BSP, serially sampled data and lack of precision in the sampling dates

[Ursulla]

Hi,
I am new to BEAST and I have got a question about the Bayesian skyline
plot using serially sampled data.
I am working on human flu and I have got sequences spanning over 4
years. I was wondering whether one can use BSP using only the year of
sampling and to expect to see accurate variations in the genetic
diversity.

My concern is that without knowing the month of sampling, a virus
sampled in December 2005 should more closely related to a January 2006
one than a January 2005.

The plot shows variations but I'm not sure if it is right since flu
seasons overlap two years. I am pretty sure I could not see any intra-
season variation but I was hopping to see inter-season variation.

My guess is that I can not do anything without the month of sampling
(the day of sampling would be even better), can anyone tell me what to
do?

Thank you very much

Ursulla

[Julian W Tang]

Hi Ursulla,
 
I have played around with influenza sequence Beast analysis for our data set (10 years of HA, NA and MP full-length sequences collected between 1997-2006 - about 30 sequences a year from each year. These sequences are described in more detail in: http://www.plosone.org/article/info:doi/10.1371/journal.pone.0002768).
 
You can get some results from the BSP analysis that almost match the real influenza incidence over the same period (make sure you reverse the time axis and plot on the BSP from Tracer before comparing it to your incidence plot).
 
But you need to have the exact dates of collection for each sample - otherwise your temporal resolution is so poor (as you pointed out) that you cannot tell whether you are looking at a sample collected at the beginning or the end of an influenza season, if they both occur in the same year.
 
Also, practically, you need to incorporate these exact collection dates into the sequence titles in a way that Beauti can use 'Guess Dates' to read the dates of collection. Otherwise, you have to type them all in individually, which is tedious. There is a quick way to do this using Excel.
 
Then you have to choose a suitable evolutionary model for these sequences. I have tried several combinations based on advice from Alexei and Andrew. It depends on what hypothesis you are trying to test...
 
Hope this helps,
 
Julian
 

 

> Date: Fri, 7 Nov 2008 01:13:52 -0800
> Subject: BSP, serially sampled data and lack of precision in the sampling dates
> From: ursu...@gmail.com
> To: beast...@googlegroups.com