qtl mapping

1. Tensorqtl

Tensorqtl algorithm applied on both 21 different cell types and seven different broad cell types resulted from snRNA-seq data.

1.1. Preparing TSS file:

In Python, using recommended function from Tensorqtl tutorial (pyqtl), the TSS file was created which containing chr, start, end and gene_id.

1.2. Preparing BED and Covariate files in R

Using the TSS file, Raw and normalized data of snRNA-Seq, the BED files and the Covariate files were created in R. Briefly speaking, I used metadata from the raw file of snRNA-Seq data and merged it with the TSS file. Then, after loading normalized data, I created the sorted BED file. For the covariate file, I used num.sv from SVA package to calculate how many PCA should I include in my analysis, which for 21 cell types it ranges from 15 to 20.

Here is an example of covariate file from astro cell type :

and Here is an example of BED file from astro cell type :

1.3. Runnig Tensorqtl in BASH

Using a for loop, I run Tensorqtl with three different parameters :

1. Default parameters :

for f in ./*/*bed.gz
do
echo "Start analyses of $f"
python3 -m tensorqtl "MDD_control_sc_maf1"
$f
${f/_df.bed.gz/}
--covariates ${f/df.bed.gz/covariates.txt}
--mode cis_nominal
done

2. Add a window of 500,000 instead of a defalut value(1Mbp).

for f in ./*/*bed.gz
do
echo "Start analyses of $f"
python3 -m tensorqtl "MDD_control_sc_maf1"
$f
${f/_df.bed.gz/_modified}
--covariates ${f/df.bed.gz/covariates.txt}
--mode cis_nominal --window=500000
done

3. Add a maf=0.05 instead of a defalut value(0).

for f in ./*/*bed.gz
do
echo "Start analyses of $f"
python3 -m tensorqtl "MDD_control_sc_maf1"
$f
${f/_df.bed.gz/_modified_maf}
--covariates ${f/df.bed.gz/covariates.txt}
--mode cis_nominal --window=500000 --maf_threshold=0.05
done

1.4. Loading Parquet file and applying p-value correction :

In the final step, Parquet files from different analyses were loaded in R and after calculation FDR, SNPs with FDR < 0.05 were selected for further investigation.

2. Matrix eQTL

Matrix eQTL algorithm applied on both 21 different cell types and seven different broad cell types resulted from snRNA-seq data.