What are rates inferred by --rate and --mlrate commands?

[trilisser]

Hi there! In the tutorial, I cannot find what rates inferred by --rate and --mlrate mean. For example, I get values from 0 to 100 by --mlrate analyzing my amino acid alignment and all I understand is that these are relative rates.

[bqminh]

--rate option computes the site-rates associated with discrete rate heterogeneity among sites model such as +G or +R. At the end of the analysis all model parameters are estimated, we can compute the posterior probabilities of each site falling into each of the rate categories. --rate then computes the site rate as weighted sum of the category rates, where the weights are the posteriors, or the the posterior mean. This method is also called empirical Bayes.

--mlrate is a different method. It doesn't require a rate heterogeneity model. Given a tree T with set of branch lengths B, for each site $D_i$ of the alignment, it compute a scaling factor $r_i$ so as to maximise the likelihood of $D_i$ where all branch lengths are rescaled by the factor $r_i$: $max_{r_i}: L(D_i|T,B*r_i)$. $r_i$ is bounded between 0 and 100 in IQ-TREE. Because there is no constraint about the distribution of site-rates like --rate option, $r_i$ can sometimes go to the maximum boundary, speaking that these sites might be saturated. It'd be interesting to study the impact of these saturated sites for phylogenetic analysis.

In any case, both methods produce site rates, either by empirical Bayes (--rate) or maximum likelihood (--mlrate). The results are values that can speak for the evolutionary rate of the sites. A value smaller than 1 means that the site evolved more slowly than average, i.e. the site is conserved. When the site rate is greater than 1, the site is fast evolving.

[trilisser]

Thanks a lot! Can you please clarify one moment—do I understand right that $r_i$ is the rate (second column in the .mlrate file)?

[bqminh]

Yes, that's correct