The dangers of reconstructing evolution using mtDNA

[mummymonkey]

The always readable Jerry Coyne has written his thoughts on the use of mtDNA versus nDNA when constructing family trees. He's discussing a paper on Polar bears which I freely admit are not catfish. However I thought it was a good piece and worth drawing to the attention of forum members.

Read it here

[Mike_Noren]

Sadly his point is rubbish. If you want to see truly ridiculous trees, brutally and obviously at odds with morphology, read studies using only nuclear genes. Exhibit A is this gem, which used six nuclear genes to give us the new superfamily Paedocypridoidea, a result so silly one of the authors contradicted it just 14 pages later in the same issue!

I don't think I'm giving away too much if I say that excessive belief in the reliability of nuclear genes was at the base of why the study got such a ridiculous result, and quite possibly also the reason the authors managed to get it published.

The recent rise in the use of multiple nuclear loci in phylogenetic analyses are not because nuclear genes are better, but because new sequencing techniques have made sequencing multiple nuclear loci affordable. Nuclear genes have several mechanisms which can make them positively misleading in phylogenetic analysis, they are not inherently superior for phylogenetic reconstruction.

Oh yeah, also his claim that mitochondrial genes "move more easily between species" than nuclear genes is rubbish too.

[racoll]

Completely agree with Mike.

It's not simply an issue of "being wary of mtDNA" or nit using it, as suggested in the article; scientists should be wary of any set of data, whether nuclear, mitochondrial, morphological, chemical, or ecological etc.

There is misleading signal or noise in all kinds of data, and the skill is in avoiding erroneous conclusions wherever they are.

[Suckermouth]

I'm also suspicious of the claim that mtDNA is more likely to move between species. Nuclear DNA moves between species as well, but I assumed the larger size of the nuclear genome and the larger effective population size means you're pretty unlikely to get that particular locus that's inherited from another species. Furthermore, that introgressed haplotype won't get wiped by recombination in the mitochondria, while this is a factor in nuclear loci. mtDNA and nuDNA should both move between species, but nuDNA has more factors to clear that signal away or to make signal detection more difficult.

The issue of reliance on mtDNA has less to do with hybridization and more with reliance on a single locus.

In any case, decreased sequencing costs from new technologies allowing for inexpensive sequencing of genomic-scale datasets will reduce the reliance on few loci when hundreds to thousands can be easily sequenced.