FtsZ‐LSTM: Generative AI Model for AntiBacterial Hits Discovery
Rupal Kushwah, Sapna Nikam, Kanika Manchanda, Prasad V. Bharatam · ChemistrySelect · 2026
ABSTRACT Filamentous temperature‐sensitive mutant Z (FtsZ) is an important cytoskeletal protein target for the discovery of antibacterial agents. To accelerate this process, Generative artificial intelligence (GAI) models are being integrated into early‐stage hit identification with target‐specific constraints. However, a sequence‐based GAI framework specifically adapted for FtsZ‐targeted molecular design remains limited. In this study, a GAI pipeline was implemented to generate novel chemical entities targeting FtsZ. The model was pretrained on antibacterial molecules retrieved from the ChEMBL database in the form of SMILES. The developed model was fine‐tuned independently for the two binding domains of FtsZ (i) GTP binding site and (ii) inter‐domain cleft (IDC), using reported site‐specific inhibitors. The performance of two fine‐tuned FtsZ‐LSTM models was benchmarked against different GAI architectures (BILSTM, VAE, GRU, and Transformer) using standard evaluation metrics. Virtual screening of the generated compounds was performed against Sa FtsZ at both binding sites, followed by molecular docking, molecular dynamics, and MM‐GBSA calculations. GTP124 and IDC9 were identified as promising site‐specific candidates, with better binding affinity in comparison to the reference molecules (GDP and TXA707) within their respective domains. This work demonstrates the utility of transfer learning‐enabled LSTM‐based GAI for designing FtsZ‐targeting molecules and highlights its application in antibacterial drug discovery.