Computational approach to predict species-specific type III secretion system (T3SS) effectors using single and multiple genomes

Abstract

BackgroundMany gram-negative bacteria use type III secretion systems (T3SSs) to translocate effector proteins into host cells. T3SS effectors can give some bacteria a competitive edge over others within the same environment and can help bacteria to invade the host cells and allow them to multiply rapidly within the host. Therefore, developing efficient methods to identify effectors scattered in bacterial genomes can lead to a better understanding of host-pathogen interactions and ultimately to important medical and biotechnological applications.ResultsWe used 21 genomic and proteomic attributes to create a precise and reliable T3SS effector prediction method called Genome Search for Effectors Tool (GenSET). Five machine learning algorithms were trained on effectors selected from different organisms and a trained (voting) algorithm was then applied to identify other effectors present in the genome testing sets from the same (GenSET Phase 1) or different (GenSET Phase 2) organism. Although a select group of attributes that included the codon adaptation index, probability of expression in inclusion bodies, N-terminal disorder, and G + C content (filtered) were better at discriminating between positive and negative sets, algorithm performance was better when all 21 attributes (unfiltered) were used. Performance scores (sensitivity, specificity and area under the curve) from GenSET Phase 1 were better than those reported for six published methods. More importantly, GenSET Phase 1 ranked more known effectors (70.3%) in the top 40 ranked proteins and predicted 10–80% more effectors than three available programs in three of the four organisms tested. GenSET Phase 2 predicted 43.8% effectors in the top 40 ranked proteins when tested on four related or unrelated organisms. The lower prediction rates from GenSET Phase 2 may be due to the presence of different translocation signals in effectors from different T3SS families.ConclusionsThe species-specific GenSET Phase 1 method offers an alternative approach to T3SS effector prediction that can be used with other published programs to improve effector predictions. Additionally, our approach can be applied to predict effectors of other secretion systems as long as these effectors have translocation signals embedded in their sequences.Electronic supplementary materialThe online version of this article (doi:10.1186/s12864-016-3363-1) contains supplementary material, which is available to authorized users.

Edwardsiella piscicida: A versatile emerging pathogen of fish

“…However, it is possible to create a reliable machine learning method that is species or genus specific. Hobbs et al [83], used 21 attributes to create a reliable machine learning method called, Genome Search for Effectors Tool (GenSET) to predict T3SS effectors. Known effectors and non-effector sequences from one genome were used to train five machine learning algorithms.…”

“…An averaging algorithm was then applied to predict known and unknown effectors in a testing set. The GenSET program was speciesspecific, gave better performance, and successfully predicted effectors in four known genomes including S. Typhimurium and E. coli [83]. A similar approach can be applied to Edwardsiella strains to predict T3SS and T6SS effectors in the future.…”

Advances and Challenges in Studying Type III Secretion Effectors of Attaching and Effacing Pathogens

“…Several resources have been developed to identify new effectors. Effector-encoding genes can be predicted in silico to varying degrees of accuracy (McDermott et al, 2011 ; Hobbs et al, 2016 ; Xue et al, 2019 ). These algorithms harness experimental knowledge of typical type III effector features, such as N-terminal enrichment of small polar amino acids (e.g., serine and threonine; Arnold et al, 2009 ), conservation of regulatory motifs upstream of the gene, a differing GC content to the rest of the genome, lack of gene homology to non-T3SS-encoding strains, and gene proximity to known effectors (Teper et al, 2016 ).…”

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

“…Over the past decade, dozens of machine learning-based computational approaches have been proposed to identify different types of secreted effectors ( Zeng and Zou, 2019 ), including support vector machine (SVM) ( Samudrala et al, 2009 ; Yang et al, 2010 ; Wang et al, 2011 , 2014 , 2017 ; Dong et al, 2013 ; Zou et al, 2013 ; Goldberg et al, 2016 ; Esna Ashari et al, 2019a , b ), random forest (RF) ( Yang et al, 2013 ), artificial neural network (ANN) ( Löwer and Schneider, 2009 ), naive Bayes (NB) ( Arnold et al, 2009 ), hidden Markov model (HMM) ( Xu et al, 2010 ; Lifshitz et al, 2013 ; Wang et al, 2013 ), logistic regression (LR) ( Esna Ashari et al, 2018 ), decision tree (DT) ( Wang et al, 2019a ), gradient boosting ( Chen et al, 2020 ), deep learning (DL) ( Xue et al, 2018 , 2019 ; Açıcı et al, 2019 ; Fu and Yang, 2019 ; Hong et al, 2020 ; Li et al, 2020a ), and their ensemble methods ( Burstein et al, 2009 ; Hobbs et al, 2016 ; Wang et al, 2018 , 2019b ; Xiong et al, 2018 ; Li et al, 2020b ). Some of these methods have achieved relatively high predictive accuracy, while they can recognize only one type of secreted effector, such as SIEVE ( Samudrala et al, 2009 ), EffectiveT3 ( Arnold et al, 2009 ), T3_MM ( Wang et al, 2013 ), GenSET ( Hobbs et al, 2016 ), Bastion3 ( Wang et al, 2019a ), DeepT3 ( Xue et al, 2019 ), WEDeepT3 ( Fu and Yang, 2019 ), ACNNT3 ( Li et al, 2020a ), and EP3 ( Li et al, 2020b ) for T3SEs; T4EffPred ( Zou et al, 2013 ), T4SEpre ( Wang et al, 2014 ), DeepT4 ( Xue et al, 2018 ), PredT4SE-Stack ( Xiong et al, 2018 ), Bastion4 ( Wang et al, 2019b ), T4SE-XGB ( Chen et al, 2020 ), and CNN-T4SE ( Hong et al, 2020 ) for T4SEs; and Bastion6 ( Wang et al, 2018 ) for T6SEs. It is important to n...…”