temporal signal

[Niti Jadeja]

We have rabies virus genomes from 15 months spread across Pune city of India. Trying out TempEst lead to the inference that the measure of temporal signal in our study exhibits a low association between genetic distances and sampling dates (R2 =0.0226).

Would Molecular clock based phylogenetic reconstructions (using BEAST) be valid if this is the case? I have used Maximum Likelihood and Bayesian methods to generate trees.

[Artem B]

Greetings,

Tempest assumes strick clock-like gene behaviour. In the case of RNA viruses it may not be quite true.

You can perform the permutation test (just shuffle isolation dates between sequences several times (the more replicates the better) and do independent analyses with each replica and after that compare 95% HPDs of an evolutionary rate parameter with estimates based on true isolation dates). If 95% HPDs of rates based on true dates don't overlap with permutation replicates results, then your data have a strong temporal signal. The method is "dumb" but gives the strong evidences.

Another way (but a little bit harder) is BETS, you should calculate marginal likelihoods for models with dated tips and without tips calibration (an ultrametric tree) and then calculate Bayes factors of these models (more about BETS read here)

Cheers,

Artem

пятница, 18 декабря 2020 г. в 01:29:37 UTC+8, ntjad...@gmail.com:

[Artem B]

More about that you can also find here 

пятница, 18 декабря 2020 г. в 01:29:37 UTC+8, ntjad...@gmail.com:

We have rabies virus genomes from 15 months spread across Pune city of India. Trying out TempEst lead to the inference that the measure of temporal signal in our study exhibits a low association between genetic distances and sampling dates (R2 =0.0226).