Metagenomic sequencing of CSF samples

# Wiki

## AIM: To develop a predictive algorithm to determine whether an infectious or other non-infectious cause is likely or not.

The aim will be achieved based on

1. Human RNASeq & downstream analysis as noted specifically related to immune system genes

1. Assess the human immune system genes DNA in particular but not limited to interferon, cytokines and chemokines)

## Sample data for all the participants is on ilifu in

        /cbio/projects/017/definitive/

## Detailed information regarding participants is provided in a txt file 

        /cbio/projects/017/patients_clinical_details.txt

Of the planned 47 participants, COVC04, COVC07, COVC23 and COVC30 were excluded based on the clinical notes shared by Ruan Marais on 18 July 2022 on slack: https://cbio.slack.com/files/U02LWC4GQTE/F03PZ1H8J0J/table_1_-_clinical_details.xlsx. 

As at **10 August 2022**, one participant: COVC26 is outstanding in **/cbio/projects/017/definitive/**, as such the metadata file excludes this participant.

> /cbio/projects/017/metadata.txt

`metadata.txt` is a file that consists of the three columns of 

> /cbio/projects/017/patients_clinical_details.txt

It was created by reading the .xsls file in R and write the "samplename", "COVID-19 status" and "Neurological symptoms due to COVID-19"

## Important things to note: 

We perform the RNA seq gene count using the 

        nf-core/rnaseq pipeline. 

`nf-core/rnaseq` does read quality checks using **FASTQC** , read trimming by **TrimGalore** , read mapping by **STAR** & quantification by **SALMON**.

To run the pipeline, we create a **samplesheet.csv** for the analysis by using **fastq_dir_to_samplesheet.py** obtained from the **nf-core** by using **wget -L https://raw.githubusercontent.com/nf-core/rnaseq/master/bin/fastq_dir_to_samplesheet.py**. And changed the file permissions to executable

``` shell

        chmod 755 fastq_dir_to_samplesheet.py

```

Run the script

``` shell

 ./fastq_dir_to_samplesheet.py /cbio/projects/017/definitive/ /cbio/projects/017/analysis/samplesheet.csv --strandedness reverse

```

## Run the `nf-core/rnaseq` pipeline,

``` shell

sbatch /cbio/projects/017/rnaseq/rnaseq-pipeline.sh

```

Upon getting the quantification results **(star_salmon)**, downstream analysis is done using **R programming** language on a local machine. The **working directory** is

> /home/ephie/UCT-DATA_ANALYST/BioinformaticsSupportTeam/ruan/definitive/results/

using the **R script**

``` shell

/home/ephie/UCT-DATA_ANALYST/BioinformaticsSupportTeam/ruan/definitive/dge_downstream.R

```

We use **DESeq2** for differential gene expression analysis, and **R packages** including **ggplot** and others. In short, the **R script** does

1. Count normalization that ie creation of the DESeq2Dataset object.

1. Exploratory data analysis (PCA & heirarchical clustering) - identifying outliers & sources of variation in the data:

1. Running the DESeq2 using the "DESeq2" function

1. Check the fit of the dispersion estimates: using "plotDispEsts"

1. Create contrasts to perform Wald testing on the shrunken log2 foldchanges between specific conditions:

1. Output significant results

1. Visualize results: volcano plots, heatmaps, normalized counts plots of top genes, etc.

1. Take note of all the versions of all tools used in the DE analysis: