Mobile element discovery with MELT brings different results when chromosomes are analysed separate

Question

[optimistsso4co3]

I get around 5x less mobile elements when calling each chromosome separate compared to when monolith genome .bam is supplied. What is the reason for that? Here is my example command:

java -jar ${meltDir}/MELT.jar Single -h ${refFasta} -bamfile ${id}.${chr}.bam -n ${refDir}/hg38.genes.bed -t mei_list.txt -w . -d 10000

[GenoMax]

A quick look at the manual indicates that MELT is designed to be run on genome(s).

MELT is optimized for performing discovery in a large number of individual sequencing samples using the Sun Grid Engine (SGE). MELT also has two additional workflows: analysis without SGE (for adaptability to other parallel computing platforms) and single genome analysis.

So running it separately on each chromosome probably treats them as different genomes. So you are not using the tool as designed/intended.

[optimistsso4co3]

It's slower when computing whole genome. I also haven't tried running multiple genomes at the same time, that could bring a different results.