Correct machine learning on protein sequences: A peer-reviewing perspective