I see your question has its answer in itself. Its more or less a choice of user to make in regard to her/his data.

So there can be several things upon which the analysis can be based on:-

• What is your aim?

If the main aim is to just study the mt genome, its easy and fast to just use the data which we are interested in. It saves the computation, efforts and analysis time. And if the organism has already a well annotated genome, why would re-annotate the genome.

• Filtering out a part of data does not always mean that you will be throwing away the rest of the data. Its just a way to better manage the data and its processing.
• I would say that filtering would increase the accuracy. Which accuracy? accuracy of reads of accuracy of assembly? Accuracy is more dependent on the data, its processing and the reference used. Thus, only filtering would not guarantee the increase in accuracy. Moreover, the filtering of the mtDNA is always based on some reference, thus it is also a big factor to decide the accuracy.

>> If you are interested in mitochondrial genome, what would you do instead if you are not filtering, assembling and annotating it? Or do you want to use the whole genome too. Think of the computational power needed in case of any big and complicated genome.

Hi Abhijeet,

thanks a lot for your detailed answer! Really appreciate it.

I was keeping my question a bit more general on purpose to see if there are any general recommendations regarding this approach. But I realize that it's highly dependent on the purpose - as it is always the case actually.

What I'm currently doing is to de novo assemble mtgenomes of arthropods without an available reference mtgenome using SPAdes, mainly to build up a reference database. I could filter mtreads on the basis of closely related species with an available mtgenome. Computational power is not an issue for me.

I'm not 100% familiar with how de novo assemblies exactly work (as in the algorithms behind it) and hence am wondering how to estimate the effect of filtering on the approach. That means, which effect it has to have less reads interfering each other while having more information available, or on the other hand more reads of my actual target mtgenome with the probability of having some additional mtreads filtered out beforehand.