**2. Methods**

#### **2.1 Pharmacokinetic properties**

The canonical SMILES (Simplified Molecular Input Line Entry System) format for solanidine, α-solanine, α-chaconine, β-solanine, β-chaconine, γ-solanine, and γ-chaconine was retrieved using the PubChem database (https://pubchem.ncbi.nlm. nih.gov/). PubChem is a public database that provides access to chemical compounds and their associated information, including SMILES strings [70].

ADMET properties were tested for solanine, chaconine, and their parent molecule, solanodine using SwissADME [71]. All of the compounds were assessed for various properties, including lipophilicity measured by XLogP3, topological polar surface area (TPSA), hydrophobicity and solubility measured by Log S, saturation of carbons in sp3 hybridization (carbon fraction sp3), flexibility measured by rotatable bonds according to the Lipinski rule, ability to cross the blood-brain barrier (BBB), human intestinal absorption (HIA), potential interaction with P-glycoprotein (PGP), inhibition of cytochrome P450 isoenzymes, and skin permeation parameters.

### **2.2 Drug target prediction**

The SMILES format if α-solanine and α-chaconine was imported to three computational tools for target prediction: (a) SuperPred (https://prediction.charite. de/index.php), a tool that predicts protein-ligand binding using a combination of machine learning models, including random forest and support vector machines [72]. Targets with both probability and model accuracy greater than 70% were selected. (b) SwissTargetPrediction (http://www.swisstargetprediction.ch/), a tool

that predicts drug targets based on a combination of bioinformatics and machine learning methods. Targets with a probability greater than zero were retrieved. (c) Similarity ensemble approach (https://sea.bkslab.org/), a method that predicts drug targets by comparing the chemical structure of compounds to known human protein structures. All human targets were retrieved from the similarity ensemble approach. The targets retrieved from each tool were normalized using the UniProt database (https://www.uniprot.org/) and merged for each compound. Duplicate targets were removed.
