Mitochondrial Haplogroup Association with Non-Small Cell Lung Cancer Patients

Background Information

Mitochondrial DNA Overview

This picture is a work by Emmanuel Douzery

Small circular DNA made up of 16,569 base pairs
Multiple copies within the mitochondrion organelles of the cell
Contains 37 genes
- 2 rRNAs
- 22 tRNAs
- 13 Proteins
  - Encoded proteins are involved in oxidative phosphorylation process, generating energy through ATP production
Economically packed with overlapping genes and a very minimal non-coding region
Maternally inherited

Mitochondrial Haplogroup Overview

This image is of mtDNA Haplogroup Migration Map from 23andMeBlog

Mitochondrial Eve lived around 200,000 years ago and is considered to be living humans' most recent matralinial common ancestor
- Originating in Southeast Africa
The accumulation of mutations in our mtDNA allow for the traceback of our lineage
- Single Nucleotide Polymorphisms (SNPs) can be detected ar specific locations in the mitochondrial genome which can be grouped and assigned to global populations
- The highly polymorphic, non-coding control region in the D-loop of the mtDNA is also known as the hypervariable region and is best for determining mtDNA lineage
Haplogroup nomenclature began in North America with A, B, C, and D and branched out through all the letters of the alphabet
- The native North American haplogroups are branched out from Asia due to early migration

Non-Small Cell Lung Cancer (NSCLC)

NSCLC accounts for 80%-85% of lung cancers
- Lung Adenocarcinoma (LUAD) is the most common malignancy
Heavily related to tobacco smoking but also the most frequent lung cancer found in “never-smokers”
Tobacco carcinogenesis is believed to play a role in the vulnerable mtDNA through alterations by way of somatic mutations

Project Goals

Assign mitochondrial DNA to haplogroups
Perform and compare global differential gene expression analysis across haplogroups
Examine overall impact haplogroups may have on NSCLC LUAD outcome, stage, and metabolism

Project Workflow

Obtain controlled access data from the GDC Data Portal
Load data into cluster environment for read depth analysis
Generate file formats: MPG, VCF, FASTA, and HSD
Call probable haplogroups
Integrate haplogroups with participant clinical data
Perform RNA-seq analysis with high dimensional data visualization

Data Fetching

Obtained whole exome, aligned, paired-end sequenced reads from GDC Data Portal
- TCGA - The Cancer Genome Atlas
- CPTAC - Clinical Proteomic Tumor Analysis Consortium
Loaded data to HPC with GDC data trasnfer command

module load anaconda/2
gdc-client download -t <token> -m <manifest>

Initial Data Analysis

Goal to analyze TCGA and CPTAC samples with read depth scripts to see which to use for full study
Wrote a shell pipe to loop through each BAM to get number of mitochondrial DNA mapped reads per sample with Samtools
- Mitochondrial region may be denoted 'chrM' or 'MT', needs to be checked
Helps assess overall quality of samples

for i in samples/*/*.bam
do
    samtools idxstats ${i} > stats/$(basename ${i} .bam)_idx.bam
done
grep -H 'chrM' stats/*.bam > stats/chrM_bamstat_summary.txt

Samtools version 1.9

Next observed per-base read depth with PileUpStats.py script
- Counts number of bases with read depths greater than 5 and 10
- Returns summary file with fraction of bases with read depths of 5 or 10 or greater, and mean, median, and minimum read depths per sample
- Outputs overall stats for all samples read
The greater the read depth, the more confidence and accuracy in base calls
- Important for calling variants and when ensuring the correct haplogroup is called

Python version 3.7.2 Pysam version 0.16.0.1 Numpy version 1.21.1

Use Integrative Genomics Viewer (IGV) to visually check selected TCGA and CPTAC samples to match calculations with observations
Overall, the initial data analysis showed that the CPTAC study mitochondrial reads had a much greater coverage and read depth and would be better to utilize for this study

File and Sample Renaming

Original CPTAC folders and files are a string of ~50 random chracters which my prove to be troublesome in downstream analysis
A sample sheet is supplementary material that was downloaded with the sample files
- It can work as a name map to map back the long character string to a shorter Case ID name
I developed the python scripts "RenamerByMap.py" and "RenamerByMapCounts.py"
- These map the long name back to the Case ID and append the first letter of the sample type to the end
  - Sample could be of Tumor, Normal Tissue, or Blood origin
- The first script is used for renaming the BAM files and their folders
- The second is used for the count files when preparing to perform DGE analysis

Python Version 3.9.5 Pandas Version 1.2.4

Haplogrouping

Used Bash/Perl Script from Dr. Jamie Teer to obtain an HSD file from the BAM files
- This file contained the SampleID, Range, Haplogroup (column blank), and polymorphisms
This file is the input to the Haplogrep2 tool which calls the most probable haplogroup based on samples polymorphisms
- These polymorphisms were additionally observed in IGV to visualize variations in the genome

Haplogrep version 2.2

Clinical Data Aggregation

I then developed a Python script to parse through the clinical data in JSON format and extract information
- Race, Ethnicity, Vital Status, Smoker Status
The script output a TSV data array with the Case ID and the specified clinical information Python version 3.9.5 Numpy version 1.21.1
I obtained additinal clinical data from CPTAC-3 study site
- This contained many other parameters and helps clear up smoking status and ethnicity variables

Differential Gene Expression Preparartion

Imported data into R
- HSD File in TSV format
- CPTAC-3 Clinical data
  - From JSON in TSV format
  - From site in CSV format
- HT-seq primary tumor count data from GDC data portal
The first letter of each haplorgoup was subset and assigned to a new column, 'HaploID'
The data was clean and merged into a single data frame
Sample data distributions were visualized

Differential Gene Expression Analysis

Followed RNA-seq protocol utilizing edgeR, limma, and Glimma
- Paper from Law, C. et al.
With Asian and European haplogroups being most heavily represented they were chosen as the main comparison factor
- Specifically, HaploIDs B vs H and M vs R were compaired
No significantly differentially expressed genes were found
Further documentation and visualization for this workthrough can be found in the 'markdowns' and 'Rscripts' folders
- Including package version information

Future Directions

Add in normal tissue sample information from CPTAC study
- Was not observed here due to uneven amount of samples
Look into TCGA study through RNA-seq analysis
- Greater number of samples, although mtDNA read depth is not as great
Study gene expression in "never-smokers" to see possible effect of carcinogenesis

aobermayer4 / CPTAC_LUAD_HaploAnalysis