为你的团伙写一个 R 包

class: center, middle, inverse, title-slide

.title[
# 为你的团伙写一个 R 包
]
.author[
### 高春辉 @ 资环 × 微重 × 作重
]
.date[
### 2023-10-20
]

---

## 关于自己

### Data Scientist

![](https://vnote-1251564393.cos.ap-chengdu.myqcloud.com/20231020095533.png)

参见：[我的简历](https://bio-spring.top/cv/)

---

### GitHub A- 级开发者

![](https://vnote-1251564393.cos.ap-chengdu.myqcloud.com/20231020095321.png)

去年是 A+ 级。

---

### GitHub 记录

-   一言不合，反手 PR

![](https://vnote-1251564393.cos.ap-chengdu.myqcloud.com/20231020094512.png)

---

-   我为大佬补漏洞：<https://github.com/rstudio/gt/pull/137>

![](https://vnote-1251564393.cos.ap-chengdu.myqcloud.com/20231020094655.png)

---

### 华农 No. 1

-   被**刘浩**老师誉为华农搞 R 最**专业** 的人

![](https://vnote-1251564393.cos.ap-chengdu.myqcloud.com/20231020095938.png)

---

## 从两位生信大佬说起

- [余光创（*Y叔*）](https://yulab-smu.top/about/)

- [陈程杰（*CJ*）](https://mp.weixin.qq.com/s/FVUWFRyFpVj2Kd_alKoeFw)

---

### 余光创

![](https://vnote-1251564393.cos.ap-chengdu.myqcloud.com/20231020100330.png)

[余光创（*Y叔*）](https://yulab-smu.top/about/)：南方医科大学生物信息学教授。

---

### 陈程杰

![](https://vnote-1251564393.cos.ap-chengdu.myqcloud.com/20231020100945.png)

[陈程杰（*CJ*）](https://mp.weixin.qq.com/s/FVUWFRyFpVj2Kd_alKoeFw)：TBtools 软件的开发者，年轻的生信专家。

---

## 我的开发体会：ggVennDiagram

###   需求的产生：画一个韦恩图

![](https://vnote-1251564393.cos.ap-chengdu.myqcloud.com/20231020102112.png)

---

###   任务的分解

-   区域（集合 + 子集），填充，标注

-   标签（集合 + 子集），文字

---

-   生成 4 sets椭圆
  
![](https://vnote-1251564393.cos.ap-chengdu.myqcloud.com/20231020102205.png)
    
---

-   获取元素的子集
    
`union()`，`intersect()`，`setdiff()`

-   获取图形的子集

`st_union()`、`st_intersect()`、`st_difference()`
    
![](https://vnote-1251564393.cos.ap-chengdu.myqcloud.com/20231020104228.png)

---

-   计算中心点坐标
    
```r
library(sf)
library(ggplot2)

nc <- st_read(system.file('shape/nc.shp', package = "sf"))

# using sf
sf_cent <- st_centroid(nc)

# plot both together to confirm that they are equivalent
ggplot() + 
  geom_sf(data = nc, fill = 'white') +
  geom_sf(data = sf_cent, color = 'red') 
```

---

-   计算中心点坐标

![](2023-hzau-write-r-pkg_files/figure-html/unnamed-chunk-1-1.png)

---

-   添加文字
    
```r
# 多边形
ggpolygons = list(A=A,B=B,C=C,D=D,AB=AB,AC=AC,AD=AD,BC=BC,BD=BD,CD=CD,ABC=ABC,ABD=ABD,ACD=ACD,BCD=BCD,ABCD=ABCD)

# 标签文字
polygon_names = names(ggpolygons)

# 标签位置
data_centers = ...

# 画图
ggplot(data_ploygons,aes(x,y,fill=group)) +
  geom_polygon(show.legend = F) +
  geom_text(aes(label=group),data=data_centers)
```

---

-   在正确的位置添加文字

![](https://vnote-1251564393.cos.ap-chengdu.myqcloud.com/20231020104538.png)

---

###   不得的取舍

开发一个画 Venn 图的包，它可以：

-   [√] 接受传统上的输出，输出优于 `VennDiagram` 和 `gplots` 的 Venn 图;

-   [√] 颜色/透明度/线型/粗细等各种自定义

-   [×] 是否固定比例：参数 `scaled = True`。

-   [√] 随意显示标签(label 和 text)

-   [×] 参数：circle 还是 ellipse？（circle在4d venn中并不能显示所有组合）

-   [√] 参数：color是不同色块还是按梯度填充？

-   [×] 接受直接的 `data.frame` 或者 `matrix` 输出, 通过指定 `Threshold` 自行创建存在矩阵, 按列分组, 按行计数;

-   [√] 也可以接受 `list` 输入；

-   [×] 如果分组过多(>5), 则使用 `upsetR` 绘制多种组合条件下的无限 `Venn` 图.

-   [√] Venn 图内每个空间内再分别上色.

---

###   算法的优化

**计算多边形的参数**

-   原始版本

```r
library(sf)
A <- st_difference(st_difference(st_difference(polygons[[1]],polygons[[2]]),polygons[[3]]),polygons[[4]])
B <- st_difference(st_difference(st_difference(polygons[[2]],polygons[[1]]),polygons[[3]]),polygons[[4]])
C <- st_difference(st_difference(st_difference(polygons[[3]],polygons[[1]]),polygons[[2]]),polygons[[4]])
D <- st_difference(st_difference(st_difference(polygons[[4]],polygons[[1]]),polygons[[3]]),polygons[[2]])
# ...
```

---

**计算多边形的参数**

-   现代版本

```r
function(venn, slice = "all"){
  overlap = overlap(venn, slice = slice)
  if (slice[1] == "all" | identical(venn@sets[slice], venn@sets)){
    discern = NULL
    return(overlap)
  } else {
    discern = discern(venn, slice1 = slice)
    return(sf::st_intersection(overlap, discern))
  }
```

