[ENH] AptaDB loader and dataset

[fkiraly]

Satarupa Deb pointed out an interesting dataset, AptaDB.
https://lmmd.ecust.edu.cn/aptadb

The datasets are available for download here: https://lmmd.ecust.edu.cn/aptadb/about.php - but there are no python bindings.

I would suggest to:

• upload the data on kaggle or huggingface
• add python loaders for the raw data (csv as-is, load as pandas.DataFrame

We can then add functionality to load them in unified formats on top.

For @KubiczekD and @JaBirke, it might also be interesting to see whether binding information can be extracted from the data. To me it looks like only binding aptamers are contained in the dataset? I.e., no non-binding or negative results.

[Satarupa22-SD]

@fkiraly Please assign it to me. Thanks!

[fkiraly]

done

[KubiczekD]

@fkiraly The AptaDB Database was already included in AptaCOM.
The used databases can be seen here: https://jc-biotechaiteam.com/AptaCom/about/
Besides AptaDB, AptaCOM used UTexas, AptaBASE, AptamerBASE, AptaIndex and NAKB.

[Satarupa22-SD]

@fkiraly, in that case, we should skip adding AptaDB separately. @KubiczekD, this is the dataset, if I am not wrong, https://github.com/rpgv/AptaCom/blob/main/Database/AptaCom02.csv

[fkiraly]

Hm, I see. @rpgv wanted to add python bindings for AptaCOM.

However, it looks like that fields were omitted from AptaDB when including to AptaCOM, simply since the different databases did not have the same fields.
Plus, it might be useful to have protein sequences or their RCSB identifiers directly.

Hence I would say, it is still very much worth adding AptaDB separately, if only due to the protein codes and the fields not in AptaCOM.

[Satarupa22-SD]

@fkiraly Thanks for confirming, I have create the dataset here : https://www.kaggle.com/datasets/satarupadeb/aptamer-interactions. Working on the loader function.