glstott
commented
1 year ago Successfully ran HIV data through the pipeline under FastTree. Here are some quick comparisons.
Looking at the genetic distance threshold of 0.015:
- 3135 samples are found within 843 unique clusters size 2 or more: max cluster size: 60, mean: 8.12, standard deviation: 10.64
- 1324 samples are found within 164 unique clusters size 5 or more: max cluster size: 60, mean: 14.19, standard deviation: 12.90
- Just 8 patristic distance (PD) clusters contain elements from 2 or more clusters generated from genetic distances (GD). The rest take all samples from a single GD cluster each time. That is, just 1% of clusters ended up being a blend of clusters from the genetic distance method.
- 136 of the GD clusters have more than 1 PD cluster assigned. This is around 22% of the clusters generated through genetic distance method.
In summary, the two methods have similar results, but the patristic distance clustering method ends up breaking up some of the genetic distance clusters, leading to more smaller clusters. There are more connected samples, but fewer transmission chains of size 5 or more.
Repeat the analysis done in https://www.nature.com/articles/s41598-021-82673-8#Sec2 . In particular, steps 1 and 3 of the phylogenetic analysis are connected.