search

Bioinformatics7181 / DeepLION

Deep MuLti-Instance Learning Improves the Prediction of Cancer-associated T Cell ReceptOrs for Accurate CaNcer Detection.
Other
3 stars 1 forks source link

Return weight for each sequence #1

Open Albert-Shuai opened 1 year ago

Albert-Shuai commented 1 year ago

Hi:

Thanks for the package uploaded. I would like to learn whether it is possible to return the weight of each sequence (Then we know which sequences play an important role in our sample level prediction) in addition to giving the prediction score for each sample. Thanks again.

Bioinformatics7181 commented 1 month ago

Hi,

Sorry, I just noticed your message now. We have proposed an improved model DeepLION2 based on DeepLION, which can better capture TCR correlations in a repertoire and predict the key TCRs with their motifs. FYI, the DeepLION2 is at https://github.com/Bioinformatics7181/DeepLION2.

Best, Xinyang

Bioinformatics7181 commented 1 month ago

DeepLION2 can obtain the key TCRs (> threshold like 0.999) with their key motifs. If you are still insterested in these issue, I can implement a function for you to obtain the specific score for each TCR in a repertoire.