Deep-ABPpred: identifying antibacterial peptides in protein sequences using bidirectional LSTM with word2vec

Abstract

The overuse of antibiotics has led to emergence of antimicrobial resistance, and as a result, antibacterial peptides (ABPs) are receiving significant attention as an alternative. Identification of effective ABPs in lab from natural sources is a cost-intensive and time-consuming process. Therefore, there is a need for the development of in silico models, which can identify novel ABPs in protein sequences for chemical synthesis and testing. In this study, we propose a deep learning classifier named Deep-ABPpred that can identify ABPs in protein sequences. We developed Deep-ABPpred using bidirectional long short-term memory algorithm with amino acid level features from word2vec. The results show that Deep-ABPpred outperforms other state-of-the-art ABP classifiers on both test and independent datasets. Our proposed model achieved the precision of approximately 97 and 94% on test dataset and independent dataset, respectively. The high precision suggests applicability of Deep-ABPpred in proposing novel ABPs for synthesis and experimentation. By utilizing Deep-ABPpred, we identified ABPs in the tail protein sequences of Streptococcus bacteriophages, chemically synthesized identified peptides in lab and tested their activity in vitro. These ABPs showed potent antibacterial activity against selected Gram-positive and Gram-negative bacteria, which confirms the capability of Deep-ABPpred in identifying novel ABPs in protein sequences. Based on the proposed approach, an online prediction server is also developed, which is freely accessible at https://abppred.anvil.app/. This web server takes the protein sequence as input and provides ABPs with high probability (>0.95) as output.

AMPlify: attentive deep learning model for discovery of novel antimicrobial peptides effective against WHO priority pathogens

“…The Deep-AmPEP30 online server applies a convolutional neural network (CNN) for AMP prediction [ 22 ], though the tool is restricted to working with short peptides up to 30 amino acids (aa) in length. The Deep-ABPpred online server adopts Bi-LSTM with word2vec [ 23 ], also for short (≤ 30 aa) peptides [ 24 ]. The Bi-LSTM model from Wang and co-workers is designed for even shorter peptides (≤ 20 aa) and specializes to predicting AMPs against Escherichia coli [ 25 ].…”

PrMFTP: Multi-functional therapeutic peptides prediction based on multi-head self-attention mechanism and class weight optimization

“…Recently, with the development of artificial intelligence technology, the importance and advantage of deep learning (DL) methods in the field of bioinformatics have been well demonstrated [16][17][18][19][20][21]. Various DL methods have been utilized for therapeutic peptides prediction [22][23][24][25], such as Fang et al proposed a predictor based on DL combined with a character embedding layer for anti-fungal peptides identification [24], Li et al developed a dual-channel deep neural network (DNN) model for identifying variable-length antiviral peptides [25]. Compared with the traditional ML methods that need to extract or select features manually, the DL models can automatically learn the feature representation with limited peptide knowledge [26].…”

GM-Pep: A High Efficiency Strategy to De Novo Design Functional Peptide Sequences

“…For instance, artificial intelligence-based (AI-based) methods are used to predict potential activities of peptides, and the results of the wet experiment indicated that this technology is reliable for drug research. − Based on this purpose, AI-based methods are grouped into two categories. One is a classifier based on the binary model, such as ATSE, Deep-ABPpred, mAHTPred, and Antifp . Due to limited amount of data, most machine learning-based (ML-based) methods adopted a support vector machine (SVM) or random forest (RF) algorithm by the use of feature abstraction technologies such as the composition feature, position feature, physicochemical properties, and so forth.…”