DeepT3_4: A Hybrid Deep Neural Network Model for the Distinction Between Bacterial Type III and IV Secreted Effectors

Abstract

Gram-negative bacteria can deliver secreted proteins (also known as secreted effectors) directly into host cells through type III secretion system (T3SS), type IV secretion system (T4SS), and type VI secretion system (T6SS) and cause various diseases. These secreted effectors are heavily involved in the interactions between bacteria and host cells, so their identification is crucial for the discovery and development of novel anti-bacterial drugs. It is currently challenging to accurately distinguish type III secreted effectors (T3SEs) and type IV secreted effectors (T4SEs) because neither T3SEs nor T4SEs contain N-terminal signal peptides, and some of these effectors have similar evolutionary conserved profiles and sequence motifs. To address this challenge, we develop a deep learning (DL) approach called DeepT3_4 to correctly classify T3SEs and T4SEs. We generate amino-acid character dictionary and sequence-based features extracted from effector proteins and subsequently implement these features into a hybrid model that integrates recurrent neural networks (RNNs) and deep neural networks (DNNs). After training the model, the hybrid neural network classifies secreted effectors into two different classes with an accuracy, F-value, and recall of over 80.0%. Our approach stands for the first DL approach for the classification of T3SEs and T4SEs, providing a promising supplementary tool for further secretome studies.

layerUMAP: A tool for visualizing and understanding deep learning models in biological sequence classification using UMAP

“…We successfully applied this tool in our previous studies. 24 , 25 To further demonstrate the capabilities of layerUMAP, we presented two case studies in this article. We expect layerUMAP to provide insight into the mechanisms of DL models and to become a standard tool for the visual analysis of DL models.…”

The applications of deep learning algorithms on in silico druggable proteins identification

“…As the most popular hybrid deep learning methods, CNN-RNNs, which are a kind of quasi-recurrent neural network (QRNN) [71] , are proposed to integrate the advantages of CNNs and RNNs and have been successfully applied to various biological applications [72] , [73] , [74] . The hybrid CNN-RNN leverages the feature extraction capability of CNN and connects the obtained feature representations to an RNN to capture sequential dependencies between the input data.…”

A Transfer-Learning-Based Deep Convolutional Neural Network for Predicting Leukemia-Related Phosphorylation Sites from Protein Primary Sequences

“…But machine-learning methods cannot directly recognize sequences as input, so we used dictionary encoding to ensure the uniformity of sequence features. Each residue in the phosphorylation peptides is represented by an ordinal number, in which each of the 20 basic amino acids is assigned a number from 1 to 20 [44]. Thus, each peptide is represented by a one-letter code and transformed into an L-dimensional vector, where L is the length of the peptides.…”