curation.rmd

---
title: "Data curation"
output: html_notebook
---

Now where we have the assemblies and GTDB taxonomy, we will perform a bit of manual curation to end up with a high quality dataset. First let's read in the GTDB classifications and clean them up a bit. 

```{r}
library(data.table)

classified <- fread("data/gtdb/assemblies.bac120.summary.tsv")
bacteria <- classified[, .(
  assembly = user_genome,
  classification,
  fastani_ani,
  fastani_af,
  closest_placement_ani,
  closest_placement_af,
  msa_percent,
  classification_method
)]
bacteria[, "species" := gsub("_[A-Z]", "", tstrsplit(classification, ";s__")[[2]])]
bacteria
```
There are 18 genomes that could not get aligned.
Let's also remove two genomes without species assignment and *Bacillus anthracis* which we will assume is a misclassification (nobody had anthrax).

```{r}
bacteria <- bacteria[!is.na(species) & species != "Bacillus anthracis"]
nrow(bacteria)
```

Okay this is our clean list that we will use. 

```{r}
fwrite(bacteria, "data/curated_assemblies.csv")
```

## Species distribution

```{r, fig.width=4, fig.height=6}
library(ggplot2)
theme_minimal() |> theme_set()

bacteria[, "genus" := tstrsplit(species, " ")[[1]]]
counts <- bacteria[, .N, by="genus"]
ggplot(counts, aes(x=N, y=reorder(genus, N))) + 
  geom_vline(xintercept=c(0.01, 0.05) * sum(counts$N), linetype="dashed") +
  geom_bar(stat="identity") +
  labs(y="", x="#assemblies")
ggsave("figures/genus_counts.png", width=4, height=6)
```