EffectorO: motif-independent prediction of effectors in oomycete genomes using machine learning and lineage-specificity

Abstract

Oomycete plant pathogens cause a wide variety of diseases, including late blight of potato, sudden oak death, and downy mildew of many plants. These pathogens are major contributors to losses in many food crops. Oomycetes secrete "effector" proteins to manipulate their hosts to the advantage of the pathogen. Plants have evolved to recognize effectors, resulting in an evolutionary cycle of defense and counter-defense in plant-microbe interactions. This selective pressure results in highly diverse effector sequences that can be difficult to computationally identify using sequence similarity. We developed a pipeline, EffectorO, that uses two complementary approaches to predict effectors in oomycete pathogen genomes: (1) a machine learning-based pipeline that predicts effector probability based on the biochemical properties of the N-terminal amino acid sequence of a protein and is trained on experimentally verified oomycete effectors and (2) a pipeline based on lineage-specificity to find proteins that are unique to one species or genus, a sign of evolutionary divergence due to adaptation to the host. We tested EffectorO on Bremia lactucae, which causes lettuce downy mildew, and Phytophthora infestans, which causes late blight of potato and tomato, and predicted many novel effector candidates, while still recovering the majority of known effector candidates. EffectorO will be useful for discovering novel families of oomycete effectors without relying on sequence similarity to known effectors.

EffectorP 3.0: prediction of apoplastic and cytoplasmic effectors in fungi and oomycetes

“…The same holds for EffectorP 2.0, which was not trained on oomycete effectors and only predicts 43.8% of oomycete effectors correctly. The oomycete effector prediction tool EffectorO (Nur et al , 2021) also predicts a high proportion of the oomycete validation set correctly (14 of 16 effectors, 87.5%). Taken together, EffectorP 3.0 is a more sensitive tool for effector prediction in both fungi and oomycetes than EffectorP 2.0 or deepredeff.…”

“…EffectorP is such a machine learning prediction tool for fungal effector prediction (Sperschneider et al, 2016(Sperschneider et al, , 2018. However, current effector prediction methods such as EffectorP 2.0 (Sperschneider et al, 2018a), deepredeff (Kristianingsih & MacLean, 2021) or EffectorO (Nur et al, 2021) give a yes-or-no answer as to whether a protein is a likely effector and do not deliver vital information for prioritization: Is the effector extracellular and functions in the plant apoplast or is it cytoplasmic and enters plant cells? In oomycetes, the presence of conserved sequence motifs such as the RxLR motif have been used extensively for cytoplasmic effector prediction.…”

“…A deep learning method called deepredeff can predict effectors in bacteria, fungi and oomycetes (Kristianingsih & MacLean, 2021), but does not return effector localization. EffectorO is a machine learning method for effector prediction trained on the N-terminus of effector sequences, but it is applicable to oomycetes only (Nur et al, 2021).…”

“…A deep learning method called deepredeff can predict effectors in bacteria, fungi and oomycetes (Kristianingsih & MacLean, 2021), but does not return effector localization. EffectorO is a machine learning method for effector prediction trained on the N-terminus of effector sequences, but it is applicable to oomycetes only (Nur et al , 2021). Here, we introduce EffectorP 3.0, the first method that can predict if a secreted fungal or oomycete protein is one of these three types: an apoplastic effector, a cytoplasmic effector or a non-effector.…”

Predicting Lifestyle from Positive Selection Data and Genome Properties in Oomycetes

“…We validated proteins discussed in the manuscript through comparison with matches from the NCBI datatabase using BLAST [ 33 ]. We annotated the effectors in the stramenopile dataset by predicting the secretion signal using the tool SignalP 5.0b followed by an annotation with the model EffectorO [ 34 , 35 ]. We annotated the presence/absence of functional annotations from each genome with the Genome Properties database, performed the clustering with the Python library SciPy and visualized it with the package Seaborn [ 36 , 37 ].…”