Fig 1.
(a) Mean and standard deviation, along pseudotime, are plotted in log-log scales for each gene. For illustrative purposes, solid lines correspond to the exponential distribution (β = 1 and log(V) = 0), and dashed lines to the Poisson distribution (β = 0.5 and log(V) = –1.7), where log(V) was chosen to be –1.7 for better visualization. The RSI of each gene is color coded. 3207, 4703 and 4433 cells; and 7978, 7567 and 8333 genes were present in each matrix for hBECs, colon and ileum cells, respectively. (b) Distributions of genes in hBECs, as a function of TPT counts, that are to the left of the breakpoint (Poisson, left panel), near the breakpoint (transition, central panel) and to the right of the breakpoint (exponential, right panel), intended to illustrate the transition Poisson to exponential distribution.
Table 1.
Parameters of the biphasic fit to Taylor’s law for infected cells.
Matrix sparsity (proportion of zeros), number of genes and number of cells are also shown.
Fig 2.
Evolution of Taylor’s law parameters along the course of infection.
(a) Taylor’s parameters, V and β, estimated in pseudotime intervals (bins) for infected cells and for the simulated datasets. (b) Taylor plots for the points in panel (a). Mitochondrial genes are shown in green. Selected nuclear genes in hBECs that generally followed an exponential distribution regardless of rank are shown in orange. Red lines correspond to an exponential distribution (β = 1 and log(V) = 0), and blue lines to a Poisson distribution (β = 0.5 and log(V) = –1.7), where log(V) = –1.7 was chosen for better visualization.
Table 2.
P-values from ANCOVA analysis of Taylor’s parameters for cells infected with SARS-CoV-2 and the simulated dataset.
Partial η2 values are shown in parenthesis (partial η2 ≥ 0.15 are conventionally taken as large effects).
Fig 3.
(a) Comparison of RSI with PSI of all genes for hBECs, colon and ileum cells. Each point represents one gene. Genes with significant punctual stability (false discovery rate FDR < 0.05) are shown in red. (b) Top 50 enriched GO terms, ranked by P-value, for genes that displayed signal of punctual rank stability in all three cell types. The median RSI of the genes in each GO term is shown for each cell type.
Fig 4.
Estimation of H exponents from rank data.
(a) Distribution of H exponents estimated from gene rank data from infected cells and control datasets for hBECs, colon and ileum cells. (b) Taylor’s law plots showing the H value (color coded) of each gene for each cell type. Solid lines correspond to the exponential distribution (β = 1 and log(V) = 0), and dashed lines to the Poisson distribution (β = 0.5 and log(V) = –1.7, chosen for better visualization). (c) Plots of the H exponents calculated from infected cells vs the mean expression of genes present in the Taylor’s plots in panel (b). A dashed red line shows indicates the value H = 0.7 (Hurst phenomenon). H > 0.7 corresponds to strong persistent behavior. A solid red line indicates H = 0.5. Remind that H > 0.5 corresponds to persistent behavior.
Fig 5.
Functional analysis of genes exhibiting strong persistent behavior.
(a) Comparison of the H exponents from infected and simulated data for hBECs, colon and ileum cells. Each point represents one gene, which are colored depending on whether they show anti-persistent (H < 0.5, black) or persistent (0.7 > H > 0.5, gray) behavior or if Hurst phenomenon (H > 0.7, red) is seen. The dashed lines are bisecting lines. Black lines represent kernel density of the data. (b) Top 100 enriched GO terms, ranked by P-value, for genes exhibiting evidence of persistent rank behavior in all three cell types. The median H of the genes in each GO term is shown for each cell type.