---

**不规则图形的中心点**

-   `colorfulVennPlot` 的“硬核代码”

```r
# Midpoints for hard-coded crossover regions
midpoints <- data.frame(
  x = c(-4.37352, -6.04042, -6.04042, -0.43516, -0.32341, -2.19859, -2.19859, 5.65737, 6.16114, 7.03031, 6.04042, 2.54627, 2.54627, 2.53192, 2.53192),
  y = c(-0.65737, 5.43516, 2.45373, 11.04042, -2.03031, 7.19859, 2.46808, 9.36716, -1.16074, 5.31864, 2.45373, 11.04042, -1.04042, 7.19859, 2.46808),
  TF = c('1000', '0100', '1100', '0010', '1010', '0110', '1110', '0001', '1001', '0101', '1101', '0011', '1011', '0111', '1111'))
```

---

-   现在的代码

```r
process_region_data <- function(venn){
  region_items <- get_region_items(venn)
  counts <- sapply(region_items, length)
  region_ids <- get_region_ids(venn)
  region_names <- get_region_names(venn)
  tibble::tibble(
    component = "region",
    id = region_ids,
    item = region_items,
    count = counts,
    name = region_names
  )
}
```

---

**现在准备画图数据的代码**

```r
function(venn, ...){
  shape <- get_shape_data(nsets = length(venn@sets), ...)
  plot_data <- VennPlotData(
    setEdge = filter(shape, component == "setEdge") %>% pull(xy),
    setLabel = filter(shape, component == "setLabel") %>% dplyr::pull(xy)
  )
  plotData_add_venn(plotData = plot_data, venn = venn)
}
```

---

###   问题的解决

- 准备示例数据

```r
library(ggVennDiagram)
genes <- paste("gene",1:1000,sep="")
set.seed(20210419)
x <- list(A=sample(genes,300),
          B=sample(genes,525),
          C=sample(genes,440),
          D=sample(genes,350))
```

---

- 绘制韦恩图

```r
library(ggplot2)
ggVennDiagram(x) + scale_fill_gradient(low="white",high = "red")
```

![](2023-hzau-write-r-pkg_files/figure-html/unnamed-chunk-3-1.png)

---

-   更多维的数据

```r
x <- list(A=sample(genes,300),
          B=sample(genes,525),
          C=sample(genes,440),
          D=sample(genes,350),
          E=sample(genes,200),
          F=sample(genes,150),
          G=sample(genes,100))
```

---

-   七维

```r
ggVennDiagram(x, label = "none", edge_size = 2) + 
  scale_fill_distiller(palette = "RdBu")
```

![](2023-hzau-write-r-pkg_files/figure-html/unnamed-chunk-5-1.png)

---

-   六维

```r
ggVennDiagram(x[1:6], label = "none", edge_size = 2) +
  scale_fill_distiller(palette = "RdBu")
```

![](2023-hzau-write-r-pkg_files/figure-html/unnamed-chunk-6-1.png)

---

-   五维

```r
ggVennDiagram(x[1:5], label = "none", edge_size = 2) +
  scale_fill_distiller(palette = "RdBu")
```

![](2023-hzau-write-r-pkg_files/figure-html/unnamed-chunk-7-1.png)

---

-   四维

```r
ggVennDiagram(x[1:4], label = "none", edge_size = 2) +
  scale_fill_distiller(palette = "RdBu")
```

![](2023-hzau-write-r-pkg_files/figure-html/unnamed-chunk-8-1.png)

---

```r
ggVennDiagram(x[1:3], label = "none", edge_size = 2) +
  scale_fill_distiller(palette = "RdBu")
```

![](2023-hzau-write-r-pkg_files/figure-html/unnamed-chunk-9-1.png)

---

```r
ggVennDiagram(x[1:2], label = "none", edge_size = 2) +
  scale_fill_distiller(palette = "RdBu")
```

![](2023-hzau-write-r-pkg_files/figure-html/unnamed-chunk-10-1.png)

---

### 论文发表

- 被引很多次的论文

![](https://vnote-1251564393.cos.ap-chengdu.myqcloud.com/20231020145601.png)

---

### 包治百病

![](https://vnote-1251564393.cos.ap-chengdu.myqcloud.com/20231020144423.png)

---

## 如何开发一个 R 包？

### 实用工具

-   RStudio

-   devtools

-   usethis

-   GitHub/Gitee

-   R Markdown

---

###   工作流程

- 初始化

- 添加代码和文档

- 测试和上传

- 分发（dev/stable）

---

### 演示

*在线演示*

<http://122.205.95.109:8787/auth-sign-in?appUri=%2F>

---

### 如何维护？

**非常重要的事情**

-   GitHub Issues

-   Pull request

-   CRAN update

---

## 我是不是华农最会搞 R 的人？

-   请大家站在我的肩膀上，向上发展

-   请大家加入我的团伙，共同进步

-   <https://github.com/gaospecial/rconf2020-my-org-first-pkg>

-   <https://r-pkgs.org/>

---

## 为自己写一个 R 包？

-   既为自己：你的**需求**是什么？

-   也为他人：领域的**需求**是什么？

-   安装包的方式

-   `devtools::install_github("gaospecial/ggVennDiagram")`
  
    -   `remotes::install_github("gaospecial/ggVennDiagram")`
    
    -   `pak::pak("gaospecial/ggVennDiagram")`

---

## 谢谢！

-   我的微信：<@gaospecial>

-   其它交流方式

-   个人网站：<https://bio-spring.top>

-   Github 主页：<https://github.com/gaospecial>

-   邮件联系：<gaoch@mail.hzau.edu.cn>