I have about ~3000 genomes of organisms and presence (recorded as 1)/absence (recorded as 0) data of ~300 genes in them. Among 3000 genomes, we have experimental data on 20 organisms, which were represented by 2 different phenotype groups (a, b). According to a previous study, the 300 genes I mentioned are believed to be associated with phenotypes 'a' and 'b'.

To predict the phenotypes of 2980 genomes with no phenotype information, I did hierarchical clustering of 3000 genomes based on the presence/absence of the 300 genes. I cut the resulting dendrogram at a certain height that allowed distinction of clusters that included either group 'a' or 'b', but not both. This gave me 120 clusters, including clusters with no group 'a' or 'b' organisms, clusters with at least one group 'a' organisms, and clusters with at least one group 'b' organisms. Among these 120 clusters, there are 15 clusters that had either group 'a' or 'b' organisms.

I am questioning which collections of genes are responsible for causing phenotypes 'a' or 'b'. To address this, I am thinking about assigning different names for the 15 clusters that contain group 'a' or 'b' organisms, and make a classification tree using these 15 clusters. Does this sound like an appropriate application of classification tree approach? If not, could there be a more sound way to address this question?



1 Answer 1


This isn't a cool strategy because dendrograms are very poor representations of the relationships between genomes, for example they use UPGMA style metrics fail to account for mutation rate variation between genomes and prone to long-branch attraction artefacts.

I would recommend doing whole genome alignments using e.g. muscle5 -super option and then performing a formal phylogenetic tree using e.g. iqtree or even just MEGAX (distance methods) and using bootstrapping to identify the robustness of clades comprising a and b.

If two clades are supported by bootstraps > 80% then your hypothesis looks fine. At present your downstream analysis could be an artefact of the dendrogram phylogeny.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.