prepare_metadata.Rmd

---
title: "Format input files"
output: html_document
---

```{r setup, include=FALSE}
knitr::opts_chunk$set(echo = TRUE)
```


## Make metadata file

```{r}
library(rgeos)
library(sp)
library(sf)
library(rgdal)
library(tidyverse)
library(ggplot2)
library(viridis)
library(ggmap)
library(geonames)
library(metacoder)
metadata <- read_csv('data/Pram_MetaData.csv')
```

## Reformat metadata

```{r}
metadata <- metadata %>%
  rename(strain = isolate_id_orig,
         isolate_id = "Isolate_ID(GL)",
         gps_coord = "GPS Coordinates")
```

Add "source" column as a hybrid of host and environment

```{r}
metadata <- metadata %>% 
  mutate(source = paste(ifelse(is.na(Host_genus), "", Host_genus), ifelse(is.na(Host_species), "", Host_species)),
         source = trimws(source),
         source = ifelse(source == "", Host_Environment, source)) 
```

fix date format

```{r}
metadata$date <- paste0(metadata$Year, '-01-01')
```

Clean up place names so it is easier to look up coordinates

```{r}
replace_key <- c(
  "N.Ireland" = "Northern Ireland"
)
metadata$Country[metadata$Country %in% names(replace_key)] <- replace_key[metadata$Country[metadata$Country %in% names(replace_key)]]
```

save metadata file

```{r}
write_tsv(metadata, file = file.path('data', 'metadata_modified.tsv'))
```


## Make coordinate file

```{r}
# https://gis.stackexchange.com/questions/240018/quickly-assign-coordinates-to-region-names-using-r
options(geonamesUsername = "fosterz")
get_coords <- function(name, group, ...) {
  res <- GNsearch(name_equals = name, ...)  
  if ("fcode" %in% colnames(res)) {
    res <- filter(res, name == name, fcode %in% c("AREA", "PCLI", "ADM1"))
  }
  res <- res[1, ]
  out <- tibble(group = group, name = name, lat = res$lat, lon = res$lng)
  return(out)
}

coord_data <- bind_rows(
  map_dfr(unique(metadata$Country, na.rm = TRUE), get_coords, group = 'Country'),
  map_dfr(unique(metadata$State), get_coords, group = 'State', country = 'USA', fcode = "ADM1")
) %>%
  filter(!is.na(name))

write_tsv(coord_data, file = file.path("config", "lat_longs.tsv"), col_names = FALSE)
```


## Make dropped strains file

These will be filtered out of the analysis.
First, I will remove the reference strain, since that gets assigned a date in 1980 for some reason and that is confusing:

```{r}
dropped_ids <- c('MT_DQ832718.1')
```

And all IDs associated with a missing location or time:

```{r}
dropped_ids <- c(dropped_ids, metadata$strain[is.na(metadata$Country)])
dropped_ids <- c(dropped_ids, metadata$strain[is.na(metadata$Year)])
```

Ignore isolates with no FASTA sequence:

```{r}
seqs <- read_fasta("data/global_n661_mt.fasta")
dropped_ids <- c(dropped_ids, metadata$strain[! metadata$strain %in% names(seqs)])
```

And write the file, one ID per line:

```{r}
write_lines(dropped_ids, file = file.path("config", "dropped_strains.txt"))
```


## Make color file

```{r}
color_data <- coord_data %>%
  select(group, name) %>%
  mutate(color = viridis(length(group)))
color_data$color <- substr(color_data$color, start = 1, stop = 7)
write_tsv(color_data, file = file.path("config", "colors.tsv"), col_names = FALSE)
```