Graph Attention Network on Extracting Feature from Simplified Molecular-Input Line-Entry System for HIV Classification
Gregory Hugo, Vincentius Loanka Sinaga, Ignatius Michael Dinata, Felix Indra Kurniadi, Maria Susan Anggreainy · 2022
Human immunodeficiency virus targets the immune system and weakens people defenses against the variety of infections. Several research try to detect HIV but mostly used Protein as their input. However, the protein structure is hard to understand to the complexity of protein, in this paper we used different approach by using Simplified Molecular-Input Line-Entry System (SMILES). The SMILES provides easier analysis because the linear structure compared to Protein structure. In this paper we proposed Graph Attention Network on extracting Simplified Molecular-Input Line-Entry System (SMILES). We compared our proposed method to renown library to extract SMILES features namely RDKIT. The result showed the GAT gave the best result in accuracy with the 96.5%. However, it has many disadvantages in handling the imbalanced data.