This might be more appropriate for atomworks, but the MMSeqs server automatically added a Null character (\x00) to my a3m file resulting in an error during rf3 inference. Traced it back to the parse_a3m() atomworks function which parsed the Null character as a sequence, causing an "inhomogenous shape" numpy error when the parsed sequences were loaded into the array (msa_array = np.array([list(seq) for seq in msa], dtype="S")). Since enough people will likely use the MMSeqs server for MSA generation, can you add a step to strip the trailing null character during a3m parsing?
This might be more appropriate for atomworks, but the MMSeqs server automatically added a Null character (\x00) to my a3m file resulting in an error during rf3 inference. Traced it back to the parse_a3m() atomworks function which parsed the Null character as a sequence, causing an "inhomogenous shape" numpy error when the parsed sequences were loaded into the array (msa_array = np.array([list(seq) for seq in msa], dtype="S")). Since enough people will likely use the MMSeqs server for MSA generation, can you add a step to strip the trailing null character during a3m parsing?