precision-recall.rmd

---
title: Calculate precision and recall of Tigmint
author: Shaun Jackman
output:
  html_document:
    keep_md: true
params:
  breakpoints_count_tsv:
    label: "Input TSV file of parameters and metrics"
    value: "breakpoints.count.tsv"
    input: text
  samtobreak_tsv:
    label: "Input TSV file of parameters and metrics"
    value: "samtobreak.tsv"
    input: text
  output_tsv:
    label: "Output TSV file of precision, recall, and G-score"
    value: "precision-recall.out.tsv"
    input: text
---

```{r setup, message=FALSE}
library(dplyr)
library(ggplot2)
library(ggrepel)
library(knitr)
library(readr)
library(tidyr)

knit_print.data.frame <- function(x, ...) kable(x) %>% paste(collapse = "\n") %>% asis_output
breakpoints_count_tsv <- params$breakpoints_count_tsv
samtobreak_tsv <- params$samtobreak_tsv
output_tsv <- params$output_tsv
```

# Read the data

```{r read-data}
breakpoints_count <- read_tsv(breakpoints_count_tsv)

samtobreak_orig <- read_tsv(samtobreak_tsv)
samtobreak <- samtobreak_orig %>% drop_na()
Positives <- max(samtobreak_orig$Total_breakpoints)

metrics <- left_join(breakpoints_count, samtobreak, by = c("Depth", "Starts")) %>%
	transmute(Depth, Starts, PP, FN = Total_breakpoints, P = Positives)
```

# Compute the precision, recall, and F1 score.
```{r compute-precision-recall}
precision_recall <- metrics %>%
	mutate(
		Label = paste("Clipped =", Starts),
		TP = P - FN,
		FP = PP - TP,
		PPV = TP / PP,
		TPR = TP / P,
		F1 = 2 * PPV * TPR / (PPV + TPR),
		G = sqrt(PPV * TPR))
```

# Precision vs recall scatter plot
```{r precision-recall}
ggplot(precision_recall) +
	aes(x = TPR, y = PPV, label = Label) +
	geom_point() +
	geom_text_repel() +
	scale_x_continuous(name = "Recall (TPR)", limits = c(0, 0.15)) +
	scale_y_continuous(name = "Precision (PPV)", limits = c(0, 1)) +
	theme_bw(base_size = 20) +
	theme(panel.border = element_blank())
```

# Precision vs recall table
```{r precision-recall-table}
precision_recall %>% select(-Label) %>% write_tsv(output_tsv)
precision_recall
```