Fig. 6. The key hydrophobic residues appear in different positions in the sequence similarity network (SSN).

The SSN is constructed with each active variant as a node and an edge connecting every pair of variants that differ at exactly one position. Partitioning the SSN with the Leiden algorithm detected nine substantially populated clusters in this SSN, such that the variants are coloured according to cluster membership and node size is proportional to the number of active variants accessible from that variant via a single mutation. The location of the library variants used for affinity screening is indicated. The Ala insertion at position 8a causes a registry shift which partitions the SSN in two equal-sized parts. Within each half, each cluster is distinguished by a sequence fingerprint (WebLogos in boxes) with different anchoring hydrophobic residues.