scripts for plink clustering (MDS and PCA) using either plink or VCF formats as input

plink_cluster_bfile.sh

#!/bin/bash

## PLINK CLUSTERING (plink1.9)
## takes input (any path) and creates clustering reports in working dir (to not mix data and results)

## OUTPUT PREFIX
prefix=$1

## MDS DIMENSIONS (optional, default = 2)
dims=${2:-2}

## LD PRUNNING
## https://www.cog-genomics.org/plink/1.9/ld#indep
plink --indep 50 5 2 --bfile $prefix --out $prefix

## DISTANCE
## https://www.cog-genomics.org/plink/1.9/ibd
## https://www.cog-genomics.org/plink/1.9/distance
plink --bfile $prefix --genome --out $prefix --extract ${prefix}.prune.in --recode # --distance --distance-matrix

## CLUSTERING
## https://www.cog-genomics.org/plink/1.9/strat

## MDS
plink --file $prefix --read-genome ${prefix}.genome --cluster --mds-plot $dims --out ${prefix}_${dims} # --read-dists ${prefix}.dists

## PCA
plink --file $prefix --read-genome ${prefix}.genome --cluster --pca --out ${prefix} # --read-dists ${prefix}.dists

plink_cluster_vcf.sh

#!/bin/bash

## PLINK CLUSTERING (plink1.9)
## takes input (any path) and creates clustering reports in working dir (to not mix data and results)

## OUTPUT PREFIX
prefix=$(basename -s .gz $1) # remove .gz if present
prefix=$(basename -s .vcf $prefix) # remove also .vcf

## MDS DIMENSIONS (optional, default = 2)
dims=${2:-2}

## LD PRUNNING
## https://www.cog-genomics.org/plink/1.9/ld#indep
plink --indep 50 5 2 --vcf $1 --out $prefix

## DISTANCE
## https://www.cog-genomics.org/plink/1.9/ibd
## https://www.cog-genomics.org/plink/1.9/distance
plink --vcf $1 --genome --out $prefix --extract ${prefix}.prune.in --recode # --distance --distance-matrix

## CLUSTERING
## https://www.cog-genomics.org/plink/1.9/strat

## MDS
plink --file $prefix --read-genome ${prefix}.genome --cluster --mds-plot $dims --out ${prefix}_${dims} # --read-dists ${prefix}.dists

## PCA
plink --file $prefix --read-genome ${prefix}.genome --cluster --pca --out ${prefix} # --read-dists ${prefix}.dists

TODO

• add README file (for git clone) with links and explanation of different options (e.g. distances)

These scripts are probably fine for a one-off analysis of samples. But after working with them today, rerunning stuff a few times, I realized there are several rather dumb decisions and the scripts can be somewhat improved. Things like:

• Doing --pca and --mds-plot in the same command. They really don't need to be run separately.
• Not using a text file set with --recode for the prunned data. This comes from a tutorial I've found back in the day and I guess it was there to solve a name collision, as using the same --bfile would cause it, but using --file can avoid it. Or, you know, just make a new name for the prunned dataset, so you can keep using --bfile (i.e. the smaller and more efficient format).
• Using --genome and --read-genome can be completelly ommited. Plink will do it automatically with --cluster (or whatever different option if feels is better atm). Also this way it can be removed from the prunning step, so all clustering can be done within one step. I may add the options again later (only to the last clustering steps) to make rerunning the scripts more efficient, but that will need some detection of files to work.
• Some log files are getting overwritten by the following commands, because they use the same --out names. This should be easy to fix, but needs a little thought for rerunning (i.e. when some steps may get skipped).

There may be some other improvements, probably more involved, like detecting if prunning was already done and skipping the step if so, to make rerunning easier, but that will be for another time.