The construction of a Fyn SH2 domain variant library and quality control. (a) Amino acid sequence and secondary structure of Fyn SH2 domain. The eight residues for mutation were numbered. Mutagenesis were introduced by Kunkel method by three primers, in which primer 1 targeting residue 1 (Region 1), primer 2 targeting residues 2,3,4,5 and 6 (Region 2) and primer 3 targeting residue 7 and 8 (Region 3) to generate the “stop template”. Then three degenerated primers, ie. primer4(targeting Region 1), prime5 (targeting Region 2) and primer6(targeting Region 3), were applied to introduced combinatorial mutations at Region 1, 2, and 3, respectively to generate the library based on the “stop template.” (b) The difference between the actual amino acid distribution based on the deep‐learning screening and the theoretical amino acid distribution at the eight positions were shown in the heat map