The study introduces AntiBP3, a method for predicting antibacterial peptides (ABPs) against gram-positive, gram-negative, and gram-variable bacteria. The researchers first developed an alignment-based approach using BLAST, which achieved poor sensitivity. They then employed a motif-based approach, which provided high precision but low sensitivity. To address these issues, they developed alignment-free methods using machine/deep learning techniques, incorporating various peptide features such as amino acid composition, binary profiles of terminal residues, and fastText word embeddings. The models were evaluated using a five-fold cross-validation technique and showed maximum AUC values of 0.93, 0.98, and 0.94 for gram-positive, gram-negative, and gram-variable bacteria, respectively, on an independent dataset. The method outperformed existing approaches and was implemented as a user-friendly web server, standalone package, and pip package to facilitate peptide-based therapeutics. The study highlights the importance of capturing distinct features within each bacterial group and the effectiveness of machine learning models in predicting ABPs.The study introduces AntiBP3, a method for predicting antibacterial peptides (ABPs) against gram-positive, gram-negative, and gram-variable bacteria. The researchers first developed an alignment-based approach using BLAST, which achieved poor sensitivity. They then employed a motif-based approach, which provided high precision but low sensitivity. To address these issues, they developed alignment-free methods using machine/deep learning techniques, incorporating various peptide features such as amino acid composition, binary profiles of terminal residues, and fastText word embeddings. The models were evaluated using a five-fold cross-validation technique and showed maximum AUC values of 0.93, 0.98, and 0.94 for gram-positive, gram-negative, and gram-variable bacteria, respectively, on an independent dataset. The method outperformed existing approaches and was implemented as a user-friendly web server, standalone package, and pip package to facilitate peptide-based therapeutics. The study highlights the importance of capturing distinct features within each bacterial group and the effectiveness of machine learning models in predicting ABPs.