Clustering analysis of large-scale phenotypic data in the model filamentous fungus Neurospora crassa

Abstract

Abstract Background With 9730 protein-coding genes and a nearly complete gene knockout strain collection, Neurospora crassa is a major model organism for filamentous fungi. Despite this abundance of information, the phenotypes of these gene knockout mutants have not been categorized to determine whether there are broad correlations between phenotype and any genetic features. Results Here, we analyze data for 10 different growth or developmental phenotypes that have been obtained for 1168 N. crassa knockout mutants. Of these mutants, 265 (23%) are in the normal range, while 903 (77%) possess at least one mutant phenotype. With the exception of unclassified functions, the distribution of functional categories for genes in the mutant dataset mirrors that of the N. crassa genome. In contrast, most genes do not possess a yeast ortholog, suggesting that our analysis will reveal functions that are not conserved in Saccharomyces cerevisiae. To leverage the phenotypic data to identify pathways, we used weighted Partitioning Around Medoids (PAM) approach with 40 clusters. We found that genes encoding metabolic, transmembrane and protein phosphorylation-related genes are concentrated in subsets of clusters. Results from K-Means clustering of transcriptomic datasets showed that most phenotypic clusters contain multiple expression profiles, suggesting that co-expression is not generally observed for genes with shared phenotypes. Analysis of yeast orthologs of genes that co-clustered in MAPK signaling cascades revealed potential networks of interacting proteins in N. crassa. Conclusions Our results demonstrate that clustering analysis of phenotypes is a promising tool for generating new hypotheses regarding involvement of genes in cellular pathways in N. crassa. Furthermore, information about gene clusters identified in N. crassa should be applicable to other filamentous fungi, including saprobes and pathogens.