search

trvrb / flux

Integrating influenza antigenic dynamics with molecular evolution
http://bedford.io/papers/bedford-flux/
12 stars 5 forks source link

Serum effects #3

Closed trvrb closed 11 years ago

trvrb commented 11 years ago
  • The improvement from inclusion of serum effects deserves comment and explanation if possible (are the effects due to the serum source)?
  • An interesting result is that estimating the serum effects improves the performance a great deal (more than including the phylogenetic data, in fact!). It suggests that there may be systematic differences between the sera. Can this be attributed to the serum itself (such as whether it is goat versus ferret), or to the particular lab or study? Could the authors correlate the estimated serum effects back to the originating source for that serum to check?
trvrb commented 11 years ago

We would suggest that a serum with a larger effect contains a more concentrated and active set of antibodies than a serum with a smaller effect. The increased number of antibodies could be due to experimental variation in serum extraction and processing or due to variation in immune response and timing between ferrets, i.e. some ferrets may mount an immediate and strong immune response, while others may mount a weaker response. This causes variation in overall strength of the serum. We try to control for the overall strength when looking at patterns of cross-reactivity through the use of 'serum effects'. Some differences in serum effects could be tracked down to particular studies, but this work is beyond the scope of the current study. Here, we've essentially treated serum effect as a nuisance parameter. We've revised the text to discuss this as well as describe why estimating serum effects improves model performance over fixing serum effects.