Reputation: 2119
There are various machine learning models (Claude, chatGPT, etc) which can be used to extract machine-readable information from images. Has anyone seen cases of successfully extracting Newick format data (or equivalent) from published images of phylogenetic trees (of which there are many on the internet)?
Upvotes: 1
Views: 80
Reputation: 2119
I just tried a test case with Claude 3.5 (sonnet) and it isn't too bad, although it does place some polytomies in places where it's visually clear that they should be bifurcations, for example, for the tree below
With the prompt "Please can you extract a newick-format string from the attached image of a phylogenetic tree", I get
(Capsaspora,((Monosiga,Salpingoeca),(Amphimedon,(((Acropora,Nematostella),(Clytia,Hydra)),(Drosophila,(Ciona,(Branchiostoma,Capitella))),(Polyplacotoma_mediterranea_H0,((Trichoplax_H17,(Trichoplax_adhaerens_H1,Trichoplax_H2)100)100,(Cladtertia_collaboinventa_H23,(H19,H24)100,(Hoilungia_hongkongensis_H13,Hoilungia_H4)100,(Hoilungia_H15,Hoilungia_H25)100)100)100)100)30)100)100);
Which corresponds to
Upvotes: 0