Pete CarltonI've been gazing at per-residue #embeddings from the #protT5 model and it gives me the feeling of having a secret tome of insight but it's in a lost language that we need to decipher; and while large parts of the tome are likely nonsense there are certainly some interesting things lurking in there. Just a simple matrix multiplication of one protein (DSB-1) with itself shows some intriguing patterns:
