IUPAC and soft-masking codes

[arodel21]

Hello,

This might be a silly question, but I couldn't find information in the code.
Are IUPAC ambiguity codes (R, Y, M, etc) and soft-masked sequences (a, c, g, t) included in the DNA sequence embedding?

Thank you!

[divya0709]

If I am understanding correctly, this is what I found....
From the one_hot_encode function in enformer.py, only ‘ACGT’ are one-hot encoded. Characters outside this alphabet (including IUPAC ambiguity codes and lowercase bases) map to an all-zero vector, same as ‘N’. So they are not explicitly supported unless preprocessed.