Install PICRUSt on windows

class: center, middle, inverse, title-slide

# Install PICRUSt on windows
### <br/>Hsiu J. Ho
### 2017/08/18

---

### PICRUSt
- PICRUSt: Phylogenetic Investigation of Communities by Reconstruction of Unobserved States
- 就是functional prediction, 可以透過greengene的OTUID轉換成KEGG或COG各功能的含量, 其中KEGG又可進一步知道pathway
- 進一步資訊請看[官方網頁](http://picrust.github.io/picrust/index.html)
- [the PICRUSt GitHub repository](https://github.com/picrust/picrust)

---
### 安裝PYTHON
- 安裝phyton 2.7版, 其中又分為32位元及64位元, 這會影響之後套件安裝的版本, [官方網頁](https://www.python.org/downloads/)上下載預設是32位元, 若要64位元要在[網頁](https://www.python.org/downloads/release/python-2713/)中搜尋下載點, 本例是用32位元的python.
- 安裝python套件的方法:
```yaml
pip install foo
pip install foo.whl
pip install -U foo
pip uninstall foo
```
其中pip及foo.whl要完整路徑, 而foo.whl是在[Unofficial Windows Binaries for Python Extension Packages](http://www.lfd.uci.edu/~gohlke/pythonlibs/)下載的套件, -U表示更新, uninstall為反安裝.

---
### 安裝PICRUSt
- 在 [Installing PICRUSt](http://picrust.github.io/picrust/install.html#install) 頁面中有簡易的安裝流程.
- 下載 [VCForPython27.msi](https://www.microsoft.com/en-us/download/confirmation.aspx?id=44266) 並安裝
- 下載 [numpy+mkl.whl](http://www.lfd.uci.edu/~gohlke/pythonlibs/#numpy) 及 [scipy.whl](http://www.lfd.uci.edu/~gohlke/pythonlibs/#scipy) 的32位元版
- [下載PICRUSt的python程式](https://github.com/picrust/picrust/releases/download/1.1.1/picrust-1.1.1.tar.gz)並解壓縮
- 在[頁面](http://picrust.github.io/picrust/picrust_precalculated_files.html#id1)下載PICRUSt’s Precalculated Files, 在項目Greengenes v13.5 (and IMG 4)及Greengenes 18may2012之下的檔案都要下載, 並在安裝PICRUSt前, 把檔案移入上一步解壓後路徑picrust/data資料夾中.
- 開啓命令提示字元(cmd), 輸入以下指令, 請自行增加檔案所在路徑:
```yaml
pip install numpy-1.13.1+mkl-cp27-cp27m-win32.whl
pip install scipy-0.19.1-cp27-cp27m-win32.whl
pip install cogent
pip install biom-format
pip install h5py
```
- 安裝PICRUSt, 會把PICRUSt’s Precalculated Files複製到系統路徑底下, 因為下載程式及檔案就可以跑PICRUSt了, 但會讀取在系統路徑底下的Precalculated Files, 其中...是忽略部分路徑, 實際是要完整路徑資訊才能執行.
```yaml
C:\Python27\Scripts\pip install E:\...\picrust-1.1.1\.
```

---
### Example 1
- 在PICRUSt程式壓縮檔中, 有官方提供的教學資料
```yaml
.\picrust-1.1.1\tutorials\hmp_mock_16S.biom
```
- 因為教學資料中的OTUID是參照Greengenes 18may2012
- 在cmd中, 預設是預測KEGG
```yaml
cd E:\fastq\picrust\Precalculated\v1.1.1\picrust-1.1.1
e:
c:\python27\python .\scripts\normalize_by_copy_number.py -g 18may2012 -i .\tutorials\hmp_mock_16S.biom -o .\tutorials\normalized_otus.biom
c:\python27\python .\scripts\predict_metagenomes.py -g 18may2012 -i .\tutorials\normalized_otus.biom -o .\tutorials\metagenome_predictions.biom
c:\python27\python .\scripts\predict_metagenomes.py -g 18may2012 -f -i .\tutorials\normalized_otus.biom -o .\tutorials\metagenome_predictions.tab
c:\python27\python .\scripts\metagenome_contributions.py -g 18may2012 -i .\tutorials\normalized_otus.biom -l K00001,K00002,K00004 -o .\tutorials\ko_metagenome_contributions.tab
```
---
### Example 1
- 也可以改為預測COG
```yaml
c:\python27\python .\scripts\predict_metagenomes.py --type_of_prediction cog -g 18may2012 -i .\tutorials\normalized_otus.biom -o .\tutorials\metagenome_predictions_COG.biom
c:\python27\python .\scripts\predict_metagenomes.py --type_of_prediction cog -g 18may2012 -f -i .\tutorials\normalized_otus.biom -o .\tutorials\metagenome_predictions_COG.tab
dir .\tutorials
```
- 可以得到兩種格式的輸出檔biom及tab, 其中biom的資訊量會比tab多.
- 可參考[Analyzing PICRUSt predicted metagenomes](http://picrust.github.io/picrust/tutorials/downstream_analysis.html#downstream-analysis-guide)的介紹, 進一步用其他軟體分析

---
### 在R環境下匯入及匯出BIOM格式檔

```r
install.packages("devtools") # if not already installed
library("devtools")
install_github("joey711/biom")
```

```r
path="E:/fastq/picrust/Precalculated/v1.1.1/picrust-1.1.1/tutorials"
a1=biom::read_biom(file.path(path,"metagenome_predictions.biom"))
a1
```

```
# biom object. 
# type: OTU table 
# matrix_type: dense 
# 6885 rows and 2 columns
```

---
### 在R環境下匯入及匯出BIOM格式檔

```r
a2=readr::read_tsv(file.path(path,"metagenome_predictions.tab"),skip = 1)
dim(a2)
```

```
# [1] 6885    4
```

```r
head(a2,3)
```

```
# # A tibble: 3 x 4
#   `#OTU ID` staggered   even                           KEGG_Description
#       <chr>     <dbl>  <dbl>                                      <chr>
# 1    K00001    431689 416486         alcohol dehydrogenase [EC:1.1.1.1]
# 2    K00002      4465  16404 alcohol dehydrogenase (NADP+) [EC:1.1.1.2]
# 3    K00003    107292 141415      homoserine dehydrogenase [EC:1.1.1.3]
```

---
### Example 2: 在R中完成PICRUSt

```r
b1=biom::make_biom(phylo@otu_table@.Data)
setwd("E:/fastq/picrust/Precalculated/v1.1.1/picrust-1.1.1")
#export
biom::write_biom(b1,"./tutorials/b1.biom")
#gene copy number
system("c:/python27/python ./scripts/normalize_by_copy_number.py -i ./tutorials/b1.biom -o ./tutorials/b1_normalized_otus.biom")
#KEGG functional prediction
system("c:/python27/python ./scripts/predict_metagenomes.py -i ./tutorials/b1_normalized_otus.biom -o ./tutorials/b1_metagenome_predictions.biom")
#彙整pathway
system("c:/python27/python ./scripts/categorize_by_function.py -i ./tutorials/b1_metagenome_predictions.biom -c KEGG_Pathways -l 3 -o ./tutorials/b1_metagenome_predictions.L3.biom")
#COG functional prediction
system("c:/python27/python ./scripts/predict_metagenomes.py --type_of_prediction cog -i ./tutorials/b1_normalized_otus.biom -o ./tutorials/b1_cog_metagenome_predictions.biom")
#COG_Category
system("c:/python27/python ./scripts/categorize_by_function.py -i ./tutorials/b1_cog_metagenome_predictions.biom -c COG_Category -l 1 -o ./tutorials/b1_cog_metagenome_predictions.L1.biom")
system("c:/python27/python ./scripts/categorize_by_function.py -i ./tutorials/b1_cog_metagenome_predictions.biom -c COG_Category -l 2 -o ./tutorials/b1_cog_metagenome_predictions.L2.biom")
```

---
### Example 2: 在R中完成PICRUSt
- 匯入PICRUSt結果

```r
setwd("E:/fastq/picrust/Precalculated/v1.1.1/picrust-1.1.1/tutorials")
b2=biom::read_biom("b1_metagenome_predictions.biom")
b3=biom::read_biom("b1_metagenome_predictions.L3.biom")
b4=biom::read_biom("b1_cog_metagenome_predictions.biom")
b41=biom::read_biom("b1_cog_metagenome_predictions.L1.biom")
b42=biom::read_biom("b1_cog_metagenome_predictions.L2.biom")
```
- 資料整理: biom to matrix

```r
require(pipeR)
b5=do.call(rbind,b4$data)
rownames(b5)=biom::rownames(b4)
b6=biom::observation_metadata(b4)
b6$COG_Category %>>%
  (regmatches(., gregexpr(pattern="\\[[A-Z]\\]", .))) %>>%
  sapply(length) %>>%
  table()
```

---
### Debug
- 在安裝套件及執行PICRUSt的過程中, 可能會出現錯誤, 這時就GOOGLE錯誤資訊尋找解決的方法.
- 可能需要更新套件
- 可能缺少套件, 補齊後再重新安裝之前的套件
- 也有可能位元版本不一致, 下載正確的版本後再重新安裝之前的套件