Open walt676 opened 3 months ago
I think I got it: LAM is used for MSAs, the SSW is for local pairwise alignment, these is no convolutional layer in SSW.
But I still don't understand how the similarity score a_{i,j}
between two elements X_i
and Y_j
in the original Smith-Waterman algorithm is calculated when only SSW is used.
Shuji, Thank you for contributing such valuable code! After reading your code and the original paper "End-to-end learning of multiple sequence alignments with differentiable Smith-Waterman", I have one question about Learned Alignment Module(LAM) in paper: I did not see the related code about LAM. For example, no context-dependent alignment score matrix
a_i_j
computed by convolutional layer. Do I have some misunderstandings about the paper? Your reply will be of great help to me. Thank you again for your contributions.Zexin Zhang