An intrinsically interpretable neural network architecture for sequence-to-function learning.

The source code, including scripts for the analysis of key findings, can be found at https://github.com/boooooogey/ATAConv, implemented in Python.