Gene Cluster Phylogenetic Tree

Question

[biohacker_tobe]

Hello community :)

Question 1. I am comparing the usage of two clustering algorithms, MCL and antiSMASH over a set of predetermined genomes. I have done quite some exploration and found interesting results. However, I wanted to create a phylogenetic tree from the clusters obtained (all .gbk files), would this be possible? Any suggestions on which algorithms would be interesting to view or how to?

Question 2. On another note, I ran the BiG-SCAPE algorithm on my gene clusters to measure their phylogenetic relationship. This algorithm basically gives a measure of distance between gene clusters and is represented in a dataframe such as:

BGC1      BGC2     Distance
------------------------------ 
BGC31     BGC34     0.6
BGC34     BGC45     0.7
BGC34     BGC53     0.2
BGC53     BGC31     0.8

Would it be possible to construct a tree just based on this type of data? I have been able to create network visualizations from this data through Cytoscape but not possibly a tree. Any further suggestions for this particular example?

Thanks once again for your input :)

[juliuspasion1994]

Hey Tobe,

Any luck with this endeavor? I am about to get into the same type of direction with my research, and I am not able to use CORASON due to it using RAST. I'm trying to build a tree with distance similarity of BGCs.

[biohacker_tobe]

Yes i was, I can share my code but it's on R

[juliuspasion1994]

Great, would love to see it, do you have a GitHub?

Answer 1

[Fatima]

Hi,

I believe you can convert it to a distance matrix and use MEGA: https://www.researchgate.net/post/How_to_generate_the_phylogenetic_tree_if_I_have_distance_matrix_rather_than_alignment

Or you can use Neighbor joining algorithm on the data that you have. Since each node is a cluster of genes, I'm not sure if the phylogenetic tree that you get can truly represent the evolutionary history.

Answer 2

[Mensur Dlakic]

There are ways to construct trees from sequence distances. I have tried this one and it works well. Yet another one - which I haven't tried - is here.