I know this is an elementary question, but I've been struggling to find an answer.
RAxML calculates "pairwise ML distances" (with the option -f x), what does these distances mean? what are the units?
In the RAxML manual says "This option will just compute the pair-wise distances (branch lengths) between the sequences of an alignment using a maximum likelihood estimate". When I compute this i get values grater than 1. Is it okay? How should I interpret these values?
This is the command i'm using to get the distances:
>> raxml -f x -t my_tree -m GTRGAMMA -n otuput_distace_matrix -s genes_aligned.fasta -T 2
The simplest measure of "distance" between two DNA sequences is the percent identity. If two sequences are 95% identical, they have a distance of 5% or five bases different per every 100 sites compared on average.
Over time, as mutations accumulate in genomes, the accumulation of the mutations is not random. Transitions happen more often than transversions, for example, and most coding regions of the genome are under selection pressures which eliminate detrimental mutations and/or select for beneficial mutations. So if we observe the evolution of HIV-1 for example within infected individuals or in local epidemics, we see that each lineage tends to change at a rate of roughly 0.5% per year, so two viruses that shared a common ancestor one year ago are now 1% distant from each other (each having changed 0.5% from the ancestor). However, after ten years time the distance is a bit less than 10% because of selection pressures and nonrandom patterns of evolution. Any two random strings of TGCTAGCGACTTG DNA should be roughly 25% identical to each other, so after 100 years time we will not observe 100% distance or 0% sequence identity even if there was no selection at all. Instead we observe that over longer timescales, the rate of distance growth decreases.
So phylogenetic distances are an attempt to correct for these issues with of nonrandom change over time to crate a distance:time scale that is more accurate than the simple % identity. A phylogenetic distance of 1.0 does indicate that each site in the DNA has mutated one time on average, but what we observe is that some sites have mutated several times, while most sites have not mutated (or more correctly, individuals with those mutations did not reproduce because the mutation was lethal or selected against). So the simple percent identity between two sequences with a phylogenetic distance of 1.0 is not zero percent identity. And it is possible to analyze sequences that have phylogenetic distances even greater than 1.0. However, it is important to keep in mind that the greater these distances are, the greater the "correction" between percent identity and computed "phylogenetic distance" is, and thus the calculation of the exact distance and timescale runs into problems.
Comparing HIV-1 to HIV-2 for example, we can only guess how long ago the common ancestor of these two primate lentiviruses lived. The phylogenetic distance alone would tell us that the common ancestor could have lived as little as a few hundred years ago, but many other lines of evidence suggest a time scale closer to millions of years ago. So it is important to consider fossils and other evidence as well as genetic distances.
Dear Maria
Branch length interprets the phylogenetic distance of a taxon from the root. In fact, when a branch is longer, compared with other branches and scale, includes longer-standing evolutionary history and higher distance from a common ancestor.
Maria Cadavid Velez
Good luck
Afsaneh
As may be seen from the previous answers, branch lengths can be measured or estimated in a variety of ways, and mean different things to different people. They can be affected by differential weighting of characters and character state transformations, as well as by whether the character state transformations are optimized as accelerated or delayed ("ACCTRAN" vs. "DELTRAN").
I do not consider your question elementary, I consider it to cut to the core of the challenges of using these methods: if the scientific community cannot agree on common definitions of its concepts and articulate them intelligibly, then maybe they are not very useful concepts.
Brian Thomas Foley
Dear Brian, thanks a lot for your answer, it is much more clear to me now.
I was wondering why I was getting this "huge" distances as im using very closely related genomes. Thanks for explaining.
Afsaneh Asgharzadeh and Andrew V Z Brower Thanks for your answers!
- Badges
- Science method