-
Notifications
You must be signed in to change notification settings - Fork 162
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
How to use viz_sequence. #110
Comments
Hi Fritz, based on your email to me it seems like you figured out the visualization. I am not sure I understand your second question i.e. "how do i make the nucleotides that contribute the same for the true positives and true negative look like the contribute nothing" - aren't the sequences different between the true positive and true negative sequences? If so, in what sense is a given nucleotide contributing "the same" to both true positives and true negatives? Do you mean to identify locations within the sequence that are specifically contributing to true positives? If so, you could probably compare the average score across all true positives to the average score across all true negatives - but I'm not sure this is what you are looking for. |
Hi @AvantiShri,
i have two questions:
i have used deeplift to compute importance scores for my sequences (true positive and true negative predictions). I wanted to get a logo representation as the one in the notebook under examples which uses deeplift.visualization. How will you suggest i do that. I want the sequence logo to point out the nucleotides with really highscores that are responsible for shifting the decision to the particular class. I am working with a binary classification problem, where sequences are either expressed or unexpressed. The output dimension of my sequence inputs are (1, 4, 1000, 1).
My second question is how do i make the nucleotides that contribute thesame for the true positives and true negative look like the contribute nothing. I want to focus on only the positions which contribute to either true negative or true positive.
Below is small part of my deeplift code that calculates the contribution scores.
scores_tp = np.array(deeplift_contrib_func(task_idx=1, input_data_list=[tp_data],
input_references_list=[tp_shuff_data],
batch_size=10, progress_update=1000))
scores_tn = np.array(deeplift_contrib_func(task_idx=1, input_data_list=[tn_data],
input_references_list=[tn_shuff_data],
batch_size=10, progress_update=1000))
The text was updated successfully, but these errors were encountered: