Context-specific house keeping genes

Purpose

The purpose of this project lies in analysing the expression levels of various genes of different tissues and discovering a "context-specific" housekeeping gene. "Context-specific" refers to consistency in the gene expression across different cell types, disease and developmental stage.

Set up

SSH

SSH key A public key was set and saved in /home/yerihan/.ssh/id_ed25519.pub

Guide to set up a SSH key

ssh-keygen -t ed25519 -C "[email protected]"

Github

git clone [email protected]:hanyeri/cshkg

User information

git config --global user.email "[email protected]"
git config --global user.name "Yeri Han"

Rstudio

On Fedora 36, install ...

sudo dnf install rstudio-desktop

Developmental files-devel, R header files

sudo dnf install make automake gcc gcc-c++ kernel-devel
sudo dnf install R-core-devel R-java-devel libRmath-devel

R packages

• io: read from, write to, plot to in an unified manner
• ggplot2
• dplyr
• annotate
• hgu133a.db: human genome u133a expression array annotation data

Bioconductor packages

install.packages("BiocManager")
BiocManager::install(c("annotate", "hgu133a.db"))

CRAN packages

install.packages(c("io", "ggplot2", "dplyr"))

Datasets

This repo is organized by different datasets:

1. biogps
2. pancan
3. gtex

Download and preprocess data

cd to the data subdirectory.

Within a dataset directory, run ./get.sh.

Prepare the data by running Rscript preprocess.R.

Analysis

Set R working directory to be the same as the script.

[Session] -> [Set Working Directory] -> [To Source file location]

Then, run analyze.R in pancan.

Mathematcis

We calculate variations within and between the groups as follows.

For each gene, between-group standard deviation \sigma_{bw} is calculated as

\sigma_{bw} = \sqrt{ (\sum_g^G m_g - \bar{m})^2 / (G - 1) }
\bar{m} = \frac{1}{G} \sum_g^G m_g

where G is the number of groups, and m_g is the mean expression within group g.

The within group standard deviation \sigma_{wi} is calculated by

\sigma_{wi} = \sqrt{ (\sum_g^G ((x_i^{(g)} - \bar{x}^{(g)})^2) / ( \sum_g^G N_g - 1 )  }
\bar{x}^{(g)} = \frac{1}{N_g} \sum_i^{N_g} x_i^{(g)}

where N_g is the number of samples within group g, and x_i^{(g)} is the expression of sample i within group g.

Git

• add: take modified file from working directory to Git index (staging area)
• commit: creates a new revision log, you can only commit after you add
• push: add the change to the GitHub
• diff: outputs the difference between two inputs

Guide on Git

Troubleshooting

Bioconductor

Bioconductor packages could not be downloaded completely at first because R was an older version and the outdated packages were relocated to a backup server, taking too much time to be retrieved.
Both Fedora and R are updated and biocManager packages were then be able to be downloaded. Guide on updating Fedora