---
title: "Import to CAMERA from local data"
output: rmarkdown::html_vignette
vignette: >
  %\VignetteIndexEntry{Import to CAMERA from local data}
  %\VignetteEngine{knitr::rmarkdown}
  %\VignetteEncoding{UTF-8}
---

```{r, include = FALSE}
knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>",
  cache=TRUE
)
```

```{r}
library(CAMeRa)
```

You can supply to CAMERA from text files or from OpenGWAS (currently not a mixture of both, though you can download the studies in OpenGWAS to be used as raw files). 

## Generating the data manually

You need to supply the following data:

- `instrument_raw` = a data frame of pooled instruments across all ancestries, that has been extracted from each ancestry for the exposure traits. Optionally also provide the same for the outcome traits.
- `instrument_outcome` = instruments in `instrument_raw` extracted from the outcome datasets
- `instrument_regions` = named list of data frames of length number of unique instruments in `instrument_raw`. Names of each item are the instruments. Each item is a list of regional extracts around the instrument from each population exposure study.
- `instrument_outcome_regions` = as above but for the outcome datasets.

Examples of these datasets can be seen in the following file:

```{r}
load(system.file(package="CAMeRa", "extdata/example-local.rdata"))
instrument_raw
```

```{r}
instrument_outcome
```

```{r}
instrument_regions[[1]]
```

```{r}
instrument_outcome_regions[[1]]
```

For example scripts on how these data were generated see https://github.com/yoonsucho/CAMERA_analysis/tree/main/scripts/ldl_stroke_analysis

## Generating the data using `CAMERA_local`

We have developed a separate set of functions to organise data from text files to generate the data above. 

```{r}
metadata <- readRDS(system.file(package="CAMeRa", "extdata/example-metadata.rds"))
metadata
```

```{r}
ld_ref <- dplyr::tibble(
    pop = unique(metadata$pop),
    bfile = file.path("path/to/plink_files/", pop)
)
ld_ref
```

```{r, eval=FALSE}
localdata <- CAMERA_local$new(metadata = metadata, ld_ref = ld_ref, plink_bin = "path/to/plink")
localdata$organise()
```

This will read the files specified in the metadata and attempt to arrange the data as described above, generating `localdata$instrument_raw`, `localdata$instrument_outcome` etc.

You can then generate the `CAMERA` object e.g.

```{r}
l <- CAMERA$new()
l$import_from_local(
  instrument_raw=instrument_raw, 
  instrument_outcome=instrument_outcome, 
  instrument_regions=instrument_regions, 
  instrument_outcome_regions=instrument_outcome_regions, 
  exposure_ids=unique(instrument_raw$id), 
  outcome_ids=unique(names(instrument_outcome_regions[[1]])),
  pops=c("AFR", "EAS", "EUR", "AMR", "SAS")
)

l$instrument_heterogeneity()
```


```{r, eval=FALSE, echo=FALSE}
library(devtools)

load_all()
load("inst/extdata/example-local.rdata")

l <- CAMERA$new()
l$import_from_local(
  instrument_raw=instrument_raw, 
  instrument_outcome=instrument_outcome, 
  instrument_regions=instrument_regions, 
  instrument_outcome_regions=instrument_outcome_regions, 
  exposure_ids=unique(instrument_raw$id), 
  outcome_ids=unique(names(instrument_outcome_regions[[1]])),
  pops=c("AFR", "EAS", "EUR", "AMR", "SAS")
)

l$instrument_heterogeneity()
l$estimate_instrument_specificity(l$instrument_raw)
l$harmonise()

l$fema_regional_instruments()

l$cross_estimate()
l$plot_cross_estimate()

l$estimate_instrument_heterogeneity_per_variant()
```