---
title: "cpgdirectionData: the data layers behind cpgdirection"
author: "Tsachi Ein-Dor"
output: rmarkdown::html_vignette
vignette: >
  %\VignetteIndexEntry{cpgdirectionData: the data layers behind cpgdirection}
  %\VignetteEngine{knitr::rmarkdown}
  %\VignetteEncoding{UTF-8}
---

Bioconductor software packages are capped at 5 MB per file. Eight of the
data layers behind the `cpgdirection` package are larger than that, so they
live here, on ExperimentHub, and are fetched on first use -- then cached
locally, so the download cost is paid once per machine.

Most users never call this package directly: `cpgdirection` resolves every
layer through its data backend, using a bundled copy when one is installed
and falling back to these Hub resources otherwise.

## What is served

```{r resources}
library(cpgdirectionData)
cpgdData_resources()[, c("name", "genome")]
```

Direct retrieval from the Hub (cached after the first download):

```{r hub-mode, eval = FALSE}
ann <- getEpicv2ProbeGeneAnnotation()
head(ann)
```

## Local mode

Air-gapped machines, containers without network access, and analyses that
must pin an exact data snapshot can serve the same files from a local
directory instead of the Hub. The chunk below demonstrates the mechanism
with a stand-in file, so it runs anywhere:

```{r local-mode}
d <- tempfile()
dir.create(d)
saveRDS(data.frame(cpg_id = "cg00000029", chr = "chr16", pos = 53434200L),
        file.path(d, "cpg_positions_hg19.rds"))
old <- options(cpgdirectionData.local = d)
head(getCpgPositionsHg19())
options(old)
```

## Provenance

Every resource is built by `inst/scripts/make-data.R` from the full
cpgdirection build. The resources themselves are archived on Zenodo
(DOI [10.5281/zenodo.22225164](https://doi.org/10.5281/zenodo.22225164)) and retrieved from there by ExperimentHub. The multi-source EPIC v2 probe-to-gene annotation is
the union of five independent tracks (Illumina EPIC-8v2-0_A2 manifest UCSC
RefGene and GencodeV41; the Zhou lab SeSAMe GENCODE v41 annotation; the
Exeter GENCODE 47 re-annotation, Zenodo record 15181885, including its
regulatory-element assignments), with per-track support flags so downstream
consumers can weight assignments by agreement.

## Session info

```{r session-info}
sessionInfo()
```