. 2023 Jun 22;8(26):23566–23578. doi: 10.1021/acsomega.3c01218

Table 2. Different Descriptor Spaces (DS) Used to Represent the 2D Structure of Insulin Analogs^a^,^b.

name	type of descriptors	examples	descriptor space dimension
DS1	overall numeric molecular descriptors calculated from the entire sequence of the insulin analogs	size, charge, hydrophobicity	15
DS2	physiochemical molecular descriptors of the fatty acid side chain (acylation group)	surface area, LogP, molecular weight	7
DS3	NLP embedding approach ESM-1b,³⁰ encoding the entire backbone sequence (insulin and attached amino acids/sequences)	GIVEQCCTSICSL	1280
DS4	SMILES representation of the fatty acid side chain Mol2Vec³¹ used for embedding of SMILES	NC(C)C(=O)O	100

Abbrevations: NLP, Natural Language Processing; ESM, Evolutionary Scale Modeling; SMILES, Simplified Molecular-Input Line-Entry System.

For more details on descriptors, see Supporting Information Table S3, Table S4, and Figure S3.

Table 2. Different Descriptor Spaces (DS) Used to Represent the 2D Structure of Insulin Analogsa,b.