Article Figures & Data
Figures
- FIG 1
Generating a coexpression network for C. albicans. (A) A gene-by-environment heatmap generated from collected C. albicans RNAseq experiments from the SRA. The C. albicans genes are on the y axis, and conditions are on the x axis. (B) A gene-by-gene heatmap generated from Spearman rank correlation. (C) UMAP embedding reveals functional clusters. Annotations were determined by GO term enrichment of genes in each cluster.
- FIG 2
CalCEN provides a robust orthogonal approach to identifying gene function. (A) UpSet plot of network overlap. The BlastP network has a threshold at an E value of <1e−5, CalCEN has a threshold at the top 1%. Each bar in the upper region shows the number of gene nodes in the intersection of the set of networks indicated by the rows with filled circles in the lower region. (B) For each annotated GO term colored by ontology biological process (BP), cellular component (CC), or molecular function (MF), the CalCEN neighbor voting guilt by association (GBA) area under the ROC curve (AUROC) is plotted as a function of the degree-null (genes predicted based on their network degree) AUROC. (C) Mean neighbor voting GBA performance for individual and combined networks, indicted by the rows with filled circles in the lower region, for subontology terms (colored dots), all terms (black) with error bars representing the standard errors of the mean sover 10-fold cross validation replicas. Degree-null predictive accuracy for each network is shown in gray. (D) Mean neighbor voting GBA performance for the CalCEN built over random subsets of RNAseq studies. The blue curve represents a mean of a nonparametric locally estimated scatterplot scattering (LOESS) fit with standard deviation in dark gray. As the number of studies increases, the performance increases.
- FIG 3
Retrospective analysis identifies functional clusters of genes in C. albicans. (A) Ribosomal proteins form a densely connected coexpression cluster. (B) Histone proteins are highly connected, except for the HHT1 variant histone protein. (C) Hsp90 is coexpressed with Hsp90 cochaperones, as well as multiple other functional classes of proteins. (D) Erg11 is coexpressed with other components of the ergosterol biosynthetic cascade. Nodes represent the genes, and the edge width corresponds to the degree of coexpression.
- FIG 4
GPI-anchored proteins in C. albicans show distinct coexpression clusters. Twenty-seven GPI-anchored proteins were used as seeds to generate coexpression clusters. Genes were included as coexpressed if they passed the top 2% cutoff.
- FIG 5
Identification of C4_06590W, or CCJ1, as a novel cell cycle regulator. (A) Ccj1 contains a J domain. TrRosetta was used to generate a de novo fold for the Ccj1 protein, and the structure was overlaid on the S. cerevisiae Sis1 (ScSis1) DnaJ protein crystal structure. (B) The coexpression network for Ccj1 identifies multiple cell cycle proteins. (C) Repression of CCJ1 with 5 μg/ml doxycycline (DOX) results in aberrant filamentation in the absence of an inducing cue. Nuclei were stained using Hoechst. WT, wild type; DIC, doxycycline. (D) Repression of CCJ1 with 5 μg/ml DOX results in hypersensitivity to 12.5 mM hydroxyurea (HU). Data are from two biological replicates, with two technical replicates each. Data were analyzed by t test; ** indicates P < 0.01. (E) Repression of Ccj1 results in hyperfilamentation in response to 12.5 mM hydroxyurea. Cells were incubated in YPD with 12.5 mM hydroxyurea and the presence or absence of 5 μg/ml of doxycycline overnight before imaging.
Supplemental Material
FIG S1
(A) A total of 1,399 RNAseq runs from 18 identified studies are scatter-plotted as the number of genes with nonzero expression versus the fraction transcripts that map exactly once. (B) The 853 runs that have nonzero expression for at least half of the genes (3,113), which are used to construct the CalCEN network, are shown in black. Download FIG S1, PDF file, 0.8 MB.
Copyright © 2021 O’Meara and O’Meara.
This content is distributed under the terms of the Creative Commons Attribution 4.0 International license.
TABLE S1
RNA-Seq studies for Candida albicans coexpression analysis. Download Table S1, XLSX file, 0.01 MB.
Copyright © 2021 O’Meara and O’Meara.
This content is distributed under the terms of the Creative Commons Attribution 4.0 International license.
TABLE S2
Coexpression clusters from UMAP. Download Table S2, XLSX file, 1.3 MB.
Copyright © 2021 O’Meara and O’Meara.
This content is distributed under the terms of the Creative Commons Attribution 4.0 International license.
FIG S2
For each annotated GO term colored by ontology biological process (BP), cellular component (CC), or molecular function (MF), the indicated network neighbor-voting guilt-by-association (GBA) area under the ROC curve (AUROC) is plotted as a function of the degree-null (genes predicted based on their network degree) AUROC. Download FIG S2, PDF file, 0.1 MB.
Copyright © 2021 O’Meara and O’Meara.
This content is distributed under the terms of the Creative Commons Attribution 4.0 International license.
FIG S3
The 853 RNAseq runs are embedded by reducing the 6,226 dimension gene expression profiles to 100 dimensions using principal component analysis and then to two dimensions using UMAP using min_dist = 0.5 and n_neighbors = 30. Runs were then labeled by their study accession and plotted using ggplot2. Download FIG S3, PDF file, 0.05 MB.
Copyright © 2021 O’Meara and O’Meara.
This content is distributed under the terms of the Creative Commons Attribution 4.0 International license.
