break long (>120 character) lines to make the content more readable?

Thanks kevin for your kind words. I think that problem raise from description that TDC provide. This can be solved by break target description and URI links, other things follow the same workflow. Please, tell me if you have time now to do that during review or I should do by myself. If this doesn't work, we will need to see if description need to be summarized

[phalem]

Add benchmark field

[kjappelbaum]

so the target is a sequence and the input is the full sequence?

[phalem]

so the target is a sequence and the input is the full sequence?

Yeah, I forget to comment on that. Input was sequence and target is sequence as well we could make it categorical, but with a lot of category. I don't know what is your opinion of that.

[MicPie]

Yes, this is how it is currently implemented. I set it up to only give the corresponding epitope sequence without the _ for the other parts of the sequence.

[kjappelbaum]

Once more: Thanks 💯
(And similar comments)

[phalem]

Before I forget @kjappelbaum we need someone to ensure if index of start with 0 or 1. As it will be different sequences and situation

[kjappelbaum]

@MicPie we need to carefully look into this. I don't think this is a simple classification/regression task.

[kjappelbaum]

Also need to check how this overlaps with #67