Abstract Plant-pathogenic fungi cause crop disease via a range of secreted effector proteins that interact with specific receptors of host plant cells. Effector identification can enable the diagnosis of disease outcomes and enable selection or breeding of disease-resistant crop cultivars. Bioinformatic methods have been developed to predict proteins with ‘effector-like’ properties, but the resulting number of candidates tends to be larger than can be feasibly validated and may contain numerous false positives. Challenges to effector discovery include the obfuscating effects of genome-wide mutations common to Fungi, such as Repeat-Induced Point (RIP) mutations. Refining effector predictions by incorporating disease phenotyping into genome-wide association studies (GWAS) have had mixed success for a handful of pathogen species. But the utility of GWAS approaches may be limited by low ‘signal-to-noise’ caused by widespread RIP-like SNP mutations across the genomes of most fungal pathogens. This study presents an alternative method for effector candidate refinement called ‘EffectorFisher’. EffectorFisher extends the output of Predector – a tool that automates and combines results of several bioinformatic tools commonly used in effector discovery – to apply pangenome-derived protein-isoform profiling to remove candidate effector protein isoforms with weak association with virulent phenotypes. This method was benchmarked using corresponding pangenome and phenotype data for two model wheat pathogens, each with multiple known effectors: the necrotroph Parastagonospora nodorum and the hemibiotroph Zymoseptoria tritici . Compared to prior methods based on effector-like protein properties, EffectorFisher improved predicted rankings of known effectors and reduced the total number of effector candidates. We present EffectorFisher ( https://github.com/ccdmb/EffectorFisher-core ) as a useful tool for refining effector predictions with phenotype data, which is broadly applicable to many fungal pathogen species, and is capable of predicting effectors involved in both gene-for-gene and inverse gene-for-gene effector-receptor interactions.
Hossain et al. (Wed,) studied this question.