DeORFanizing Candida albicans Genes using Coexpression

Abstract

Functional characterization of open reading frames in nonmodel organisms, such as the common opportunistic fungal pathogen Candida albicans, can be labor-intensive. To meet this challenge, we built a comprehensive and unbiased coexpression network for C. albicans, which we call CalCEN, from data collected from 853 RNA sequencing runs from 18 large-scale studies deposited in the NCBI Sequence Read Archive. Retrospectively, CalCEN is highly predictive of known gene function annotations and can be synergistically combined with sequence similarity and interaction networks in Saccharomyces cerevisiae through orthology for additional accuracy in gene function prediction. To prospectively demonstrate the utility of the coexpression network in C. albicans, we predicted the function of underannotated open reading frames (ORFs) and identified CCJ1 as a novel cell cycle regulator in C. albicans. This study provides a tool for future systems biology analyses of gene function in C. albicans. We provide a computational pipeline for building and analyzing the coexpression network and CalCEN itself at http://github.com/momeara/CalCEN. IMPORTANCE Candida albicans is a common and deadly fungal pathogen of humans, yet the genome of this organism contains many genes of unknown function. By determining gene function, we can help identify essential genes, new virulence factors, or new regulators of drug resistance, and thereby give new targets for antifungal development. Here, we use information from large-scale RNA sequencing (RNAseq) studies and generate a C. albicans coexpression network (CalCEN) that is robust and able to predict gene function. We demonstrate the utility of this network in both retrospective and prospective testing and use CalCEN to predict a role for C4_06590W/CCJ1 in cell cycle. This tool will allow for a better characterization of underannotated genes in pathogenic yeasts.

Leveraging machine learning essentiality predictions and chemogenomic interactions to identify antifungal targets

“…In order to generate a comprehensive assessment of genes with evidence for essentiality in C. albicans , we first collected a set of functional genomic features for use in training a machine-learning model (Supplementary Data 1 ). These included features derived from a collection of gene expression datasets such as gene expression level (Transcripts Per Kilobase Million (TPM) median), gene expression variance, and degree of co-expression (number of partners in a co-expression network) 22 . It also encompassed a codon adaptation index (CAI), which measures the bias in codon usage across each gene 21 ; the number of SNPs per nucleotide for each gene across a set of sequenced C. albicans strains 23 ; the presence of an essential ortholog in S. cerevisiae 12 ; and the presence of a duplicated set of paralogs in S. cerevisiae that exhibited a synthetic sick/lethal genetic interaction 24 .…”

“…We utilized co-expression clustering analysis to determine whether any biological processes are enriched among essential genes in the C. albicans genome. Using a non-linear dimensionality reduction, we clustered C. albicans genes based on co-expression 22 (Fig. 2c ) and used GO term enrichment of the genes in each cluster to assign a putative function.…”

“…The function of C6_03200W could not be predicted using BLASTp or conserved domain searches. Thus, we leveraged the C. albicans co-expression dataset used in our machine-learning model, as co-expression partners are predicted to have related functions 22 . C6_03200W co-clustered with mitochondrial genes, especially those involved in mitochondrial translation (Fig.…”

“…To further define the function of fungal-specific essential genes that lack a S. cerevisiae homolog, we next turned to EMF1 . Leveraging co-expression datasets 22 , we predicted it to have a role in the mitochondria. Our discovery that Emf1 localizes to the C. albicans mitochondria and is important for maintaining mitochondrial integrity illustrates the power of co-expression partners in predicting gene function, a concept that has been employed for other species 67 – 70 .…”

Inactivating the mannose-ethanolamine phosphotransferase Gpi7 confers caspofungin resistance in the human fungal pathogen Candida albicans

“…6 C), such as CSH1 , IFD6 , and ORF19.7214 ( Table S4 ). CSH1 and IFD6 have been predicted to participate in protein hydrolysis during cell-wall deconstruction ( O’Meara and O’Meara, 2021 ). ORF19.7214 is an uncharacterized gene in C. albicans , but its homologue in S. pombe , Exg3, encodes a glucan 1,6-β-glucosidase involved in cell-wall decomposition ( Dueñas-Santero et al, 2010 ).…”