README

quantbayes provides a minimal Bayesian transform for evidence sufficiency from a binary matrix of zero and one entries. The method is simple, portable, and independent of any rule set.

Installation

# Install development version
remotes::install_github("switzerland-omics/quantbayes")

Data layout

head(core_test_data)

x <- as.matrix(core_test_data[, -1])
rownames(x) <- core_test_data[[1]]

Run quantbayes in one step

res <- quant_es_core(x)

res$global
head(res$variants)

Global contains posterior summaries. Variants contains per variant theta, credible intervals, and percentiles.

Default plots

plots <- quant_es_plots(res, x)

plots$p_global
plots$p_overlay
plots$p_matrix
plots$p_p_hat
plots$p_theta_ci
plots$p_combined

These cover density, overlay of top candidates, evidence matrix, observed proportions, credible intervals, and a combined panel.

Highlight variants of interest

highlight_demo <- list(
  list(id = rownames(x)[1], colour = "#ee4035", size = 4),
  list(id = rownames(x)[5], colour = "#2f4356", size = 4)
)

plots2 <- quant_es_plots(res, x, highlight_points = highlight_demo)
plots2$p_overlay

Custom palettes

pal10 <- colorRampPalette(c("black", "grey"))(10)
pal20 <- colorRampPalette(c("skyblue", "navy"))(20)

plots_custom <- quant_es_plots(
  res,
  x,
  palette10 = pal10,
  palette20 = pal20
)

plots_custom$p_overlay

Any plot returned by quant_es_plots is a standard ggplot, so users can layer themes or labels.

plots$p_overlay + ggplot2::theme_minimal()

File based input

tmp <- tempfile(fileext = ".tsv")
write.table(core_test_data, tmp, sep = "\t", quote = FALSE, row.names = FALSE)

res_file <- quant_es_from_binary_table(tmp)
res_file$global

Save plots

outdir <- "quantbayes_output"
if (!dir.exists(outdir)) dir.create(outdir)

ggplot2::ggsave(
  file.path(outdir, "overlay.png"),
  plots$p_overlay,
  width = 6,
  height = 4,
  dpi = 120
)

Save tables

write.csv(
  res$variants,
  file.path(outdir, "variants_results.csv"),
  row.names = FALSE
)

write.csv(
  as.data.frame(res$global),
  file.path(outdir, "global_summary.csv"),
  row.names = FALSE
)

Full workflow in six lines

data(core_test_data)

x <- as.matrix(core_test_data[, -1])
rownames(x) <- core_test_data[[1]]

res <- quant_es_core(x)
plots <- quant_es_plots(res, x)
plots$p_combined

Core results example

Clinical genetics example

After Whole Genome Sequencing, a proprietary candidate selection tool identified potential causal variants. A clinical laboratory requires verifiable evidence to support or refute these findings. Each candidate variant was evaluated using a minimal and independent evidence set that records whether supporting evidence is present or absent under a Qualifying Variant Evidence Standard.

res_df <- as.data.frame(res$variants)
global_df <- as.data.frame(res$global)

res_df <- res_df[order(res_df$theta_mean, decreasing = TRUE), ]

head(res_df)
head(global_df)

variant_id	k	m	theta_mean	theta_lower	theta_upper	percentile
2-54234474-G-A_AR	18	24	0.7308	0.5487	0.8793	99.875
6-72183475-CG-N_AR	18	24	0.7308	0.5487	0.8793	99.875
1-14682421-G-A_AD	17	24	0.6923	0.5061	0.8505	99.375
7-751853912-C-T_AR	17	24	0.6923	0.5061	0.8505	99.375
X-224319469-CT-C_XR	16	24	0.6538	0.4650	0.8203	97.000
X-414698664-CT-C_XD	16	24	0.6538	0.4650	0.8203	97.000

Scenario: autosomal recessive disease XXX with a primary candidate variant. Variant: 2-54234474-G-A_AR

These values reflect the relative strength of evidence for this variant within the tested panel.

Total evaluated variants: 400 Global evidence sufficiency: 0.52 Credible interval: 0.38 to 0.65. Across all variants, the mean theta is 0.52 and the median is 0.54.

quantbayes