AI Predicted the Peptide Would Bind to Its Target, but the Laboratory Disagreed: Why Computer Models Cannot Replace Experimental Evidence