I'm trying to generate confusion matrices and an ROC curve from ALICE, and having a problem getting the ALICE output. Hope you can help look into it.
I input 963 sequences (list of one data table in _alice_in) and got 143 ALICE hits back in alice_res_. Code below:
alice_res = ALICE_pipeline(DTlist = alice_in,folder="res3",cores=1,iter=10,nrec=5e5)
If I understand correctly:
ALICE returns only positive classes as ALICE hits, the number of negative classes can be calculated by subtracting from the input data.
So 143 is the number of positive TCR binding events/responses that ALICE detects, which allows me to build one confusion matrix. And as I change P_thres in ALICE_pipeline function, I should get different output for more confusion matrices. But no matter what P_thres I give to the function, I'm always only getting 143 ALICE hits.
If I understood the algorithm incorrectly please point out. And if you can please explain how the ROC curve in the paper was generated that would be great.
Hi Mr. Pogorely,
I'm trying to generate confusion matrices and an ROC curve from ALICE, and having a problem getting the ALICE output. Hope you can help look into it.
I input 963 sequences (list of one data table in _alice_in) and got 143 ALICE hits back in alice_res_. Code below: alice_res = ALICE_pipeline(DTlist = alice_in,folder="res3",cores=1,iter=10,nrec=5e5)
If I understand correctly: ALICE returns only positive classes as ALICE hits, the number of negative classes can be calculated by subtracting from the input data.
So 143 is the number of positive TCR binding events/responses that ALICE detects, which allows me to build one confusion matrix. And as I change P_thres in ALICE_pipeline function, I should get different output for more confusion matrices. But no matter what P_thres I give to the function, I'm always only getting 143 ALICE hits.
If I understood the algorithm incorrectly please point out. And if you can please explain how the ROC curve in the paper was generated that would be great.
Looking forward to hear back from you.
SIncerely, Louis